According to the Ministry of Agriculture, Forestry, and Fisheries of Japan, the number of rearing houses has been decreasing in Japan in recent years due to lower business volumes. However, the number of rearing animals per house has been increasing, and in such situations, management of a herd of cows becomes very important. However, although systems such as a milking robot and an automatic feeding machine have been designed and implemented, an automatic measurement system to evaluate the body condition score (BCS), which is used for nutrition management of dairy cows, has not yet become popular. There have been many prior studies on this subject; however, none of them have succeeded in creating an inexpensive and highly accurate system that is capable of capturing images over a long period of time. The purpose of this study was to develop a system that continuously and automatically captures images of cows using a camera over a long period of time and to carry out a highly accurate determination of BCS. By attaching a three-dimensional camera to a sorting gate of a milking robot, we have developed a system that automatically captures images of cows as they pass through the gate. Data obtained from the captured images are automatically accumulated in a server. Thus, we successfully obtained a huge amount of data within two months. All parts of the image except the dairy cows were removed from the obtained three-dimensional data, and the three-dimensional data were then converted into two-dimensional images. Subsequently, the two-dimensional images were analyzed using a convolutional neural network, resulting in 89.1% of the answers being correct. When the acceptable error was ±0.25, the rate of correct answers is 94.6%, and the average absolute error, which is the average of the difference between the predicted BCS and the actual BCS for all test data, is 0.05. Although we used images that do not cover the entire body of the cow because of the position of the camera and the variation in captured parts (depending on images), we have successfully achieved a high accuracy. This promises that even higher accuracy can be achieved by automating the flow of data and carrying out the appropriate treatment of data to determine BCS.

1. [1] S. Arima and N. Kondo, “Cucumber Harvesting Robot and Plant Training System,” J. of Robotics and Mechatronics, Vol.11, No.3, pp. 208-212, 1999.
2. [2] M. Monta, N. Kondo, S. Arima, and K. Namba, “Robotic Vision for Bioproduction Systems,” J. of Robotics and Mechatronics, Vol.15, No.3, pp. 341-348, 2003.
3. [3] Y. Fukumoto, T. Hamada, J. Suyama, A. Yamamoto, and T. Naka, “Development of Flowering Stage Inspection Equipment for Small-Flowered Chrysanthemum,” J. of Robotics and Mechatronics, Vol.23, No.2, pp. 310-315, 2011.
4. [4] M. Kurita , N. Kondo, H. Shimizu, P. Ling, P. D. Falzea, T. Shiigi, K. Ninomiya, T. Nishizu, and K. Yamamoto, “A Double Image Acquisition System with Visible and UV LEDs for Citrus Fruit,” J. of Robotics and Mechatronics, Vol.21, No.4, pp. 533-540, 2009.
5. [5] A. J. Edmonson, I. J. Lean, L. D. Weaver, T. Farver, and G. Webster, “A body condition scoring chart for Holstein dairy cows,” J. of Dairy Science, Vol.72, pp. 68-78, January 1989.
6. [6] J. D. Ferguson, D. T. Galligan, and N. Thomsen, “Principal descriptors of body condition score in Holstein cows,” J. of Dairy Science, Vol.77, pp. 2695-2703, 1994.
7. [7] E. E. Wildman, G. M. Jones, P. E. Wagner, R. L. Boman, H. F. Troutt Jr., and T. N. Lesch, “A dairy-cow body condition scoring system and its relationship to selected production characteristics,” J. of Dairy Science, Vol.65, pp. 495-501, March 1982.
8. [8] T. Leroy, J. M. Aerts, J. Eeman, E. Maltz, G. Stojanovski, and D. Berckmans, “Automatic determination of body condition score of cows based on 2D images,” Precision Livestock Farming, Vol.5, pp. 251-255, 2005.
9. [9] J. M. Bewley, A. M. Peacock, O. Lewis, R. E. Boyce, D. J. Roberts, M. P. Coffey, S. J. Kenyon, and M. M. Schutz, “Potential for Estimation of Body Condition Scores in Dairy Cattle from Digital Images,” J. of Dairy Science, Vol.91, pp. 3439-3453, 2008.
10. [10] M. Hansen, M. Smith, L. Smith, I. Hales, and D. Forbes, “Non-intrusive automated measurement of dairy cow body condition using 3D video,” Proc. of the Machine Vision of Animals and their Behaviour (26th British Machine Vision Conf.), pp. 1.1-1.8, September 2015.
11. [11] R. Spoliansky, Y. Edan, Y. Parmet, and I. Halachmi, “Development of automatic body condition scoring using a low-cost 3-dimensional Kinect camera,” J. of Dairy Science, Vol.99, pp. 7714-7723, 2016.
12. [12] Y. LeCun, L. Bottou, and Y. Bengio, “Gradient-based learning applied to document recognition,” Proc. IEEE, Vol.86, No.11, pp. 2278-2324, 1998.
13. [13] R. B. Rusu and S. Cousins, “3D is here: Point Cloud Library (PCL),” IEEE Int. Conf. on Robotics and Automation (ICRA), 2011.
14. [14] R. B. Rusu, “Semantic 3D Object Maps for Everyday Manipulation in Human Living Environments,” German J. on Artificial Intelligence, Vol.8, pp. 345-348, 2010.
15. [15] I. Halachmi, P. Polak, D. J. Roberts, and M. Klopcic, “Cow body shape and automation of condition scoring,” J. of Dairy Science, Vol.91, pp. 4444-4451, 2008.
16. [16] A. Bercovich, Y. Edan, V. Alchanatis, U. Moallem, Y. Parmet, H. Honig, E. Maltz, A. Antler, and I. Halachmi, “Development of an automatic cow body condition scoring using body shape signature and Fourier descriptors,” J. of Dairy Science, Vol.96, pp. 8047-8059, 2013.
17. [17] J. Salau, J. H. Haas, W. Junge, U. Bauer, J. Harms, and S. Bieletzki, “Feasibility of automated body trait determination using the SR4K time-of-flight camera in cow barns,” SpringerPlus, Vol.3, p. 225, 2014.
18. [18] A. Weber, J. Salau, J. H. Haas, W. Junge, U. Bauer, J. Harms, S. Bieletzki, O. Suhr, K. Schönrock, H. Rothfuß, and G. Thaller, “Estimation of backfat thickness using extracted traits from an automatic 3D optical system in lactating Holstein-Friesian cows,” Livistock Science, Vol.165, pp. 129-137, 2014.
19. [19] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” Proc. NIPS, 2012.