An Application For Motion Detection In Sheep With Dense Optical Flow


Tayyar Bati C., Ser G., Karaca S.

2nd International Conference on Artificial Intelligence Studies, 17 - 18 September 2022, pp.125

  • Publication Type: Conference Paper / Summary Text
  • Page Numbers: pp.125
  • Van Yüzüncü Yıl University Affiliated: Yes

Abstract

Today, the world population is increasing rapidly and the demands for animal products are increasing at the same rate. Therefore, productivity needs to be increased while maintaining animal health and welfare standards. Animal behavior traits are shown as an important indicator of animal welfare in precision livestock. Monitoring and classifying animal behavior plays an especially important role in animal welfare, as changes in animal behavior can often herald conditions such as lameness, injury, estrus, an d pregnancy. Today, the methods used for the classification of animal behavior are generally contact technologies and non vision methodscontact detection methods using offer an important albased methods using sensor computer vision technologies . Computer ternative in this field, due to the disadvantages of sensor technologies, which are used extensively in precision livestock farming, such as injury and stress in animals. In farms monitored by camera systems, deep learning methods such as image classificat ion, object detection or image segmentation are frequently used in image processing. In this study, motion detection in sheep was performed through the opencv library by pre processing a video data for use in animal production, dense optical flow. In this process, it is determined whether each frame in the videos contains motion information by calculating the distance that the pixels move with the “cv2.cartToPolar” option. As a result, it is possible to say that dense optical flow can provide an important a sheep behaviors in deep learning studies as a predvantage in determining and classifying process.