Research On Individual Recognition Algorithm Of Cow In Unconstrained Scene

With the improvement of living standards,people pay more attention to diet structure and health.As an inexpensive supplemental program of protein,the demand for dairy products is rising rapidly,large-scale dairy farming industry is thus rapidly rising and gradually expanding.The introduction of modern technology in dairy farming to make it develop in the direction of standardization,intelligence and refinement o f healthy farming and management,is the trend ofsustainable development of dairy farming industry.At present,China's dairy farming intelligent fine management level is still in the exploratory stage,so it is particularly necessary to carry out the relevant key research of farming.Among them,the intelligent dairy cow individual recognition technology is one of the core research content,accurate recognition of dairy cow and its tracking management is an important means to improve the efficiency of dairy farming.In view of C hina's large-scale dairy farming fine breeding level is low,the degree of standardization and intelligent is not high and so on.This paper takes the cow head image as the research object,and based on Bag of Feature(BOF)model in image recognition algorithm,to carry out the dairy cow individual recognition technology research.In view of the series of shortcomings in this model,this paper focuses on the improvement of recognition accuracy and calculation performance of the model.The main research results are as follows:(1)Selecting histogram intersection kernel for SVM classifier,and it is found that the kernel function can improve the recognition accuracy of the algorithm compared with the conventional Gaussian kernel function.Furthermore,it is found that the introduction of Spatial Pyramid Matching(SPM)principle in BOF model can further improve the accuracy of algorithm recognition.(2)Introducing the optimized 31-Dimensional Histogram of Oriented Gradient(HO G)feature as the image feature descriptor to further improve the computational performance under the premise of ensuring the recognition accuracy.In addition,the optimized HOG feature improves the recognition accuracy and computational performance compared to the traditional HO G feature as the image descriptor.(3)Using the gray logarithmic transformation to enhance the processing of the cow head imageto improve the image visual effect,increase the image contrast,and especially to enhance the details of the dark area image.And thenfurther improving the recognition accuracy and computational performance.The experimental data set consists of the cow head images taken in unconstrained scene in a cattle farm,the recognition accuracy is measured by the correct percentage,and the computational performance is measured by the run time of algorithm.