Research On Deer Body Pose Behavior Recognition Based On Target Features

In the modern breeding industry,scientific breeding technology is constantly advancing,and the intensive breeding of sika deer occupies a major position in the sika deer breeding industry.Traditional manual monitoring in the sika deer intensive breeding industry not only requires a lot of work and high labor costs,but also easily causes the sika deer to produce a stress response during the monitoring process,which makes it difficult to monitor and identify.Therefore,in order to solve the above problems,this paper analyzes and summarizes the research status of domestic and foreign livestock body gesture behavior recognition,and uses machine vision technology to realize the research of sika deer body gesture behavior recognition in sika deer farms.In order to improve the segmentation effect and ensure the completeness of sika deer target extraction,this paper proposes an improved OTSU segmentation algorithm based on color space,which uses color space to process the sika deer image and then uses the segmentation algorithm for target extraction,and performs morphological processing such as expansion and erosion.Optimize the extracted target,and compare the results of K-means,color space processing,and binary filter segmentation.It can be seen that the improved OTSU segmentation algorithm can clearly segment the sika deer target and background,which solves the problem of the traditional segmentation algorithm when it is easy to segment the sika deer.Problems such as serious noise interference,inaccurate extraction of target individuals,and many voids inside the extracted target are caused.In order to achieve real-time and accurate monitoring of sika deer’s daily body posture behavior,this paper proposes an improved VGG recognition and classification algorithm based on dense link blocks.Comprehensively considers the daily behavior characteristics of sika deer,selects different behavior images of sika deer to build a data set,and uses the improved algorithm to identify and classify.The recognition and classification results of Res Net(Residual Neural Network),Alex Net and VGG(Visual Geometry Group)show that the improved algorithm is superior to other deep networks.It can optimize feature extraction and accurately recognize sika deer body posture behavior,which solves the problem of low real-time performance of manual monitoring of sika deer behavior.Problems such as low accuracy.