Research On Dairy Goat Object Detection Algorithms Based On YOLOv3 In Intelligent Video Surveillance

The use of intelligent video surveillance in the sheep shed can realize early warning of abnormal conditions in dairy goats,improve management efficiency,and reduce breeding costs.Object detection is at the foundation of intelligent surveillance video,and plays an important role in the follow-up object tracking and object behavior analysis.In this study,the image of the dairy goat farm in the Animal Husbandry Base of Northwest A&F University was selected as the research object.Based on the Mobile Netv2-YOLOv3 object detection and the KG-YOLOv3 object detection algorithm,the object detection of the dairy goat in the dairy goat farm was realized.The main research contents and conclusions of this article are:1.Data collection and labeling.First introduced the time,location,and equipment of data collection.Then,based on the surveillance video of the farm,the key frames containing the dairy goats were screened out to construct the dairy goat data set.Thirdly,the label Img open source tagging tool was used to mark the specific location of the dairy goat in the picture,and The dairy goat data is organized according to the format of the PASCALVOC data set,and finally the data enhancement algorithm is introduced to realize the expansion of the dairy goat data set.2.Object detection based on Mobile Netv2-YOLOv3.In view of the model redundancy problem of the YOLOv3 algorithm,a lightweight object detection algorithm based on depth wise separable convolution is adopted.Through the change of network infrastructure and the use of depth wise separable convolutional neural network to replace the conventional standard convolution,the backbone is greatly reduced.The convolution operation part of the network reduces the overall calculation amount of the network and reduces the computational complexity.The core layer of Mobile Netv2 is a deeply separable convolutional layer,so Mobile Netv2 is used to replace YOLOv3’s backbone network.Experimental results show that the method in this paper effectively reduces the amount of parameters and calculations,and at the same time can achieve better detection results.3.Based on the KG-YOLOv3 Object detection algorithm.The K-means clustering method is used to determine the number and dimensions of the object candidate boxes on the data set,and the GIOU box regression loss function is used to improve the positioning accuracy of the dairy goat regression box.At the same time,the model is optimized through multi-scale training,and the improved YOLOv3 network is used to return the object ctategory and position,which realizes end-to-end target detection,and realizes the goal of monitoring video for dairy goat farms while taking into account accuracy and speed.Detection.Experimental results show that compared with SSD and YOLOv3,the method in this paper can achieve higher accuracy and faster recognition speed.The average accuracy rate can reach 94.58%,and the number of recognized frames per second is 52.86.In summary,based on the Mobile Netv2-YOLOv3 object detection and the KG-YOLOv3 object detection algorithm,this paper achieves accurate and efficient detection of dairy goat objects.It will play a positive role in accelerating the informatization and standardized management of the dairy goat breeding farm,improving its management efficiency and enhancing its market competitiveness.