Research And Application Of Mouse Behavior Analysis System Based On Target Tracking And Posture Recognition

Brain function involves sensory perception,emotion,cognition,learning,memory and motor control,and all functions ultimately play a role in regulating animal behavior.Therefore,if we want to understand how the brain works,we must consider its function in the context of behavior.Combining mouse electrical activity data with behavioral data contributes to understand the role of specific brain regions or neural circuits in regulating behavior.In this context,a large number of studies have emphasized the need for an accurate,comprehensive,and automated behavior analysis system that can provide stable and comprehensive behavior evaluation with minimal manual intervention.The current animal behavior analysis system has a single function and is expensive.For example,it requires complicated operation with the help of a hardware device.Or it is only suitable for specific experiments or requires geometric modeling,which is not universal.This paper systematically studied the behavior analysis algorithm of mice and built a behavior analysis system based on target tracking and posture recognition.The multi-target tracking system described in this paper does not depend on any external markers.Different pre-processing methods and background modeling algorithms can be selected according to the different experimental conditions to accurately extract the target foreground.Karman filter and Hungarian algorithm are used to match the corresponding trajectory for each individual.Due to the occlusion because of the interaction between multiple targets,resulting in matching errors,or even error propagation.In this paper,a hash algorithm is used to establish "fingerprint information" for the image to perform second correction,correct the matching errors that occur after target interaction,and improve the accuracy of target.tracking.In addition,the multi-target system is equipped with a concise and clear software interface.Users can load multiple data and set different processing parameters for each data,and then perform batch operations.Mouse movement parameters combined with electrophysiological data can reflect information such as the mouse’s mobility,position preference,and position cells,and the posture change of the mouse reflects more specific psychological information.The mouse posture recognition and classification system described herein includes posture marking software and posture classification algorithms,both of which can be used alone or complement each other.Users can use posture marking software to mark the posture of mice at different times,combined with the electrophysiological data recorded at the same time,preliminary analysis of the correlation between specific neural circuits and posture changes of mice.This tool is useful for screening mutant animals or studying some unexpected way to change the neural intake of behavior,that is,it is particularly helpful in situations where people cannot know a priori which behavior to analyze.The data generated by the posture labeling software is used as a training set,and the foreground pixels of the mouse are extracted using the background modeling method described above,and the posture recognition network is trained,thereby obtaining posture classification results of a large amount of data.The experimental results show that the mouse behavior analysis system described in this paper can be used in a variety of animal behavior experiment programs.It has comprehensive functions,high accuracy,and easy to use.Even non-engineering personnel can easily use it.It can play an important role in animal behavior analysis and neuroscience research.