Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

An analysis of visitors' length of stay through noninvasive Bluetooth monitoring in the Louvre Museum

An analysis of visitors' length of stay through noninvasive Bluetooth monitoring in the Louvre... An  analysis  of  visitors’  length  of  stay  through  noninvasive  Bluetooth   monitoring  in  the  Louvre  Museum   a b a Yuji  Yoshimura ,  Anne  Krebs ,  Carlo  Ratti   a   SENSEable  City  Laboratory,  Massachusetts  Institute  of  Technology,  77   Massachusetts  Avenue,  Cambridge,  MA  02139,  USA;   b   Dominique-­‐Vivant  Denon  Research  Centre,  muse  du  Louvre,  75058,  Paris,  Cedex   01,  France;   Abstract:  Art  Museums  traditionally  employ  observations  and  surveys  to  enhance   their  knowledge  of  visitors’  behavior  and  experience.  However,  these  approaches   often   produce   spatially   and   temporally   limited   empirical   evidence   and   measurements.   Only   recently   has   the   ubiquity   of   digital   technologies   revolutionized   the   ability   to   collect   data   on   human   behavior.   Consequently,   the   greater  availability  of  large-­‐scale  datasets  based  on  quantifying  visitors’  behavior   provides   new   opportunities   to   apply   computational   and   comparative   analytical   techniques.  In  this  paper,  we  attempt  to  analyze  visitors’  behavior  in  the  Louvre   Museum   from   anonymized   longitudinal   datasets   collected   from   noninvasive   Bluetooth  sensors.  We  examine  visitors’  length  of  stay  in  the  museum  and  consider   this   relationship   with   occupation   density   around   artwork.   This   data   analysis   increases   the   knowledge   and   understanding   of   museum   professionals   related   to   the  experience  of  visitors.   Introduction   The   recent   development   of   emerging   technologies   and   their   rapid   diffusion   into   our  daily  life  has  caused  a  structural  change  in  human  behavior  analysis.  Indeed,   the   ubiquitous   presence   of   wired   and   wireless   sensors   in   contemporary   urban   environments   produces   an   empirical   record   of   individual   activities   at   detailed   levels.   In   addition,   to   the   ubiquity   of   this   technology,   computationally   advanced   computer   systems   make   accumulating   large   datasets   of   human   behavior   at   high   frequencies  possible-­‐sometimes  even  in  real  time.   Contrary  to  the  common  use  of  such  data-­‐collection  technology,  the  data  collection   of  visitors’  behaviors  in  large-­‐scale  art  museums  has  not  advanced  much  over  the   past  few  decades.  The  traditional  pencil-­‐and-­‐paper  based  tracking  method  is  still   widely   used   in   the   form   of   “timing   and   tracking”   [1].   Furthermore,   many   of   the   emerging  technologies  don’t  work  appropriately  in  the  museum  setting  for  several   reasons:  active  mobile  phone  tracking  with  GPS  [2]  doesn’t  work  inside  buildings   and  passive  mobile  phone  tracking  [3,4]  cannot  distinguish  visitors’  presence  and   movement   between   rooms   because   its   detection   range,   based   on   the   antenna’s   coverage,  is  too  large.  Video  camera  based  tracking  technologies  are  useful,  but  the   substantially  higher  cost  do  not  allow  for  the  necessary  infrastructure  in  a  museum   environment.   RFID   [5],   ultra   wideband   [6]   and   mobile   phone   centered   wifi   tracking   method   are   promising,   but   they   require   visitors   to   be   equipped   with   certain   devices   or   to   download   the   proper   application   in   advance.   This     1   participatory  process  prevents  us  from  generating  large-­‐scale  datasets,  which  are   necessary  for  relevant  analysis. For   this   purpose,   this   paper   employs   a   Bluetooth   detection   technique   [7-­‐10].   Bluetooth   detection   is   unobtrusive,   making   use   of   the   visitors’   digital   footprint   they   unconsciously   leave   behind   [11].   Furthermore,   these   datasets   can   be   anonymized  in  order  to  protect  the  users’  personal  data.  As  a  result,  this  technique   enables  us  to  generate  large-­‐scale  datasets.  Additionally,  we  can  expect  datasets  to   be  obtained  without  any  behavioral  bias.  Participants  might  adapt  their  behaviors   if   they   become   conscious   of   being   observed.   Furthermore,   we   can   perform   data   collection  for  a  longer  period  than  just  one  day  or  a  few  days,  which  is  a  typical   data   collection   technique,   to   form   a   hypothetical   visitor,   who   can   represent   a   whole  population. Thus,  our  methodology  is  to  analyze  “real”  and  large-­‐scale  empirical  data,  which  is   contrary  to  that  of  traditionally  employed  manual-­‐based  methods.  This  proposed   method   sheds   light   on   unknown   aspects   of   visitors’   behaviors   in   terms   of   their   length  of  stay  and  the  influence  of  the  crowd  over  a  visitor.   Bluetooth  tracking  system  in  the  Louvre  Museum   Bluetooth  detection  systems  are  widely  used  to  track  people  inside  buildings  [10],   in   urban   settings   [9],   and   to   generate   traffic   information   based   on   detecting   vehicles   [12].   Bluetooth   detection   systems   work   as   follows:   once   a   Bluetooth-­‐ activated  mobile  device  enters  the  detectable  area,  the  sensor  continues  to  receive   the   emitted   signal   from   the   mobile   device   until   the   signal   disappears.   Thus,   the   sensor  registers  the  time  at  which  the  signal  of  a  mobile  device  appears,  also  called   check-­‐in  time.  Afterward,  when  the  signal  of  a  mobile  device  disappears,  the  sensor   records   the   check-­‐out   time.   Then   the   difference   between   each   mobile   device’s   check-­‐in  and  check-­‐out  time  can  be  calculated,  which  defines  the  length  of  stay  at   the  node.  Similarly,  by  looking  at  the  first  check-­‐in  time  and  the  last  check-­‐out  time   over  all  nodes,  it  is  possible  to  calculate  how  long  a  visitor  stays  in  the  Denon  wing   of   the   Louvre   museum.   Such   a   series   of   check-­‐in   and   check-­‐out   time   data   registered   by   all   the   installed   sensors   makes   it   possible   to   construct   a   visitor’s   trajectory   through   the   Denon   wing   of   the   museum,   including   their   travel   time   between   nodes.   All   of   this   information   can   be   collected   without   invading   visitor’   privacy  because  Secure  Hash  Algorithm  (SHA)  encryption  [13]  is  applied  to  each   sensors  by  converting  each  device’s  MACID  in  to  a  unique  identifier  [7].   2     Figure   1.  Location  of  eight  sensors  (E,  D,  V,  B,  S,  G,  C,  P)  indicating  their  approximating   sensing  range.  The  arrow  and  its  width  represents  visitors’  flow  between  nodes.   Eight   Bluetooth   sensors   have   been   deployed   throughout   the   Denon   wing   of   the   museum,   covering   key   places   to   capture   visitors’   behavior.   These   sensors   were   placed   along   one   of   the   busiest   trails   identified   by   Louvre   Museum,   which   lead   visitors   from   the   entrance   to   the   Venus   de   Milo.   The   sensors   are   placed   at   the   Entrance   Hall   (E),   Gallery   Daru   (D),   Venus   de   Milo   (V),   Salle   des   Caryatides   (C),   Sphinx  (P),  Great  Gallery  (B),  Victory  of  Samothrace  (S),  and  Salle  des  Verres  (G).   The   data   collection   was   performed   at   different   periods   by   a   different   number   of   sensors   during   a   five-­‐month   period   from   April   2010   to   August   2010.   After   data   cleanup   and   data   processing,   which   adjusted   the   data   to   remove   any   inconsistencies,  81,498  unique  devices  were  selected  to  be  analyzed  for  this  paper.   By   comparing   the   number   of   detected   mobile   devices   and   ticket   sales,   we   found   that,  on  average,  8.2%  of  visitors  had  activated  Bluetooth  on  their  mobile  phone.   Additionally,   we   previously   uncovered   visitors’   transition   probability   between   nodes  and  their  mobility  patterns  considering  their  length  of  stay  in  the  museum   (see  Figure  1).  Based  on  this  previous  study,  this  paper  analyzes  visitors’  length  of     3   stay   in   each   node   as   well   as   the   entire   duration   of   their   visit,   which   is   how   this   study  differs  from  the  previous  one.   Analyzing  three  factors  regarding  length-­‐of-­‐stay  in  the  museum   We  analyze  three  different  factors  related  to  visitors’  length  of  stay  in  the  museum.   The  first  factor  relates  to  entry  time,  which  can  be  used  to  assess  the  distribution   of   visitors’   length   of   stay   in   the   museum   depending   on   when   they   entered   the   museum.  The  second  factor  provides  visitors’  length  of  stay  at  each  specific  node.   The  third  factor  is  the  relationship  between   the   length   of   stay  at   a   specific   node   and  the  number  of  visitors  around  the  node  (i.e.,  density).   The  length  of  stay  in  the  museum   (a)                                                                                                                                                                                      (b)   Figure  2.  (a)  The  number  of  unique  node  visited  against  the  length  of  stay  (b)  The  number   of  total  visited  nodes  against  the  length  of  stay.   Figure  2  (a)  shows  that  the  median  length  of  stay  is  very  similar  across  all  amounts   of   unique   visited   nodes.   We   used   a   non-­‐parametric   correlation   analysis   (Spearman’s   rank   correlation   coefficient)   because   the   variables   do   not   seem   to   follow  a  normal  distribution.  We  also  include  a  series  of  boxplots  to  better  explain   the  relationship  between  variables.  A  very  low  correlation  value  (ρ = 0.073, p<2.2e- 16) suggests that there is no relationship between these two variables. That is, the number of unique nodes visited seems to be independent of the length of stay. On the other hand, Figure 2 (b) shows a different relationship between the length of stay and the total amount of visited nodes. The correlation coefficient (ρ = 0.186, p-value<2.2e- 16) suggests a weak association between the two variables. Our interpretation is that when people stay for a longer period inside the museum, they tend to limit the number of visited nodes and prefer to dedicate more time to exploring those they visit thoroughly (sometimes visiting them more than once), instead of visiting a higher number of different nodes. We   also   examine   a   distribution   of   an   average   of   visitors’   length   of   stay   in   the   museum   classified   by   the   hour   of   the   day   they   visit   the   museum.   This   analysis   reveals   whether   or   not   visitors’   entry   time   affects   their   length   of   stay   at   the   museum.  We  divide  a  day  of  the  week  into  two  groups  depending  on  the  closing     4   time  of  the  museum.  The  first  group  consists  of  Monday,  Thursday,  Saturday  and   Sunday,  when  the  door  closes  at  18:00  and  the  second  group  includes  Wednesday   and  Friday,  when  the  museum  closes  at  21:45.   visitor's length of stay vs entry time to the museum visitor's length of stay vs entry time to the museum 4:48:00 04:48:00 03:36:00 3:36:00 02:24:00 2:24:00 2"may& 3"may& 6"may& 8"may& 01:12:00 1:12:00 30"abr& 5"may& 7"may& 30"jun& 2"jul& 9"may& 1"jul& 3"jul& 4"jul& 5"jul& 8"jul& 12"ago& 14"ago& 7"jul& 9"jul& 13"ago& 18"ago& average& 15"ago& 16"ago& 19"ago& average& 00:00:00 0:00:00 Visitor's entry time to the museum Visitor's entry time to the museum (a)                                                                                                                                                                                          (b)   Figure   3.   (a)   The   distribution   of   the   average   stay   time   by   visiting   hour   on   Monday,   Thursday,  Saturday  and  Sunday.  (b)  The  distribution  of  the  average  stay  time  by  visiting   hour  on  Wednesday  and  Friday.   Figure   3   (a)   presents   a   clear   tendency   that   the   length   of   stay   time   decreases   toward  the  closing  hours  of  the  museum.  The  earlier  a  visitor  enters  the  museum,   the  longer  that  visitor  tends  to  stay  in  the  museum.  This  indicates  that  the  closing   time  of  the  museum  works  as  a  constraint  to  limit  the  length  of  visitors’  time  in  the   museum.  Thus,  visitors  seem  to  stay  longer  within  their  limited  available  time  in   order  to  maximize  their  benefits.   Conversely,   the   results   on   Wednesday   and   Friday   show   a   different   tendency   of   visitors   (see   Figure   3   (b)).   As   with   the   previous   analysis,   we   can   observe   a   tendency  that  the  length  of  stay  at  the  museum  decreases  with  the  advance  of  the   time.  However,  the  decrease  in  length  of  stay  is  slightly  mitigated  in  the  middle  of   the  day.  Just  after  the  opening  of  the  museum  (i.e.,  10:00-­‐11:00),  the  length  of  stay   is  greatest,  but  in  the  late  afternoon  (i.e.,  17:00-­‐18:00),  the  length  of  stay  increases   slightly.  This  data  makes  us  suppose  there  might  be  two  kinds  of  visitors.  That  is,   while  some  intend  to  maximize  their  utility  (e.g.,  staying  time)  within  the  limited   time   the   museum   is   open   by   visiting   earlier,   others   try   to   take   advantage   of   the   longer  hours  and  wait  until  the  evening  to  visit.   All  of  these  analyses  and  results  indicate  that  the  time  visitors  enter  can  be  used  to   predict   visitors’   length   of   stay   in   the   museum,   but   their   length   of   stay   in   the   museum  doesn’t  have  any  correlation  with  the  number  of  visited  nodes  over  the   course   of   their   visit.   While   the   longer   length   of   stay   slightly   suggests   a   larger   number   of   visited   nodes,   their   relationship   is   not   significant.   Such   a   result   is   particularly  useful  for  the  daily  management  of  peak  periods  and  rush  hours  by  the   staff.   The  length  of  stay  at  each  node     5   9:00-10:00 10:00-11:00 11:00-12:00 12:00-13:00 13:00-14:00 14:00-15:00 15:00-16:00 16:00-17:00 Average length of stay in the museum Average length of stay in the museum 9:00-10:00 10:00-11:00 11:00-12:00 12:00-13:00 13:00-14:00 14:00-15:00 15:00-16:00 16:00-17:00 17:00-18:00 18:00-19:00 19:00-20:00 20:00-21:00   We   examine   how   long   a   visitor   stays   at   each   node   in   a   disaggregated   way   and   uncover  the  feature  of  each  node  by  comparing  the  analysis  of  an  accumulation  of   individual  visitor’s  behavior.   Our  preliminary  analysis  shows  that  node  E  and  node  S  have  a  much  longer  length   of   stay   than   the   other   nodes.   The   median   length   of   stay   at   nodes   E   and   S   are   00:16:29  and  00:19:03,  respectively,  while  the  average  length  of  stay  at  the  other   nodes   is   00:03:14.   Node   E   is   situated   in   the   ticket   sale   desk,   indicating,   unsurprisingly,  that  visitors  may  wait  for  a  long  queue  in  order  to  purchase  a  ticket.   Winged   Victory   of   Samothrace,   where   sensor   S   is   located,   is   one   of   the   most   attractive  exhibits  in  the  museum-­‐together  with  the  huge  staircase  in  front  of  the   exhibit.  Besides,  many  visitors  use  the  stairs  as  a  improvised  chair  to  take  a  rest   during  their  visit.  Those  two  factors  make  node  E  and  node  S  different  from  other   nodes.   We   speculate   that   the   unique   uses   for   these   spaces   results   in   their   much   longer  length  of  stay,  and  this  fact  motivates  us  to  exclude  these  nodes  from  our   following  analysis.   (a)                                                                                                                                                                            (b)   Figure   4.  (a)  The  boxplot  of  the  length  of  stay  in  each  node.  (b)  The  comparative  visitor   decay  curves.   Figure  4  (a)  presents  the  boxplot  of  visitors’  length  of  stay  at  each  node.  We  can   observe   that   two   groups   exist:   node   V   (Venus   de   Milo),   node   C   (Salle   des   Caryatides),  and  node  B  (Great  Gallery)  experience  a  longer  stay;  node  D  (Gallery   Daru),  node  P  (Sphinx),  and  node  G  (Salle  des  Verres)  experience  a  shorter  stay.   While  the  median  length  of  stay  for  the  former  group  is  03:02,  the  median  for  the   latter  is  00:44.  Among  these  nodes  the  visitors’  length  of  stay  and  its  range  in  node   P  is  much  shorter  than  that  of  other  nodes,  and  as  a  result,  98%  of  all  visitors  to   node  P  have  a  length  of  stay  between  19  seconds  and  113  second.   Conversely,  Figure  4  (b)  shows  the  comparative  “visitor  survival  curve”.  We  plot   the   percentage   of   visitors   at   given   times.   This   plot   is   frequently   used   in   visitor   studies   in   order   to   analyze   the   length   of   stay,   when   half   of  the   visitors   leave,   an   exhibit  or  room  (Bicknell,  1995).  We  can  observe  that  the  length  of  stay  is  largely   varied  among  nodes:  00:20-­‐00:30  for  node  D,  01:30-­‐01:40  for  node  V,  01:30-­‐01:40     6   for  node  C,  00:00-­‐  00:10  for  node  P,  02:10-­‐02:20  for  node  B,  and  00:00-­‐00:10  for   node   G.   However,   most   of   the   nodes   experience   a   length   of   stay   within   a   few   minutes.  Again,  we  can  observe  that  two  kinds  of  nodes  exist,  which  we  can  classify   as  shorter  and  longer  stay  type  nodes.  Within  the  former  group,  half  of  visitors  left   the  node  by  10-­‐20  seconds  in  the  cases  of  nodes  G  and  P.  In  the  case  of  the  node  D,   this  duration  is  little  bit  longer,  but  half  of  visitors  stayed  for  just  20-­‐30  seconds.   Relationship  between  visitors’  length  of  stay  and  density   The   perspective   on   length   of   stay,   however,   greatly   changes   when   examining   visitors’   duration   of   stay   in   relationship   to   the   degree   of   the   occupancy   of   each   node.  Figure  5  shows  the  relationship  between  each  node’s  occupancy  normalized   by  the  maximum  number  of  visitors  in  the  area  (x-­‐axis)  and  the  average  duration   of  stay  expressed  by  seconds  (y-­‐axis).  As  we  can  see,  a  clear  tendency  exists  among   all  data.  The  average  duration  of  stay  first  goes  up  with  the  room  occupancy  from   point   W   to   point   X,   then   stays   around   the   maximum   on   some   occupancy   level   interval  (point  X  to  Z).  After  that,  length  of  stay  drops  down  as  the  occupancy  level   starts  to  exceed  a  certain  threshold  (point  Z).   Figure  5.  Distribution  of  the  normalized  occupancy  vs  the  length  of  stay.   Table  1.  The  threshold  of  the  normalized  occupancy  in  each  node.                                                        Point  X                                                                Point  Y                                                                        Point  Z   Node  D                                                183sec  (0.235)                                  204sec  (0.498)                                          188sec  (0.636)                       Node  V                                                289sec  (0.235)                                  315sec  (0.368)                                          271sec  (0.781)                   Node  C                                                225sec  (0.221)                                  249sec  (0.352)                                          187sec  (0.753)                       Node  P     7                                                    94sec  (0.214)                                    118sec  (0.5)                                                        79sec  (0.764)                   Node  B                                                353sec  (0.238)                                  375sec  (0.504)                                          293sec  (0.772)               Node  G                                                167sec  (0.267)                                  208sec  (0.351)                                          145sec  (0.657)                     The  length  of  stay  for  point  W  is  extremely  short  because  it  contains  visitors,  who   just   pass   by   the   area   rather   than   stay   there,   which   we   demonstrated   in   the   previous   sections.   When   the   occupancy   level   increases   from   point   W   to   X,   the   number   of   visitors   who   tend   to   stay   longer   also   increases.   Point   X   can   be   considered  the  equilibrium  between  an  ideal  length  of  stay  for  a  visitor  when  he  or   she  is  free  to  stay  as  long  as  he  or  she  wants.  This  is  because  visitors  are  free  to   look  at  the  artwork  between  point  W  and  X  without  any  obstacles  due  to  the  low   density  of  other  visitors.  Conversely,  from  point  X  onward,  the  average  length  of   stay   remains   almost   flat   until   point   Z,   at   which   point   visitors’   length   of   stay   drastically  starts  to  decrease.  We  speculate  that  this  is  because  the  exceeding  high-­‐ density   of   other   visitors   may   affect   a   visitor’s   comfort,   resulting   in   a   desire   to   escape  the  crowd.   The  length  of  stay  at  point  X  is  largely  varied  depending  on  the  node.  Node  B  at   point   X   shows   the   longest   length   of   stay   (353  seconds)  with  a   0.238   normalized   occupancy  level  among  the  other  nodes.  Then,  node  B  remains  a  flat  until  point  Z.   Conversely,  node  P  at  point  X  represents  the  shortest  length  of  stay  (94  seconds)   with   a   0.214   normalized   occupancy   level   among   the   other   nodes.   Although   the   occupancy  level  is  similar  between  those  two  nodes,  the  former’s  length  of  stay  is   almost   four   times   longer   than   the   latter’s.   However,   the   normalized   occupancy   level   of   point   X   for   node   B   is   higher   than   that   of   node   P   (i.e.,   0.238   vs.   0.214,   respectively).   Additionally,   these   nodes   correspond   with   the   maximum   and   minimum  lengths  of  stay  at  point  X,  although  the  occupancy  level  of  both  nodes  is   quite   similar:   both   of   them   are   around   0.50.   Furthermore,   the   length   of   stay   at   node  P  starts  to  decrease  earlier  than  that  of  node  B.     Regarding  the  relationship  between  node  V  and  node  B,  although  point  X  for  both   of   them   shows   a   similar   length   of   stay   (i.e.,   0.235   vs.   0.238,   respectively),   the   maximum  length  of  stay  of  node  V  has  a  much  lower  occupation  level  than  those  of   nodes  P  and  B  (i.e.,  0.368  for  node  V).  Additionally,  node  V  has  the  highest  density   for  point  Z  (0.781)  with  the  second  longest  length  of  stay  (271sec),  whereas  the   longest  length  of  stay  is  at  node  B  (293seconds).   All   of   these   facts   indicate   that   the   artworks   at   node   B   attract   and   hold   visitors   much  more  strongly  than  the  artworks  of  node  V  and  P.  Node  B  seems  to  inspire   visitors   to   stay   longer,   even   with   higher   occupation   density,   while   node   V   and   P   seem  to  cause  visitors  to  stay  for  shorter  durations  when  experiencing  the  same   density.   Also,   the   data   shows   that   the   average   length   of   stay   of   a   visitor   and   the   occupation   level   of   a   node   form   a   clear   pattern.   The   crowd   density   around   the   artwork  largely  affects  a  visitor’s  length  of  stay  either  positively  or  negatively,  and   the  type  of  effect  is  largely  dependent  on  the  node.     8   We  speculate  that,  although  it  also  depends  on  the  nature  of  the  galleries  and  the   type  of  visitors,  up  to  certain  occupancy  limits  visitors  are  actually  attracted  by  the   crowd:  however,  once  the  crowd’s  size  reaches  a  certain  level,  visitors  will  try  to   avoid  the  node.  This  indicates  that  we  may  use  point  Z  as  a  threshold  to  distinguish   visitors’  level  of  comfort,  which  enables  us  to  manage  the  environment  in  order  to   avoid  exceeding  this  density  threshold  in  the  environment.  Also,  we  may  consider   some  characteristic  points  such  as  point  X  and  Y  to  enabling  to  manage  different   types  of  environments/crowds  inside  the  galleries.   Conclusions  and  discussion   The   limitation   of   our   proposed   method   is   as   follows:   first,   the   sensor   can   detect   only  mobile  devices  in  which  the  Bluetooth  function  is  activated.  This  indicates  the   representativeness   of   the   sample   may   have   a   strong   bias   toward   certain   groups,   such   as   the   upper   class,   higher-­‐educated   people,   and   the   younger   generation   rather   than   seniors   and   children.   Moreover,   sensors   are   not   capable   of   distinguishing   individual   visit   or   group   visits.   The   representativeness   of   our   sample   can   be   calculated   by   comparing   the   sample   obtained   from   Bluetooth   detection  and  the  head  count  by  hand  over  a  shorter  period  (e.g,  a  few  hours  or   few  days).  In  fact  we  applied  a  systematic  comparison  over  a  longer  period  (one   month)   with   the   number   of   ticket   sales   in   the   desk   where   a   Bluetooth   sensor   is   installed.  Second,  the  sensor  enables  us  to  detect  visitors’  presence  in  the  specific   area,  which  is  determined  by  the  radius  of  each  sensor’s  detection  range,  but  such   sensors   cannot   specify   whether   visitors   are   actually   looking   at   the   artwork   or   if   they  are  simply  in  the  area.  This  shortcoming  is  strongly  related  to  the  following   limitation:  “while  time  is  a  necessary  condition  for  learning,  time  in  a  gallery  does   not   correspond   directly   to   time   spent   attending   to   exhibitions”   [14].   Third,   Bluetooth  detection  techniques  cannot  disclose  the  visitors’  motivations  and  inner   thoughts  in  any  way:  this  method  merely  identifies  their  presence  and  the  precise   length  of  their  stay.  Finally,  sensors  are  not  capable  to  collect  socio-­‐demographics   (i.e.,  origin,  age,  gender,  profession)  as  for  other  traditional  behavioral  variables. As   a   backdrop   to   this   situation,   the   proposed   new   approach   enables   us   to   shed   light  on  some  unknown  aspects  of  visitors’  behavior.  Our  proposed  system  and  the   results   we   obtained   through   an   adequate   statistical   analysis   can   work   as   a   new   tool   for   the   museum   management.   Our   system   effectively   captures   an   individual   visitor’s  length  of  stay  in  different  ways  and  enables  us  to  store  this  information  as   a  large-­‐scale  dataset.  Such  a  method  appears  to  be  particularly  useful  for  medium   or  small  size  museums  where  it  could  be  possible  to  install  a  complete  system  of   sensors.   This   method   results   in   a   dataset   that   is   very   different   from   the   conventional  dataset  in  two  ways:  the  quantitative  size  of  the  dataset  and  the  finer-­‐ grain  detection  of  visitor  behavior  both  spatially  and  temporally.   All  of  our  findings  are  helpful  for  the  management  of  visitor  flow  in  order  to  reduce   congestion  at  specific  areas  and  around  specific  pieces  of  artwork.  It  is  also  useful   in  order  to  bring  to  light  less  visited  or  less  “attractive”  artworks/rooms  inside  the   museum,  in  order  to  propose  interpretation  tools  and/  or  walking  tours  capable  to   increase  the  value  of  such  “neglected”  spaces.  Additionally,  our  data  suggest  that   visitor  behavior  is  based  on  some  patterns,  which  make  it  possible  to  foresee  their     9   future   movement   in   a   dynamic   way.   Also,   these   data   are   significantly   useful   for   designing  the  spatial  arrangement  (e.g.,  changes  in  the  layout  of  exhibits,  facilities   and  advertisements),  depending  on  visitor  activities  and  use  of  space.  Finally,  our   findings   indicate   that   efficient   and   effective   congestion   management   of   the   museum  can  be  realized  by  limiting  the  number  of  visitors  that  are  able  to  enter   based  on  the  time  of  the  day.   References   1. Yalowitz, S. S. & Bronnenkant, Kerry. (2009) ”Timing and Tracking: Unlocking Visitor Behavior” Visitor Studies 12(1): 47-64. 2. Shoval N, McKercher B, Birenboim A, Ng E, 2013, “The application of a sequence alignment method to the creation of typologies of tourist activity in time and space” Environment and Planning B: Planning and Design advance online publication, doi:10.1068/b38065 3. González M C, Hidalgo C A, Barabási A L, 2008, “Understanding individual human mobility patterns” Nature 453 779-782 4. Ratti C, Pulselli R, Williams S, Frenchman D, 2006, “Mobile Landscapes: using location data from cell phones for urban analysis” Environment and Planning B: Planning and Design 33(5) 727-748 5. Kanda T, Shiomi M, Perrin L, Nomura T, Ishiguro H, Hagita N, 2007, “Analysis of people trajectories with ubiquitous sensors in a science museum” Proceedings 2007 IEEE International Conference on Robotics and Automation (ICRA’07) 4846-4853 6. Tröndle, M, Greenwood, S, Kirchberg, V, Tschacher, W, 2014, “An Integrative and Comprehensive Methodology for Studying Aesthetic Experience in the Field: Merging Movement Tracking, Physiology, and Psychological Data“, Environment and Behavior, 46 (1) pp102-135. 7. Yoshimura Y, Sobolevsky S, Ratti C, Girardin F, Carrascal J P, Blat J, Sinatra R, 2014, “An analysis of visitors’ behaviour in The Louvre Museum: a study using Bluetooth data” Environment and Planning B: Planning and Design 41 (6) 1113-1131 8. Kostakos V, O'Neill E, Penn A, Roussos G, Papadongonas D, 2010, “Brief encounters: sensing, modelling and visualizing urban mobility and copresence networks” ACM Transactions on Computer Human Interaction 17(1) 1-38 9. Versichele M, Neutens T, Delafontaine M, Van de Weghe N, 2011, “The use of Bluetooth for analysing spatiotemporal dynamics of human movement at mass events: a case study of the Ghent festivities” Applied Geography 32 208-220 10. Delafontaine M, Versichele M, Neutens T, Van de Weghe N, 2012, “Analysing spatiotemporal sequences in Bluetooth tracking data” Applied Geography 34 659-668 11. Mayer-Schönberger V, Cukier K, 2013, Big Data: A Revolution That Will Transform How We Live, Work and Think (John Murray, London) 12. Barceló J, Montero L, Marqués L, Carmona C, 2010, “Travel Time Forecasting and Dynamic Origin-Destination Estimation for Freeways Based on Bluetooth Traffic Monitoring” Transportation Research Record: Journal of the Transportation Research Board 2175: 19-27 13. Stallings W, 2011, Cryptography and Network Security: Principles and Practice, 5th Edition (Prentice Hall, Boston MA) 14. Hein G, 1998, Learning in the Museum (Routledge, London)   10   http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Physics arXiv (Cornell University)

An analysis of visitors' length of stay through noninvasive Bluetooth monitoring in the Louvre Museum

Physics , Volume 2020 (1605) – Apr 30, 2016

Loading next page...
 
/lp/arxiv-cornell-university/an-analysis-of-visitors-length-of-stay-through-noninvasive-bluetooth-00gwmeB0Rf
ISSN
1536-1268
eISSN
ARCH-3341
DOI
10.1109/MPRV.2017.33
Publisher site
See Article on Publisher Site

Abstract

An  analysis  of  visitors’  length  of  stay  through  noninvasive  Bluetooth   monitoring  in  the  Louvre  Museum   a b a Yuji  Yoshimura ,  Anne  Krebs ,  Carlo  Ratti   a   SENSEable  City  Laboratory,  Massachusetts  Institute  of  Technology,  77   Massachusetts  Avenue,  Cambridge,  MA  02139,  USA;   b   Dominique-­‐Vivant  Denon  Research  Centre,  muse  du  Louvre,  75058,  Paris,  Cedex   01,  France;   Abstract:  Art  Museums  traditionally  employ  observations  and  surveys  to  enhance   their  knowledge  of  visitors’  behavior  and  experience.  However,  these  approaches   often   produce   spatially   and   temporally   limited   empirical   evidence   and   measurements.   Only   recently   has   the   ubiquity   of   digital   technologies   revolutionized   the   ability   to   collect   data   on   human   behavior.   Consequently,   the   greater  availability  of  large-­‐scale  datasets  based  on  quantifying  visitors’  behavior   provides   new   opportunities   to   apply   computational   and   comparative   analytical   techniques.  In  this  paper,  we  attempt  to  analyze  visitors’  behavior  in  the  Louvre   Museum   from   anonymized   longitudinal   datasets   collected   from   noninvasive   Bluetooth  sensors.  We  examine  visitors’  length  of  stay  in  the  museum  and  consider   this   relationship   with   occupation   density   around   artwork.   This   data   analysis   increases   the   knowledge   and   understanding   of   museum   professionals   related   to   the  experience  of  visitors.   Introduction   The   recent   development   of   emerging   technologies   and   their   rapid   diffusion   into   our  daily  life  has  caused  a  structural  change  in  human  behavior  analysis.  Indeed,   the   ubiquitous   presence   of   wired   and   wireless   sensors   in   contemporary   urban   environments   produces   an   empirical   record   of   individual   activities   at   detailed   levels.   In   addition,   to   the   ubiquity   of   this   technology,   computationally   advanced   computer   systems   make   accumulating   large   datasets   of   human   behavior   at   high   frequencies  possible-­‐sometimes  even  in  real  time.   Contrary  to  the  common  use  of  such  data-­‐collection  technology,  the  data  collection   of  visitors’  behaviors  in  large-­‐scale  art  museums  has  not  advanced  much  over  the   past  few  decades.  The  traditional  pencil-­‐and-­‐paper  based  tracking  method  is  still   widely   used   in   the   form   of   “timing   and   tracking”   [1].   Furthermore,   many   of   the   emerging  technologies  don’t  work  appropriately  in  the  museum  setting  for  several   reasons:  active  mobile  phone  tracking  with  GPS  [2]  doesn’t  work  inside  buildings   and  passive  mobile  phone  tracking  [3,4]  cannot  distinguish  visitors’  presence  and   movement   between   rooms   because   its   detection   range,   based   on   the   antenna’s   coverage,  is  too  large.  Video  camera  based  tracking  technologies  are  useful,  but  the   substantially  higher  cost  do  not  allow  for  the  necessary  infrastructure  in  a  museum   environment.   RFID   [5],   ultra   wideband   [6]   and   mobile   phone   centered   wifi   tracking   method   are   promising,   but   they   require   visitors   to   be   equipped   with   certain   devices   or   to   download   the   proper   application   in   advance.   This     1   participatory  process  prevents  us  from  generating  large-­‐scale  datasets,  which  are   necessary  for  relevant  analysis. For   this   purpose,   this   paper   employs   a   Bluetooth   detection   technique   [7-­‐10].   Bluetooth   detection   is   unobtrusive,   making   use   of   the   visitors’   digital   footprint   they   unconsciously   leave   behind   [11].   Furthermore,   these   datasets   can   be   anonymized  in  order  to  protect  the  users’  personal  data.  As  a  result,  this  technique   enables  us  to  generate  large-­‐scale  datasets.  Additionally,  we  can  expect  datasets  to   be  obtained  without  any  behavioral  bias.  Participants  might  adapt  their  behaviors   if   they   become   conscious   of   being   observed.   Furthermore,   we   can   perform   data   collection  for  a  longer  period  than  just  one  day  or  a  few  days,  which  is  a  typical   data   collection   technique,   to   form   a   hypothetical   visitor,   who   can   represent   a   whole  population. Thus,  our  methodology  is  to  analyze  “real”  and  large-­‐scale  empirical  data,  which  is   contrary  to  that  of  traditionally  employed  manual-­‐based  methods.  This  proposed   method   sheds   light   on   unknown   aspects   of   visitors’   behaviors   in   terms   of   their   length  of  stay  and  the  influence  of  the  crowd  over  a  visitor.   Bluetooth  tracking  system  in  the  Louvre  Museum   Bluetooth  detection  systems  are  widely  used  to  track  people  inside  buildings  [10],   in   urban   settings   [9],   and   to   generate   traffic   information   based   on   detecting   vehicles   [12].   Bluetooth   detection   systems   work   as   follows:   once   a   Bluetooth-­‐ activated  mobile  device  enters  the  detectable  area,  the  sensor  continues  to  receive   the   emitted   signal   from   the   mobile   device   until   the   signal   disappears.   Thus,   the   sensor  registers  the  time  at  which  the  signal  of  a  mobile  device  appears,  also  called   check-­‐in  time.  Afterward,  when  the  signal  of  a  mobile  device  disappears,  the  sensor   records   the   check-­‐out   time.   Then   the   difference   between   each   mobile   device’s   check-­‐in  and  check-­‐out  time  can  be  calculated,  which  defines  the  length  of  stay  at   the  node.  Similarly,  by  looking  at  the  first  check-­‐in  time  and  the  last  check-­‐out  time   over  all  nodes,  it  is  possible  to  calculate  how  long  a  visitor  stays  in  the  Denon  wing   of   the   Louvre   museum.   Such   a   series   of   check-­‐in   and   check-­‐out   time   data   registered   by   all   the   installed   sensors   makes   it   possible   to   construct   a   visitor’s   trajectory   through   the   Denon   wing   of   the   museum,   including   their   travel   time   between   nodes.   All   of   this   information   can   be   collected   without   invading   visitor’   privacy  because  Secure  Hash  Algorithm  (SHA)  encryption  [13]  is  applied  to  each   sensors  by  converting  each  device’s  MACID  in  to  a  unique  identifier  [7].   2     Figure   1.  Location  of  eight  sensors  (E,  D,  V,  B,  S,  G,  C,  P)  indicating  their  approximating   sensing  range.  The  arrow  and  its  width  represents  visitors’  flow  between  nodes.   Eight   Bluetooth   sensors   have   been   deployed   throughout   the   Denon   wing   of   the   museum,   covering   key   places   to   capture   visitors’   behavior.   These   sensors   were   placed   along   one   of   the   busiest   trails   identified   by   Louvre   Museum,   which   lead   visitors   from   the   entrance   to   the   Venus   de   Milo.   The   sensors   are   placed   at   the   Entrance   Hall   (E),   Gallery   Daru   (D),   Venus   de   Milo   (V),   Salle   des   Caryatides   (C),   Sphinx  (P),  Great  Gallery  (B),  Victory  of  Samothrace  (S),  and  Salle  des  Verres  (G).   The   data   collection   was   performed   at   different   periods   by   a   different   number   of   sensors   during   a   five-­‐month   period   from   April   2010   to   August   2010.   After   data   cleanup   and   data   processing,   which   adjusted   the   data   to   remove   any   inconsistencies,  81,498  unique  devices  were  selected  to  be  analyzed  for  this  paper.   By   comparing   the   number   of   detected   mobile   devices   and   ticket   sales,   we   found   that,  on  average,  8.2%  of  visitors  had  activated  Bluetooth  on  their  mobile  phone.   Additionally,   we   previously   uncovered   visitors’   transition   probability   between   nodes  and  their  mobility  patterns  considering  their  length  of  stay  in  the  museum   (see  Figure  1).  Based  on  this  previous  study,  this  paper  analyzes  visitors’  length  of     3   stay   in   each   node   as   well   as   the   entire   duration   of   their   visit,   which   is   how   this   study  differs  from  the  previous  one.   Analyzing  three  factors  regarding  length-­‐of-­‐stay  in  the  museum   We  analyze  three  different  factors  related  to  visitors’  length  of  stay  in  the  museum.   The  first  factor  relates  to  entry  time,  which  can  be  used  to  assess  the  distribution   of   visitors’   length   of   stay   in   the   museum   depending   on   when   they   entered   the   museum.  The  second  factor  provides  visitors’  length  of  stay  at  each  specific  node.   The  third  factor  is  the  relationship  between   the   length   of   stay  at   a   specific   node   and  the  number  of  visitors  around  the  node  (i.e.,  density).   The  length  of  stay  in  the  museum   (a)                                                                                                                                                                                      (b)   Figure  2.  (a)  The  number  of  unique  node  visited  against  the  length  of  stay  (b)  The  number   of  total  visited  nodes  against  the  length  of  stay.   Figure  2  (a)  shows  that  the  median  length  of  stay  is  very  similar  across  all  amounts   of   unique   visited   nodes.   We   used   a   non-­‐parametric   correlation   analysis   (Spearman’s   rank   correlation   coefficient)   because   the   variables   do   not   seem   to   follow  a  normal  distribution.  We  also  include  a  series  of  boxplots  to  better  explain   the  relationship  between  variables.  A  very  low  correlation  value  (ρ = 0.073, p<2.2e- 16) suggests that there is no relationship between these two variables. That is, the number of unique nodes visited seems to be independent of the length of stay. On the other hand, Figure 2 (b) shows a different relationship between the length of stay and the total amount of visited nodes. The correlation coefficient (ρ = 0.186, p-value<2.2e- 16) suggests a weak association between the two variables. Our interpretation is that when people stay for a longer period inside the museum, they tend to limit the number of visited nodes and prefer to dedicate more time to exploring those they visit thoroughly (sometimes visiting them more than once), instead of visiting a higher number of different nodes. We   also   examine   a   distribution   of   an   average   of   visitors’   length   of   stay   in   the   museum   classified   by   the   hour   of   the   day   they   visit   the   museum.   This   analysis   reveals   whether   or   not   visitors’   entry   time   affects   their   length   of   stay   at   the   museum.  We  divide  a  day  of  the  week  into  two  groups  depending  on  the  closing     4   time  of  the  museum.  The  first  group  consists  of  Monday,  Thursday,  Saturday  and   Sunday,  when  the  door  closes  at  18:00  and  the  second  group  includes  Wednesday   and  Friday,  when  the  museum  closes  at  21:45.   visitor's length of stay vs entry time to the museum visitor's length of stay vs entry time to the museum 4:48:00 04:48:00 03:36:00 3:36:00 02:24:00 2:24:00 2"may& 3"may& 6"may& 8"may& 01:12:00 1:12:00 30"abr& 5"may& 7"may& 30"jun& 2"jul& 9"may& 1"jul& 3"jul& 4"jul& 5"jul& 8"jul& 12"ago& 14"ago& 7"jul& 9"jul& 13"ago& 18"ago& average& 15"ago& 16"ago& 19"ago& average& 00:00:00 0:00:00 Visitor's entry time to the museum Visitor's entry time to the museum (a)                                                                                                                                                                                          (b)   Figure   3.   (a)   The   distribution   of   the   average   stay   time   by   visiting   hour   on   Monday,   Thursday,  Saturday  and  Sunday.  (b)  The  distribution  of  the  average  stay  time  by  visiting   hour  on  Wednesday  and  Friday.   Figure   3   (a)   presents   a   clear   tendency   that   the   length   of   stay   time   decreases   toward  the  closing  hours  of  the  museum.  The  earlier  a  visitor  enters  the  museum,   the  longer  that  visitor  tends  to  stay  in  the  museum.  This  indicates  that  the  closing   time  of  the  museum  works  as  a  constraint  to  limit  the  length  of  visitors’  time  in  the   museum.  Thus,  visitors  seem  to  stay  longer  within  their  limited  available  time  in   order  to  maximize  their  benefits.   Conversely,   the   results   on   Wednesday   and   Friday   show   a   different   tendency   of   visitors   (see   Figure   3   (b)).   As   with   the   previous   analysis,   we   can   observe   a   tendency  that  the  length  of  stay  at  the  museum  decreases  with  the  advance  of  the   time.  However,  the  decrease  in  length  of  stay  is  slightly  mitigated  in  the  middle  of   the  day.  Just  after  the  opening  of  the  museum  (i.e.,  10:00-­‐11:00),  the  length  of  stay   is  greatest,  but  in  the  late  afternoon  (i.e.,  17:00-­‐18:00),  the  length  of  stay  increases   slightly.  This  data  makes  us  suppose  there  might  be  two  kinds  of  visitors.  That  is,   while  some  intend  to  maximize  their  utility  (e.g.,  staying  time)  within  the  limited   time   the   museum   is   open   by   visiting   earlier,   others   try   to   take   advantage   of   the   longer  hours  and  wait  until  the  evening  to  visit.   All  of  these  analyses  and  results  indicate  that  the  time  visitors  enter  can  be  used  to   predict   visitors’   length   of   stay   in   the   museum,   but   their   length   of   stay   in   the   museum  doesn’t  have  any  correlation  with  the  number  of  visited  nodes  over  the   course   of   their   visit.   While   the   longer   length   of   stay   slightly   suggests   a   larger   number   of   visited   nodes,   their   relationship   is   not   significant.   Such   a   result   is   particularly  useful  for  the  daily  management  of  peak  periods  and  rush  hours  by  the   staff.   The  length  of  stay  at  each  node     5   9:00-10:00 10:00-11:00 11:00-12:00 12:00-13:00 13:00-14:00 14:00-15:00 15:00-16:00 16:00-17:00 Average length of stay in the museum Average length of stay in the museum 9:00-10:00 10:00-11:00 11:00-12:00 12:00-13:00 13:00-14:00 14:00-15:00 15:00-16:00 16:00-17:00 17:00-18:00 18:00-19:00 19:00-20:00 20:00-21:00   We   examine   how   long   a   visitor   stays   at   each   node   in   a   disaggregated   way   and   uncover  the  feature  of  each  node  by  comparing  the  analysis  of  an  accumulation  of   individual  visitor’s  behavior.   Our  preliminary  analysis  shows  that  node  E  and  node  S  have  a  much  longer  length   of   stay   than   the   other   nodes.   The   median   length   of   stay   at   nodes   E   and   S   are   00:16:29  and  00:19:03,  respectively,  while  the  average  length  of  stay  at  the  other   nodes   is   00:03:14.   Node   E   is   situated   in   the   ticket   sale   desk,   indicating,   unsurprisingly,  that  visitors  may  wait  for  a  long  queue  in  order  to  purchase  a  ticket.   Winged   Victory   of   Samothrace,   where   sensor   S   is   located,   is   one   of   the   most   attractive  exhibits  in  the  museum-­‐together  with  the  huge  staircase  in  front  of  the   exhibit.  Besides,  many  visitors  use  the  stairs  as  a  improvised  chair  to  take  a  rest   during  their  visit.  Those  two  factors  make  node  E  and  node  S  different  from  other   nodes.   We   speculate   that   the   unique   uses   for   these   spaces   results   in   their   much   longer  length  of  stay,  and  this  fact  motivates  us  to  exclude  these  nodes  from  our   following  analysis.   (a)                                                                                                                                                                            (b)   Figure   4.  (a)  The  boxplot  of  the  length  of  stay  in  each  node.  (b)  The  comparative  visitor   decay  curves.   Figure  4  (a)  presents  the  boxplot  of  visitors’  length  of  stay  at  each  node.  We  can   observe   that   two   groups   exist:   node   V   (Venus   de   Milo),   node   C   (Salle   des   Caryatides),  and  node  B  (Great  Gallery)  experience  a  longer  stay;  node  D  (Gallery   Daru),  node  P  (Sphinx),  and  node  G  (Salle  des  Verres)  experience  a  shorter  stay.   While  the  median  length  of  stay  for  the  former  group  is  03:02,  the  median  for  the   latter  is  00:44.  Among  these  nodes  the  visitors’  length  of  stay  and  its  range  in  node   P  is  much  shorter  than  that  of  other  nodes,  and  as  a  result,  98%  of  all  visitors  to   node  P  have  a  length  of  stay  between  19  seconds  and  113  second.   Conversely,  Figure  4  (b)  shows  the  comparative  “visitor  survival  curve”.  We  plot   the   percentage   of   visitors   at   given   times.   This   plot   is   frequently   used   in   visitor   studies   in   order   to   analyze   the   length   of   stay,   when   half   of  the   visitors   leave,   an   exhibit  or  room  (Bicknell,  1995).  We  can  observe  that  the  length  of  stay  is  largely   varied  among  nodes:  00:20-­‐00:30  for  node  D,  01:30-­‐01:40  for  node  V,  01:30-­‐01:40     6   for  node  C,  00:00-­‐  00:10  for  node  P,  02:10-­‐02:20  for  node  B,  and  00:00-­‐00:10  for   node   G.   However,   most   of   the   nodes   experience   a   length   of   stay   within   a   few   minutes.  Again,  we  can  observe  that  two  kinds  of  nodes  exist,  which  we  can  classify   as  shorter  and  longer  stay  type  nodes.  Within  the  former  group,  half  of  visitors  left   the  node  by  10-­‐20  seconds  in  the  cases  of  nodes  G  and  P.  In  the  case  of  the  node  D,   this  duration  is  little  bit  longer,  but  half  of  visitors  stayed  for  just  20-­‐30  seconds.   Relationship  between  visitors’  length  of  stay  and  density   The   perspective   on   length   of   stay,   however,   greatly   changes   when   examining   visitors’   duration   of   stay   in   relationship   to   the   degree   of   the   occupancy   of   each   node.  Figure  5  shows  the  relationship  between  each  node’s  occupancy  normalized   by  the  maximum  number  of  visitors  in  the  area  (x-­‐axis)  and  the  average  duration   of  stay  expressed  by  seconds  (y-­‐axis).  As  we  can  see,  a  clear  tendency  exists  among   all  data.  The  average  duration  of  stay  first  goes  up  with  the  room  occupancy  from   point   W   to   point   X,   then   stays   around   the   maximum   on   some   occupancy   level   interval  (point  X  to  Z).  After  that,  length  of  stay  drops  down  as  the  occupancy  level   starts  to  exceed  a  certain  threshold  (point  Z).   Figure  5.  Distribution  of  the  normalized  occupancy  vs  the  length  of  stay.   Table  1.  The  threshold  of  the  normalized  occupancy  in  each  node.                                                        Point  X                                                                Point  Y                                                                        Point  Z   Node  D                                                183sec  (0.235)                                  204sec  (0.498)                                          188sec  (0.636)                       Node  V                                                289sec  (0.235)                                  315sec  (0.368)                                          271sec  (0.781)                   Node  C                                                225sec  (0.221)                                  249sec  (0.352)                                          187sec  (0.753)                       Node  P     7                                                    94sec  (0.214)                                    118sec  (0.5)                                                        79sec  (0.764)                   Node  B                                                353sec  (0.238)                                  375sec  (0.504)                                          293sec  (0.772)               Node  G                                                167sec  (0.267)                                  208sec  (0.351)                                          145sec  (0.657)                     The  length  of  stay  for  point  W  is  extremely  short  because  it  contains  visitors,  who   just   pass   by   the   area   rather   than   stay   there,   which   we   demonstrated   in   the   previous   sections.   When   the   occupancy   level   increases   from   point   W   to   X,   the   number   of   visitors   who   tend   to   stay   longer   also   increases.   Point   X   can   be   considered  the  equilibrium  between  an  ideal  length  of  stay  for  a  visitor  when  he  or   she  is  free  to  stay  as  long  as  he  or  she  wants.  This  is  because  visitors  are  free  to   look  at  the  artwork  between  point  W  and  X  without  any  obstacles  due  to  the  low   density  of  other  visitors.  Conversely,  from  point  X  onward,  the  average  length  of   stay   remains   almost   flat   until   point   Z,   at   which   point   visitors’   length   of   stay   drastically  starts  to  decrease.  We  speculate  that  this  is  because  the  exceeding  high-­‐ density   of   other   visitors   may   affect   a   visitor’s   comfort,   resulting   in   a   desire   to   escape  the  crowd.   The  length  of  stay  at  point  X  is  largely  varied  depending  on  the  node.  Node  B  at   point   X   shows   the   longest   length   of   stay   (353  seconds)  with  a   0.238   normalized   occupancy  level  among  the  other  nodes.  Then,  node  B  remains  a  flat  until  point  Z.   Conversely,  node  P  at  point  X  represents  the  shortest  length  of  stay  (94  seconds)   with   a   0.214   normalized   occupancy   level   among   the   other   nodes.   Although   the   occupancy  level  is  similar  between  those  two  nodes,  the  former’s  length  of  stay  is   almost   four   times   longer   than   the   latter’s.   However,   the   normalized   occupancy   level   of   point   X   for   node   B   is   higher   than   that   of   node   P   (i.e.,   0.238   vs.   0.214,   respectively).   Additionally,   these   nodes   correspond   with   the   maximum   and   minimum  lengths  of  stay  at  point  X,  although  the  occupancy  level  of  both  nodes  is   quite   similar:   both   of   them   are   around   0.50.   Furthermore,   the   length   of   stay   at   node  P  starts  to  decrease  earlier  than  that  of  node  B.     Regarding  the  relationship  between  node  V  and  node  B,  although  point  X  for  both   of   them   shows   a   similar   length   of   stay   (i.e.,   0.235   vs.   0.238,   respectively),   the   maximum  length  of  stay  of  node  V  has  a  much  lower  occupation  level  than  those  of   nodes  P  and  B  (i.e.,  0.368  for  node  V).  Additionally,  node  V  has  the  highest  density   for  point  Z  (0.781)  with  the  second  longest  length  of  stay  (271sec),  whereas  the   longest  length  of  stay  is  at  node  B  (293seconds).   All   of   these   facts   indicate   that   the   artworks   at   node   B   attract   and   hold   visitors   much  more  strongly  than  the  artworks  of  node  V  and  P.  Node  B  seems  to  inspire   visitors   to   stay   longer,   even   with   higher   occupation   density,   while   node   V   and   P   seem  to  cause  visitors  to  stay  for  shorter  durations  when  experiencing  the  same   density.   Also,   the   data   shows   that   the   average   length   of   stay   of   a   visitor   and   the   occupation   level   of   a   node   form   a   clear   pattern.   The   crowd   density   around   the   artwork  largely  affects  a  visitor’s  length  of  stay  either  positively  or  negatively,  and   the  type  of  effect  is  largely  dependent  on  the  node.     8   We  speculate  that,  although  it  also  depends  on  the  nature  of  the  galleries  and  the   type  of  visitors,  up  to  certain  occupancy  limits  visitors  are  actually  attracted  by  the   crowd:  however,  once  the  crowd’s  size  reaches  a  certain  level,  visitors  will  try  to   avoid  the  node.  This  indicates  that  we  may  use  point  Z  as  a  threshold  to  distinguish   visitors’  level  of  comfort,  which  enables  us  to  manage  the  environment  in  order  to   avoid  exceeding  this  density  threshold  in  the  environment.  Also,  we  may  consider   some  characteristic  points  such  as  point  X  and  Y  to  enabling  to  manage  different   types  of  environments/crowds  inside  the  galleries.   Conclusions  and  discussion   The   limitation   of   our   proposed   method   is   as   follows:   first,   the   sensor   can   detect   only  mobile  devices  in  which  the  Bluetooth  function  is  activated.  This  indicates  the   representativeness   of   the   sample   may   have   a   strong   bias   toward   certain   groups,   such   as   the   upper   class,   higher-­‐educated   people,   and   the   younger   generation   rather   than   seniors   and   children.   Moreover,   sensors   are   not   capable   of   distinguishing   individual   visit   or   group   visits.   The   representativeness   of   our   sample   can   be   calculated   by   comparing   the   sample   obtained   from   Bluetooth   detection  and  the  head  count  by  hand  over  a  shorter  period  (e.g,  a  few  hours  or   few  days).  In  fact  we  applied  a  systematic  comparison  over  a  longer  period  (one   month)   with   the   number   of   ticket   sales   in   the   desk   where   a   Bluetooth   sensor   is   installed.  Second,  the  sensor  enables  us  to  detect  visitors’  presence  in  the  specific   area,  which  is  determined  by  the  radius  of  each  sensor’s  detection  range,  but  such   sensors   cannot   specify   whether   visitors   are   actually   looking   at   the   artwork   or   if   they  are  simply  in  the  area.  This  shortcoming  is  strongly  related  to  the  following   limitation:  “while  time  is  a  necessary  condition  for  learning,  time  in  a  gallery  does   not   correspond   directly   to   time   spent   attending   to   exhibitions”   [14].   Third,   Bluetooth  detection  techniques  cannot  disclose  the  visitors’  motivations  and  inner   thoughts  in  any  way:  this  method  merely  identifies  their  presence  and  the  precise   length  of  their  stay.  Finally,  sensors  are  not  capable  to  collect  socio-­‐demographics   (i.e.,  origin,  age,  gender,  profession)  as  for  other  traditional  behavioral  variables. As   a   backdrop   to   this   situation,   the   proposed   new   approach   enables   us   to   shed   light  on  some  unknown  aspects  of  visitors’  behavior.  Our  proposed  system  and  the   results   we   obtained   through   an   adequate   statistical   analysis   can   work   as   a   new   tool   for   the   museum   management.   Our   system   effectively   captures   an   individual   visitor’s  length  of  stay  in  different  ways  and  enables  us  to  store  this  information  as   a  large-­‐scale  dataset.  Such  a  method  appears  to  be  particularly  useful  for  medium   or  small  size  museums  where  it  could  be  possible  to  install  a  complete  system  of   sensors.   This   method   results   in   a   dataset   that   is   very   different   from   the   conventional  dataset  in  two  ways:  the  quantitative  size  of  the  dataset  and  the  finer-­‐ grain  detection  of  visitor  behavior  both  spatially  and  temporally.   All  of  our  findings  are  helpful  for  the  management  of  visitor  flow  in  order  to  reduce   congestion  at  specific  areas  and  around  specific  pieces  of  artwork.  It  is  also  useful   in  order  to  bring  to  light  less  visited  or  less  “attractive”  artworks/rooms  inside  the   museum,  in  order  to  propose  interpretation  tools  and/  or  walking  tours  capable  to   increase  the  value  of  such  “neglected”  spaces.  Additionally,  our  data  suggest  that   visitor  behavior  is  based  on  some  patterns,  which  make  it  possible  to  foresee  their     9   future   movement   in   a   dynamic   way.   Also,   these   data   are   significantly   useful   for   designing  the  spatial  arrangement  (e.g.,  changes  in  the  layout  of  exhibits,  facilities   and  advertisements),  depending  on  visitor  activities  and  use  of  space.  Finally,  our   findings   indicate   that   efficient   and   effective   congestion   management   of   the   museum  can  be  realized  by  limiting  the  number  of  visitors  that  are  able  to  enter   based  on  the  time  of  the  day.   References   1. Yalowitz, S. S. & Bronnenkant, Kerry. (2009) ”Timing and Tracking: Unlocking Visitor Behavior” Visitor Studies 12(1): 47-64. 2. Shoval N, McKercher B, Birenboim A, Ng E, 2013, “The application of a sequence alignment method to the creation of typologies of tourist activity in time and space” Environment and Planning B: Planning and Design advance online publication, doi:10.1068/b38065 3. González M C, Hidalgo C A, Barabási A L, 2008, “Understanding individual human mobility patterns” Nature 453 779-782 4. Ratti C, Pulselli R, Williams S, Frenchman D, 2006, “Mobile Landscapes: using location data from cell phones for urban analysis” Environment and Planning B: Planning and Design 33(5) 727-748 5. Kanda T, Shiomi M, Perrin L, Nomura T, Ishiguro H, Hagita N, 2007, “Analysis of people trajectories with ubiquitous sensors in a science museum” Proceedings 2007 IEEE International Conference on Robotics and Automation (ICRA’07) 4846-4853 6. Tröndle, M, Greenwood, S, Kirchberg, V, Tschacher, W, 2014, “An Integrative and Comprehensive Methodology for Studying Aesthetic Experience in the Field: Merging Movement Tracking, Physiology, and Psychological Data“, Environment and Behavior, 46 (1) pp102-135. 7. Yoshimura Y, Sobolevsky S, Ratti C, Girardin F, Carrascal J P, Blat J, Sinatra R, 2014, “An analysis of visitors’ behaviour in The Louvre Museum: a study using Bluetooth data” Environment and Planning B: Planning and Design 41 (6) 1113-1131 8. Kostakos V, O'Neill E, Penn A, Roussos G, Papadongonas D, 2010, “Brief encounters: sensing, modelling and visualizing urban mobility and copresence networks” ACM Transactions on Computer Human Interaction 17(1) 1-38 9. Versichele M, Neutens T, Delafontaine M, Van de Weghe N, 2011, “The use of Bluetooth for analysing spatiotemporal dynamics of human movement at mass events: a case study of the Ghent festivities” Applied Geography 32 208-220 10. Delafontaine M, Versichele M, Neutens T, Van de Weghe N, 2012, “Analysing spatiotemporal sequences in Bluetooth tracking data” Applied Geography 34 659-668 11. Mayer-Schönberger V, Cukier K, 2013, Big Data: A Revolution That Will Transform How We Live, Work and Think (John Murray, London) 12. Barceló J, Montero L, Marqués L, Carmona C, 2010, “Travel Time Forecasting and Dynamic Origin-Destination Estimation for Freeways Based on Bluetooth Traffic Monitoring” Transportation Research Record: Journal of the Transportation Research Board 2175: 19-27 13. Stallings W, 2011, Cryptography and Network Security: Principles and Practice, 5th Edition (Prentice Hall, Boston MA) 14. Hein G, 1998, Learning in the Museum (Routledge, London)   10  

Journal

PhysicsarXiv (Cornell University)

Published: Apr 30, 2016

There are no references for this article.