Reconstruction of kiwifruit fruit geometry using a CGAN trained on a synthetic dataset

Abstract Non-destructive crop yield estimation is a major ambition for the development of digital horticulture systems, where the goal is to have a machine that can look at fruit on the vine/tree and estimate key performance metrics such as fruit weight, size and shape. The on-orchard partial occlusion problem is a major obstacle for the creation of such systems, where it is not possible for a single camera/scanner to capture the complete fruit surface due to its limited field of view. In this work, a deep learning approach was taken to solve this problem. Framing the issue as an image-to-image translation problem, a Conditional Generative Adversarial Network was trained to realistically reconstruct the enclosed and complete surface of kiwifruit when supplied with incomplete surface data, where the surface data was missing due to occlusion. Rather than training the deep learning algorithm with empirically collected data, an alternative approach was taken: the model was trained using a synthesised dataset, a large collection of kiwifruit shapes generated computationally via a Monte-Carlo routine. This was an attempt to generalise the approach to be applicable to other crops and other domains, and to provide substantial savings in time, labour and material costs. The trained model was later applied to a smaller population of kiwifruit empirically scanned using an infrared scanner and could predict fruit weight with a mean absolute percentage error of less than 5% and was successful in realistically reconstructing the whole enclosed surface over a range of sizes, shapes and orientations.

[1]  Trevor Darrell,et al.  Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[2]  P. Martin,et al.  Refining the estimation of kiwifruit size from linear fruit dimensions , 2020 .

[3]  Charu C. Aggarwal,et al.  Advanced Topics in Deep Learning , 2018 .

[4]  Maryam Rahnemoonfar,et al.  Deep Count: Fruit Counting Based on Deep Simulated Learning , 2017, Sensors.

[5]  Andrew R. East,et al.  An automated random stacking tool for packaged horticultural produce , 2020 .

[6]  Sandeep Patil,et al.  Voxel-based representation, display and thickness analysis of intricate shapes , 2005, Ninth International Conference on Computer Aided Design and Computer Graphics (CAD-CG'05).

[7]  S. Sivakumaran,et al.  The nutritional composition of Zespri® SunGold Kiwifruit and Zespri® Sweet Green Kiwifruit. , 2018, Food chemistry.

[8]  O. A. Fawole,et al.  Preharvest and postharvest factors influencing bioactive compounds in pomegranate (Punica granatum L.)—A review , 2014 .

[9]  Hong Cheng,et al.  Early Yield Prediction Using Image Analysis of Apple Fruit and Tree Canopy Features with Neural Networks , 2017, J. Imaging.

[10]  Salah Sukkarieh,et al.  Mapping almond orchard canopy volume, flowers, fruit and yield using lidar and vision sensors , 2016, Comput. Electron. Agric..

[11]  Timo Aila,et al.  A Style-Based Generator Architecture for Generative Adversarial Networks , 2018, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[12]  P. Verboven,et al.  Modelling Cooling of Packaged Fruit Using 3D Shape Models , 2018, Food and Bioprocess Technology.

[13]  Won Suk Lee,et al.  Citrus Fruit Identification and Size Determination Using Machine Vision and Ultrasonic Sensors , 2005 .

[14]  W. Snelgar,et al.  Influence of time of shading on flowering and yield of kiwifruit vines , 1992 .

[15]  Peter Reinartz,et al.  Exploring the Potential of Conditional Adversarial Networks for Optical and SAR Image Matching , 2018, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[16]  Yawei Wang,et al.  Fruit Morphological Measurement Based on Three-Dimensional Reconstruction , 2020, Agronomy.

[17]  O. E. Apolo-Apolo,et al.  Deep learning techniques for estimation of the yield and size of citrus fruits using a UAV , 2020 .

[18]  P. Pathare,et al.  Bruise damage measurement and analysis of fresh horticultural produce-A review , 2014 .

[19]  Thijs Defraeye,et al.  The role of horticultural carton vent hole design on cooling efficiency and compression strength: A multi-parameter approach , 2017 .

[20]  Thijs Defraeye,et al.  Forced-convective cooling of citrus fruit: Cooling conditions and energy consumption in relation to package design , 2014 .

[21]  Daniel Cohen-Or,et al.  Volume graphics , 1993, Computer.

[22]  Tomas Pfister,et al.  Learning from Simulated and Unsupervised Images through Adversarial Training , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[23]  Bart Nicolai,et al.  Modelling the forced-air cooling mechanisms and performance of polylined horticultural produce , 2016 .

[24]  Dumitru Erhan,et al.  Scalable Object Detection Using Deep Neural Networks , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[25]  Chris McCarthy,et al.  Deep learning - Method overview and review of use for fruit detection and yield estimation , 2019, Comput. Electron. Agric..

[26]  Andrew Robson,et al.  Exploring the Potential of High Resolution WorldView-3 Imagery for Estimating Yield of Mango , 2018, Remote. Sens..

[27]  Namil Kim,et al.  Pixel-Level Domain Transfer , 2016, ECCV.

[28]  Alexei A. Efros,et al.  Image-to-Image Translation with Conditional Adversarial Networks , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[29]  Tomas Möller,et al.  A fast triangle-triangle intersection test , 1997 .

[30]  J. Carson,et al.  The cold chain in New Zealand – A review , 2017 .

[31]  Brijesh Verma,et al.  On-Tree Mango Fruit Size Estimation Using RGB-D Images , 2017, Sensors.

[32]  Dumitru Erhan,et al.  Going deeper with convolutions , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[33]  Lee Russell,et al.  Vehicle Detection in Infrared Imagery Using Neural Networks with Synthetic Training Data , 2018, ICIAR.

[34]  Bart Nicolai,et al.  Stochastic modelling for virtual engineering of controlled atmosphere storage of fruit , 2016 .

[35]  R. P. Singh,et al.  Modeling the forced-air cooling process of fresh strawberry packages, Part I: Numerical model , 2009 .

[36]  J. A. Martínez-Casasnovas,et al.  Mobile terrestrial laser scanner applications in precision fruticulture/horticulture and tools to extract information from canopy point clouds , 2016, Precision Agriculture.

[37]  J. R. Olatunji,et al.  A shape equation for Hayward Kiwifruit , 2019, International Journal of Food Properties.

[38]  D. Stajnko,et al.  Modelling of Apple Fruit Growth by Application of Image Analysis , 2005 .

[39]  P. Verboven,et al.  A 3D contour based geometrical model generator for complex-shaped horticultural products , 2015 .

[40]  P. Martin,et al.  Prediction of final weight for Actinidia chinensis ‘Hort1 6A’ fruit , 2003 .

[41]  Rozman Črtomir,et al.  Application of Neural Networks and Image Visualization for Early Forecast of Apple Yield , 2012, Erwerbs-Obstbau.

[42]  Manoj Karkee,et al.  Apple fruit size estimation using a 3D machine vision system , 2018, Information Processing in Agriculture.