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

Learn More →

Vision-based assessment of viability of acorns using sections of their cotyledons during automated scarification procedure

Vision-based assessment of viability of acorns using sections of their cotyledons during... Design considerationsIn the design process of the considered automaton, several contradictory demands must have been mitigated. We have identified the following main constraints: efficiency (throughput), cost, and accuracy.The device is expected to operate fast enough to provide a satisfactory number of seeds for the greenhouse in a limited period of time assuming continuous operation. To find a balance between such demand and next expectations (limited cost and high accuracy), we have selected basic functions that the automaton must implement. They correspond to the sequence of actions to which each acorn must be subjected. Such sequence is presented in Figure 1.Figure 1:Sequence of actions to which each acorn must be subjected.The first step in the considered sequence of actions is the delivery of acorns. This action is performed by a vibrating feeder as presented in Figure 2. It can take acorns from their large amount (for example, a content of full barrel) and it creates a queue going a spiral road up to the place where they are picked up by the described automaton.Figure 2:Vibratory feeder as the input container of acorns.The next step is the downloading of acorns. Automaton must pick up one acorn on one step, because all the next actions are designed for the analysis and processing of a singular acorn. It means the downloading procedure must separate one acorn from a whole queue and pass it on to further processing by the next functional elements of the machine. Elements designed for downloading acorns are presented in Figure 3.Figure 3:Interconnection between the vibratory feeder and V-shaped double belt conveyor driven by DC engines with a photoelectric sensor comprising acorn singulator module.The downloaded acorn must be checked in terms of orientation. It is important because the cutting (scarification) plane will be localized near the front tip of the acorn (assuming as the “front” in the direction of its movement on a special belt conveyor). Meanwhile, the internal structure of the acorn is such that near one end are interesting cotyledons subjected to vision-based assessment, whereas at the other end there is an embryo. Scarification must be performed in the distal cotyledon part, as cutting the embryo kills the seed. The identification of the part of the acorn in which the embryo is located is possible based on its shape analysis, because the tip of the embryo is more spike-formed. Orientation check is performed by a special computer image analysis system based on a Harris detector for the determination of the orientation of the processed acorn. Such part of the automaton structure, which is designed for checking the acorn orientation, is presented in Figure 4.Figure 4:Camera-based detector of the orientation of acorns and measurement of their length (A) and the sample image captured by the monochrome camera (B).After orientation checking, three situations are possible. The first is to detect if acorn orientation is correct (it means in front is part of the acorn without the embryo). Then, the acorn can be passed directly to the gripper and scarification device. The second situation happens when the acorn is in the wrong orientation. In this case, before gripping, the acorn must be rotated 180° in a special rotator (see Figure 5).Figure 5:Rotator fixed on the test rig and driven by a stepper motor, input channel with a photosensor, rotary channel inside the well, and two output channels (one for properly oriented acorns and one for unrecognized ones).The third situation occurs when the vision system fails to determine whether the acorns are in the correct or incorrect orientation. Such acorn is marked as “unknown” and removed from the automaton, because its further processing is pointless.The next action is the gripping and scarification process. Before gripping, the acorn can be positioned by moving up or down by means of a special cam mechanism. This process is necessary, because the lengths of individual acorns vary significantly, whereas scarification knives are always at the same level. The position of acorns that is always cut off and removed is the front part of the acorn, for example, 20% of its total length.The length of every acorn is determined by the same visual system, which determines the proper or improper acorn orientation, so positioning can be easy and precise.The acorn fixed in an electrically controlled gripper is next moved to the scarification device made of two blades (round knives) rotating in opposite directions (Figure 6). The transportation of the gripper with the acorn inside is performed by a rotating arm controlled by an industrial controller.Figure 6:Rotary blades placed between the outlet of the rotator and a separator consisting of the camera module and receiving baskets (A) and the gripper stopped in front of the camera module, close to the first receiving basket (B).An acorn before and after scarification is shown in Figure 7.Figure 7:An acorn before and after scarification.The last and most important action shown in Figure 1, namely health check and sorting, will be described in detail in the next three chapters.Health check and sortingThe most important part of the described automaton is designated to the acorn health recognition process and, based on the results of such recognition, sorting of acorns into three classes: healthy for sowing, wrong which must be rejected, and doubtful, which can be additionally assessed by human experts.The first step of the work performed by the health check and sorting subsystem is image processing. The image of the acorn section, registered by a special color camera, must be processed for the segmentation of such part of the image, in which interesting cotyledons are visible. The stages of such processing are presented in Figure 8.Figure 8:Sample images at subsequent stages of section image processing during segmentation: color conversion, detection and segmentation, and analysis and recognition.The next registered image must be classified into one of the considered classes: healthy, wrong (spoiled), and doubtful (partially spoiled). As the basis for the recognition (classification) process, the intensity of the considered part of the image of the acorn cotyledon section was selected. Figure 9 and Figure 10 show how well correlated is the considered section intensity with acorn healthiness.Figure 9:Sample section of scarified acorns and their histograms: (A) healthy, (B and C) doubtful, and (D) wrong (totally spoiled).Legend: red, histogram of cotyledon; blue, overall histogram including the holder. A 12-bit representation was used.Figure 10:Distributions of the intensity of the section of healthy acorns (green) and wrong (spoiled ones; red).Results of an experiment completed in year 2015.Several in-filed experiments have been performed during which several hundreds of acorns have been measured, cut, sections registered with a camera, and sowed in the cartridge nursery. This allowed collecting reference data for training and testing the performance of the viability prediction for particular algorithms. The results of the performed experiments are shown in Figure 10.The next step is the decision-making procedure.In the beginning, we accepted a satisfactory accuracy (84%) equal to the rate of recognitions provided by professionals who perform an assessment of acorn as their work. This was determined during in-field experiments performed at the initial stage of the project. Acquired data enabled to work out computer models of classifier as good as humans or even better in some cases [1]. This proves that images of cut cotyledons contain significant discriminative information for the classification of scarified acorns by means of image recognition and basic machine learning techniques: k-nearest neighbors, support vector machine, and artificial neural network (ANN).The recognition performance based on binary discrimination is presented in Figure 11 in terms of recognition accuracy and in Figure 12 in terms of the receiver operating characteristic (ROC) curve.Figure 11:Accuracy of sorting computed for subsequent normalized values of intensity threshold.Figure 12:ROC curve for binary separation of seed based on the intensity of the green channel of the image of the section of an acorn.Legend: point, where the accuracy of prediction reaches maximum value.These results were used as a base for optimal threshold value determination. If section intensity for a particular acorn is over the threshold, this acorn is classified as healthy. If the intensity is below the second threshold (determined experimentally), the acorn is classified as definitely wrong (spoiled). Acorns whose section intensity is between selected thresholds are classified as doubtful.DiscussionVarious imaging techniques are being used for assessing the quality of biological tissues as well as in food maintenance and postharvest crop processing. A popular and efficient method of data acquisition is hyperspectral imaging, which allows for the precise identification of spectral properties of the subject under analysis. This, however, is expensive and time-consuming and thus infeasible in real-time due to the delay introduced by the acquisition and sophisticated processing of a huge amount of data [2]. Because the aim of our design is the model of automaton intended for processing high volume of acorns in a short period of time, we used a color area scan camera sensitive to three bands (red, green, and blue) in the module of the detector of pathological changes for the acquisition of images of scarified acorns [3]. This allows finding a balance between speed, accuracy, and cost of the detector device. A similar solution was applied by other scientists for the discrimination of seeds and contamination by means of shape and color [4].At a certain stage of processing, acorns are treated with a chemical substance to prevent an attack of fungus. This substance adheres to the pericarp; thus, its color does not represent a real status of the seed. For this reason, the camera module dedicated to the detection of the orientation and measurement of its length contains a monochrome sensor and backlight illumination that allows for the reliable representation of the acorn’s contours. To achieve a satisfactory fidelity of detection, multiple recognitions of a particular object are performed and the final recognition is the result of the voting procedure. This prevents hesitant decisions that would result in discarding the viable part of the seed and thus false recognition of a healthy or partly spoiled acorn.An interesting issue is the number of fractions that acorns should be assigned to. In the very basic approach, acorns can be divided into two fractions: healthy and spoiled. This can be justified by the type of data we acquire after in-field experiments where all acorns whose sections have been registered and assessed have been collected before sowing. This allows implementing a simple discriminative model of viability prediction. However, it may be not satisfactory because cultivation in cartridge nursery is more expensive than others. Therefore, a fraction of partly spoiled acorns can be introduced for less expensive and less demanding type of farming or other purposes. Separation into three fractions can be implemented by introducing a three-class discrimination to features of cut cotyledons or introducing a confidence factor to the output of a two-class ANN architecture. In this way, the value that does not correspond to “strong healthy” or “strong spoiled” can be treated as “partly spoiled” or doubtful. The other strategy can assume two stages of classification. At first, the healthy seeds can be identified. The rest of the acorns can be subjected to further discrimination into spoiled and partly spoiled.Final remarksAt the end of this article we must stress that for all described functions adequate modules implementing these functions have been designed and manufactured. Some of them implement few functionalities, for example, the first machine vision module detects the seed, measures its length, and determines the orientation. These three factors are acquired by the algorithm in a single step.The sequential operation mode of the automaton allows for the identification of properties of the components of the automaton and to measure significant parameters of the process: duration of each action, delays, errors, etc. To achieve a higher throughput, it should be possible, however, to arrange this series of operations executed by different modules into a pipeline. In this mode, the automaton would be able to operate in parallel, that is, it will process multiple acorns concurrently, each at subsequent stage. This, however, requires a fine-tuning of the control parameters and precise synchronization of actions.Expenditures are limited by the schedule of the project and in any case must not be surpassed due to the amount of funding assigned by the agency. Time factor is also involved as the duration of the project is also limited. Thus, only critical components have been designed from scratch by the interdisciplinary team and manufactured by the industrial member of the consortium: gripper, belt conveyor, rotator, cutting blades (scarifier), rotating arm, and the chassis allowing to support and connect all mechanical components. The others have been purchased from external companies to meet milestones determined in the Gantt chart defined by the board of the managers for the research project. These are vibrating feeder, programmable-analogue controller, processing unit, and touch-screen control panel. Particular modules and infrastructure including sensors and actuators have been assembled using professional automation components: pneumatic actuators, stepper motors, encoders, photoelectric barriers, and machine vision sensors. However, in the beginning of the research, simplified models of particular modules have been arranged using off-the-shelf consumer components, rapid prototyping methods including 3D printing technology [5], [6], and model-based design including software-in-the-loop methodology. These allowed to acquire data at the initial stage necessary for the design of algorithms and to hammer out main assumptions for virtual prototyping and further manufacturing of the automaton.ConclusionsWe believe that the methodology of the design we applied and the functionality of the model consider significant issues related to oak seed nursery: economic, social, and environmental [7]. When implemented, it will allow minimizing waste of energy in greenhouse nursery due to the reduced number of empty cells in cartridges. This is the result of the separation of acorns into fractions: healthy spoiled and partly spoiled. A fully autonomous operation decreases the danger of repetitive strain injury of the employees performing manual scarification of a large amount of acorns with shears in the spring season. The automaton operates in stand-alone mode [8]. It means that the control unit and computing platforms are embedded into the device; thus, an interaction of an operator is necessary only in a few specific situations: configuration, initialization, error handling, etc. The device is highly reconfigurable; therefore, it can be adapted to specific demands by software calibration or hardware adjustment.Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.Research funding: The work presented was supported by the National Centre of Research and Development of Republic of Poland under the project “Functional model of automaton, comprising machine vision system, for scarification and assessment of acorn viability by means of automatic recognition of topography of mummification changes” (Grant No. PBS3/A8/134/2015).Employment or leadership: None declared.Honorarium: None declared.Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.References[1]Grabska-Chrząstowska J, Kwiecień J, Drożdż M, Bubliński Z, Tadeusiewicz R, Szczepaniak J, et al. Comparison of selected classification methods in automated oak seed sorting. J Res Appl Agric Eng 2017;62:31–3.Grabska-ChrząstowskaJKwiecieńJDrożdżMBublińskiZTadeusiewiczRSzczepaniakJComparison of selected classification methods in automated oak seed sortingJ Res Appl Agric Eng201762313[2]ElMasry GM, Nakauchi S. Image analysis operations applied to hyperspectral images for non-invasive sensing of food quality – a comprehensive review. Biosyst Eng 2016;142:53–82.10.1016/j.biosystemseng.2015.11.009http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000370100100005&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3ElMasryGMNakauchiSImage analysis operations applied to hyperspectral images for non-invasive sensing of food quality – a comprehensive reviewBiosyst Eng20161425382[3]Jabłoński M, Tylek P, Walczyk J, Tadeusiewicz R, Piłat A. Colour-based binary discrimination of scarified Quercus robur acorns under varying illumination. Sensors 2016;16:1–13.http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000382323200058&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3JabłońskiMTylekPWalczykJTadeusiewiczRPiłatAColour-based binary discrimination of scarified Quercus robur acorns under varying illuminationSensors201616113[4]Momin MA, Yamamoto K, Miyamoto M, Kondo N, Grift T. Machine vision based soybean quality evaluation. Comput Electro Agric 2017;140:452–60.10.1016/j.compag.2017.06.023MominMAYamamotoKMiyamotoMKondoNGriftTMachine vision based soybean quality evaluationComput Electro Agric201714045260[5]Jabłoński M, Tadeusiewicz R. Second International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP), IEEE, Kraków, 2016:1–3.JabłońskiMTadeusiewiczRSecond International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP), IEEEKraków201613[6]Przybyło J, Jabłoński M, Pociecha D, Tadeusiewicz R, Piłat A, Walczyk J, et al. Application of model-based design in prototyping of algorithms for experimental acorn scarification rig. J Res Appl Agric Eng 2017;62:166–70.PrzybyłoJJabłońskiMPociechaDTadeusiewiczRPiłatAWalczykJApplication of model-based design in prototyping of algorithms for experimental acorn scarification rigJ Res Appl Agric Eng20176216670[7]Tadeusiewicz R, Tylek P, Adamczyk F, Kiełbasa P, Jabłoński M, Bubliński Z, et al. Assessment of selected parameters of the automatic scarification device as an example of a device for sustainable forest management. Sustainability 2017;9:1–17.http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000419231500216&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3TadeusiewiczRTylekPAdamczykFKiełbasaPJabłońskiMBublińskiZAssessment of selected parameters of the automatic scarification device as an example of a device for sustainable forest managementSustainability20179117[8]Tadeusiewicz R, Tylek P, Adamczyk F, Kiełbasa P, Jabłoński M, Pawlik P, et al. Automation of the acorn scarification process as contribution to sustainable forest management: case study: common oak. Sustainability 2017;9:1–17.http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000419231500122&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3TadeusiewiczRTylekPAdamczykFKiełbasaPJabłońskiMPawlikPAutomation of the acorn scarification process as contribution to sustainable forest management: case study: common oakSustainability20179117 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Bio-Algorithms and Med-Systems de Gruyter

Vision-based assessment of viability of acorns using sections of their cotyledons during automated scarification procedure

Download PDF
8 pages

Loading next page...
 
/lp/de-gruyter/vision-based-assessment-of-viability-of-acorns-using-sections-of-their-yomFYZNfux

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
de Gruyter
Copyright
©2018 Walter de Gruyter GmbH, Berlin/Boston
ISSN
1896-530X
eISSN
1896-530X
DOI
10.1515/bams-2018-0006
Publisher site
See Article on Publisher Site

Abstract

Design considerationsIn the design process of the considered automaton, several contradictory demands must have been mitigated. We have identified the following main constraints: efficiency (throughput), cost, and accuracy.The device is expected to operate fast enough to provide a satisfactory number of seeds for the greenhouse in a limited period of time assuming continuous operation. To find a balance between such demand and next expectations (limited cost and high accuracy), we have selected basic functions that the automaton must implement. They correspond to the sequence of actions to which each acorn must be subjected. Such sequence is presented in Figure 1.Figure 1:Sequence of actions to which each acorn must be subjected.The first step in the considered sequence of actions is the delivery of acorns. This action is performed by a vibrating feeder as presented in Figure 2. It can take acorns from their large amount (for example, a content of full barrel) and it creates a queue going a spiral road up to the place where they are picked up by the described automaton.Figure 2:Vibratory feeder as the input container of acorns.The next step is the downloading of acorns. Automaton must pick up one acorn on one step, because all the next actions are designed for the analysis and processing of a singular acorn. It means the downloading procedure must separate one acorn from a whole queue and pass it on to further processing by the next functional elements of the machine. Elements designed for downloading acorns are presented in Figure 3.Figure 3:Interconnection between the vibratory feeder and V-shaped double belt conveyor driven by DC engines with a photoelectric sensor comprising acorn singulator module.The downloaded acorn must be checked in terms of orientation. It is important because the cutting (scarification) plane will be localized near the front tip of the acorn (assuming as the “front” in the direction of its movement on a special belt conveyor). Meanwhile, the internal structure of the acorn is such that near one end are interesting cotyledons subjected to vision-based assessment, whereas at the other end there is an embryo. Scarification must be performed in the distal cotyledon part, as cutting the embryo kills the seed. The identification of the part of the acorn in which the embryo is located is possible based on its shape analysis, because the tip of the embryo is more spike-formed. Orientation check is performed by a special computer image analysis system based on a Harris detector for the determination of the orientation of the processed acorn. Such part of the automaton structure, which is designed for checking the acorn orientation, is presented in Figure 4.Figure 4:Camera-based detector of the orientation of acorns and measurement of their length (A) and the sample image captured by the monochrome camera (B).After orientation checking, three situations are possible. The first is to detect if acorn orientation is correct (it means in front is part of the acorn without the embryo). Then, the acorn can be passed directly to the gripper and scarification device. The second situation happens when the acorn is in the wrong orientation. In this case, before gripping, the acorn must be rotated 180° in a special rotator (see Figure 5).Figure 5:Rotator fixed on the test rig and driven by a stepper motor, input channel with a photosensor, rotary channel inside the well, and two output channels (one for properly oriented acorns and one for unrecognized ones).The third situation occurs when the vision system fails to determine whether the acorns are in the correct or incorrect orientation. Such acorn is marked as “unknown” and removed from the automaton, because its further processing is pointless.The next action is the gripping and scarification process. Before gripping, the acorn can be positioned by moving up or down by means of a special cam mechanism. This process is necessary, because the lengths of individual acorns vary significantly, whereas scarification knives are always at the same level. The position of acorns that is always cut off and removed is the front part of the acorn, for example, 20% of its total length.The length of every acorn is determined by the same visual system, which determines the proper or improper acorn orientation, so positioning can be easy and precise.The acorn fixed in an electrically controlled gripper is next moved to the scarification device made of two blades (round knives) rotating in opposite directions (Figure 6). The transportation of the gripper with the acorn inside is performed by a rotating arm controlled by an industrial controller.Figure 6:Rotary blades placed between the outlet of the rotator and a separator consisting of the camera module and receiving baskets (A) and the gripper stopped in front of the camera module, close to the first receiving basket (B).An acorn before and after scarification is shown in Figure 7.Figure 7:An acorn before and after scarification.The last and most important action shown in Figure 1, namely health check and sorting, will be described in detail in the next three chapters.Health check and sortingThe most important part of the described automaton is designated to the acorn health recognition process and, based on the results of such recognition, sorting of acorns into three classes: healthy for sowing, wrong which must be rejected, and doubtful, which can be additionally assessed by human experts.The first step of the work performed by the health check and sorting subsystem is image processing. The image of the acorn section, registered by a special color camera, must be processed for the segmentation of such part of the image, in which interesting cotyledons are visible. The stages of such processing are presented in Figure 8.Figure 8:Sample images at subsequent stages of section image processing during segmentation: color conversion, detection and segmentation, and analysis and recognition.The next registered image must be classified into one of the considered classes: healthy, wrong (spoiled), and doubtful (partially spoiled). As the basis for the recognition (classification) process, the intensity of the considered part of the image of the acorn cotyledon section was selected. Figure 9 and Figure 10 show how well correlated is the considered section intensity with acorn healthiness.Figure 9:Sample section of scarified acorns and their histograms: (A) healthy, (B and C) doubtful, and (D) wrong (totally spoiled).Legend: red, histogram of cotyledon; blue, overall histogram including the holder. A 12-bit representation was used.Figure 10:Distributions of the intensity of the section of healthy acorns (green) and wrong (spoiled ones; red).Results of an experiment completed in year 2015.Several in-filed experiments have been performed during which several hundreds of acorns have been measured, cut, sections registered with a camera, and sowed in the cartridge nursery. This allowed collecting reference data for training and testing the performance of the viability prediction for particular algorithms. The results of the performed experiments are shown in Figure 10.The next step is the decision-making procedure.In the beginning, we accepted a satisfactory accuracy (84%) equal to the rate of recognitions provided by professionals who perform an assessment of acorn as their work. This was determined during in-field experiments performed at the initial stage of the project. Acquired data enabled to work out computer models of classifier as good as humans or even better in some cases [1]. This proves that images of cut cotyledons contain significant discriminative information for the classification of scarified acorns by means of image recognition and basic machine learning techniques: k-nearest neighbors, support vector machine, and artificial neural network (ANN).The recognition performance based on binary discrimination is presented in Figure 11 in terms of recognition accuracy and in Figure 12 in terms of the receiver operating characteristic (ROC) curve.Figure 11:Accuracy of sorting computed for subsequent normalized values of intensity threshold.Figure 12:ROC curve for binary separation of seed based on the intensity of the green channel of the image of the section of an acorn.Legend: point, where the accuracy of prediction reaches maximum value.These results were used as a base for optimal threshold value determination. If section intensity for a particular acorn is over the threshold, this acorn is classified as healthy. If the intensity is below the second threshold (determined experimentally), the acorn is classified as definitely wrong (spoiled). Acorns whose section intensity is between selected thresholds are classified as doubtful.DiscussionVarious imaging techniques are being used for assessing the quality of biological tissues as well as in food maintenance and postharvest crop processing. A popular and efficient method of data acquisition is hyperspectral imaging, which allows for the precise identification of spectral properties of the subject under analysis. This, however, is expensive and time-consuming and thus infeasible in real-time due to the delay introduced by the acquisition and sophisticated processing of a huge amount of data [2]. Because the aim of our design is the model of automaton intended for processing high volume of acorns in a short period of time, we used a color area scan camera sensitive to three bands (red, green, and blue) in the module of the detector of pathological changes for the acquisition of images of scarified acorns [3]. This allows finding a balance between speed, accuracy, and cost of the detector device. A similar solution was applied by other scientists for the discrimination of seeds and contamination by means of shape and color [4].At a certain stage of processing, acorns are treated with a chemical substance to prevent an attack of fungus. This substance adheres to the pericarp; thus, its color does not represent a real status of the seed. For this reason, the camera module dedicated to the detection of the orientation and measurement of its length contains a monochrome sensor and backlight illumination that allows for the reliable representation of the acorn’s contours. To achieve a satisfactory fidelity of detection, multiple recognitions of a particular object are performed and the final recognition is the result of the voting procedure. This prevents hesitant decisions that would result in discarding the viable part of the seed and thus false recognition of a healthy or partly spoiled acorn.An interesting issue is the number of fractions that acorns should be assigned to. In the very basic approach, acorns can be divided into two fractions: healthy and spoiled. This can be justified by the type of data we acquire after in-field experiments where all acorns whose sections have been registered and assessed have been collected before sowing. This allows implementing a simple discriminative model of viability prediction. However, it may be not satisfactory because cultivation in cartridge nursery is more expensive than others. Therefore, a fraction of partly spoiled acorns can be introduced for less expensive and less demanding type of farming or other purposes. Separation into three fractions can be implemented by introducing a three-class discrimination to features of cut cotyledons or introducing a confidence factor to the output of a two-class ANN architecture. In this way, the value that does not correspond to “strong healthy” or “strong spoiled” can be treated as “partly spoiled” or doubtful. The other strategy can assume two stages of classification. At first, the healthy seeds can be identified. The rest of the acorns can be subjected to further discrimination into spoiled and partly spoiled.Final remarksAt the end of this article we must stress that for all described functions adequate modules implementing these functions have been designed and manufactured. Some of them implement few functionalities, for example, the first machine vision module detects the seed, measures its length, and determines the orientation. These three factors are acquired by the algorithm in a single step.The sequential operation mode of the automaton allows for the identification of properties of the components of the automaton and to measure significant parameters of the process: duration of each action, delays, errors, etc. To achieve a higher throughput, it should be possible, however, to arrange this series of operations executed by different modules into a pipeline. In this mode, the automaton would be able to operate in parallel, that is, it will process multiple acorns concurrently, each at subsequent stage. This, however, requires a fine-tuning of the control parameters and precise synchronization of actions.Expenditures are limited by the schedule of the project and in any case must not be surpassed due to the amount of funding assigned by the agency. Time factor is also involved as the duration of the project is also limited. Thus, only critical components have been designed from scratch by the interdisciplinary team and manufactured by the industrial member of the consortium: gripper, belt conveyor, rotator, cutting blades (scarifier), rotating arm, and the chassis allowing to support and connect all mechanical components. The others have been purchased from external companies to meet milestones determined in the Gantt chart defined by the board of the managers for the research project. These are vibrating feeder, programmable-analogue controller, processing unit, and touch-screen control panel. Particular modules and infrastructure including sensors and actuators have been assembled using professional automation components: pneumatic actuators, stepper motors, encoders, photoelectric barriers, and machine vision sensors. However, in the beginning of the research, simplified models of particular modules have been arranged using off-the-shelf consumer components, rapid prototyping methods including 3D printing technology [5], [6], and model-based design including software-in-the-loop methodology. These allowed to acquire data at the initial stage necessary for the design of algorithms and to hammer out main assumptions for virtual prototyping and further manufacturing of the automaton.ConclusionsWe believe that the methodology of the design we applied and the functionality of the model consider significant issues related to oak seed nursery: economic, social, and environmental [7]. When implemented, it will allow minimizing waste of energy in greenhouse nursery due to the reduced number of empty cells in cartridges. This is the result of the separation of acorns into fractions: healthy spoiled and partly spoiled. A fully autonomous operation decreases the danger of repetitive strain injury of the employees performing manual scarification of a large amount of acorns with shears in the spring season. The automaton operates in stand-alone mode [8]. It means that the control unit and computing platforms are embedded into the device; thus, an interaction of an operator is necessary only in a few specific situations: configuration, initialization, error handling, etc. The device is highly reconfigurable; therefore, it can be adapted to specific demands by software calibration or hardware adjustment.Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.Research funding: The work presented was supported by the National Centre of Research and Development of Republic of Poland under the project “Functional model of automaton, comprising machine vision system, for scarification and assessment of acorn viability by means of automatic recognition of topography of mummification changes” (Grant No. PBS3/A8/134/2015).Employment or leadership: None declared.Honorarium: None declared.Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.References[1]Grabska-Chrząstowska J, Kwiecień J, Drożdż M, Bubliński Z, Tadeusiewicz R, Szczepaniak J, et al. Comparison of selected classification methods in automated oak seed sorting. J Res Appl Agric Eng 2017;62:31–3.Grabska-ChrząstowskaJKwiecieńJDrożdżMBublińskiZTadeusiewiczRSzczepaniakJComparison of selected classification methods in automated oak seed sortingJ Res Appl Agric Eng201762313[2]ElMasry GM, Nakauchi S. Image analysis operations applied to hyperspectral images for non-invasive sensing of food quality – a comprehensive review. Biosyst Eng 2016;142:53–82.10.1016/j.biosystemseng.2015.11.009http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000370100100005&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3ElMasryGMNakauchiSImage analysis operations applied to hyperspectral images for non-invasive sensing of food quality – a comprehensive reviewBiosyst Eng20161425382[3]Jabłoński M, Tylek P, Walczyk J, Tadeusiewicz R, Piłat A. Colour-based binary discrimination of scarified Quercus robur acorns under varying illumination. Sensors 2016;16:1–13.http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000382323200058&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3JabłońskiMTylekPWalczykJTadeusiewiczRPiłatAColour-based binary discrimination of scarified Quercus robur acorns under varying illuminationSensors201616113[4]Momin MA, Yamamoto K, Miyamoto M, Kondo N, Grift T. Machine vision based soybean quality evaluation. Comput Electro Agric 2017;140:452–60.10.1016/j.compag.2017.06.023MominMAYamamotoKMiyamotoMKondoNGriftTMachine vision based soybean quality evaluationComput Electro Agric201714045260[5]Jabłoński M, Tadeusiewicz R. Second International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP), IEEE, Kraków, 2016:1–3.JabłońskiMTadeusiewiczRSecond International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP), IEEEKraków201613[6]Przybyło J, Jabłoński M, Pociecha D, Tadeusiewicz R, Piłat A, Walczyk J, et al. Application of model-based design in prototyping of algorithms for experimental acorn scarification rig. J Res Appl Agric Eng 2017;62:166–70.PrzybyłoJJabłońskiMPociechaDTadeusiewiczRPiłatAWalczykJApplication of model-based design in prototyping of algorithms for experimental acorn scarification rigJ Res Appl Agric Eng20176216670[7]Tadeusiewicz R, Tylek P, Adamczyk F, Kiełbasa P, Jabłoński M, Bubliński Z, et al. Assessment of selected parameters of the automatic scarification device as an example of a device for sustainable forest management. Sustainability 2017;9:1–17.http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000419231500216&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3TadeusiewiczRTylekPAdamczykFKiełbasaPJabłońskiMBublińskiZAssessment of selected parameters of the automatic scarification device as an example of a device for sustainable forest managementSustainability20179117[8]Tadeusiewicz R, Tylek P, Adamczyk F, Kiełbasa P, Jabłoński M, Pawlik P, et al. Automation of the acorn scarification process as contribution to sustainable forest management: case study: common oak. Sustainability 2017;9:1–17.http://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcApp=PARTNER_APP&SrcAuth=LinksAMR&KeyUT=WOS:000419231500122&DestLinkType=FullRecord&DestApp=ALL_WOS&UsrCustomerID=b7bc2757938ac7a7a821505f8243d9f3TadeusiewiczRTylekPAdamczykFKiełbasaPJabłońskiMPawlikPAutomation of the acorn scarification process as contribution to sustainable forest management: case study: common oakSustainability20179117

Journal

Bio-Algorithms and Med-Systems – de Gruyter

Published: Apr 7, 2018

References

  • Machine vision based soybean quality evaluation
  • Colour-based binary discrimination of scarified Quercus robur acorns under varying illumination
  • Assessment of selected parameters of the automatic scarification device as an example of a device for sustainable forest management