Research On Dynamic Tracking Algorithm For Free Living Animals

The behavioral analysis of experimental animals is widely used in the study of advanced central nervous function,such as exercise function and learning and memory.Other parameters analysis methods,such as electrophysiology and biochemistry,have the characteristics of overall response to the whole state of experimental animals.Therefore,animal behavior experiments have become one of the important research methods in neuroscience,behavioral pharmacology and other neuroscience fields.Early studies on animal behavior mainly rely on manual observation and recording,and there are many inconveniences.In recent years,with the development of video and computer image recognition technology,the automatic monitoring and recording methods of animal spontaneous activities have been greatly improved.However,because of the diversity and complexity of animal activities,the existing methods and techniques are still inadequate,mainly two points.First,the shape of the animals will change during the movement;second,the speed of the animal's movement may have a relatively large acceleration,for example,the sudden start of the animal and the sudden stop,etc.This paper aimed at the two challenges to mice as experimental animals,mainly studies the mouse position acquisition methods,the existing tracking method improvement,to create a set of animal behavior monitoring method,which is more suitable for use(KCF)kernel correlation filtering algorithm and joined the scale adaptive method,used in trajectory tracking in video image in mice,implements the free tracking in mice,and get the trajectory.The study of animal behavior provides a better monitoring method.The main research work of this paper is summarized as follows:Considering the movement of animals with a small target,the target system of change,the sudden changes in the speed of moving target problem,this article from the demand of real-time,stability and accuracy,the design of the experimental device,the related video dynamic target tracking algorithm research and determine the trajectory of were studied.Studied the tracking algorithms for several kinds of moving object detection,by a comparison of the commonly used in study of animal behavior meanshift algorithm and kernel correlation filtering algorithm,after analysis,Kernel correlation filtering algorithm is adopted,at the same time,because the animals in motion,shape change,joined the scale adaptive method for nuclear related filtering algorithm is improved.In this way,the animals are tracked in real time and output the moving coordinates,the moving video and the movement of the animals.Research shows that by adopting the proposed nuclear related filtering algorithm based on moving target detection and tracking method,realizes the location of the mouse movement track,the determination of direction and a description of the trajectory.The study of animal behavior provides a better way to monitor and provide effective information for related biomedical research,especially animal behavior.