PSIV-A-8 Development of Automated Computer Vision Systems for Evaluating Dairy cow Body Weight and Body Condition Scores

Abstract

Animal dimensions are important indicators for monitoring the growth rates, diet efficiency, and health status of animals. A computer vision system is one of the recently emerging precision livestock farming technologies to overcome the previously unresolved challenges pertaining to labor and cost. It is common practice to collect top-view depth images for evaluating body mass or conformation traits. However, in the depth image data acquisition process, some manual interventions are often involved for controlling a camera from a laptop computer or detailed steps for automated data collection are not described. The objective of this study was to evaluate the utility of automated top-view dairy cow depth data collection methods using the picture- and video-based methods. Both methods automatically performed animal detection, trigger recording, capture depth data, and stop recording. The depth sensor camera was mounted 2.74 m above the ground over a free walk-through path between the milking parlor and feeding area. A total of 150 Holsteins and 100 Jersey cows were evaluated. A pixel location that depth is monitored was set up as a point-of-interest. The picture-based method takes the only pre-determined number of frames instead, while the video-based method takes a sequence of frames as a video. More than 90% of cows were successfully captured in both picture- and video-based methods. The success rate further improved to 98% when the video-based method was combined with the infrared cut lens. Overall, the video-based method yielded more accurate results for collecting depth data on cows. However, the picture-based method can be improved by increasing the number of pictures taken. We also discuss the utility of the developed automated system for evaluating body weight and body condition scores.