The deep learning based image target detection technology was introduced into the target detection of individual breeding images,which can improve the intelligent analysis technology of breeding video images and enhance scientific breeding capabilities.YOLO V3 algorithm was applied to pig detection in videos.Based on the actual situation of animal husbandry,we designed category selection,occlusion processing and image enhancement,and implemented a pig image target detec-tion algorithm based on deep learning technology.The algorithm proposed was used to train,verify and test the collected pig image data.The missed detection rate of YOLO V3 was about 6％,and the error detection rate was about 1％.The results show that the algorithm has good detection accuracy and fast detection speed.On our dataset,we found that mAP-50 of our algorithm was improved by 7％-8％and detection speed was increased by about 5 times compared with Fast r-cnn.mAP-50 and detection speed of our algorithm are similar to SSD.However,because the Yolo V3 algorithm is sim-pler than SSD algorithm,the algorithm migration compatibility is higher.The results show that YOLO V3 algorithm has a fast detection speed and is suitable for the requirements of intelligent recognition engineering target detection in animal husbandry images.
维普
万方数据