Animal Social Behavior And Cognitive State Analysis System Based On Machine Vision

The development of computer hardware as well as further maturity of algorithm have facilitated the development of machine vision,and enabled machine vision to be applied in a wider field,which are affecting our every aspect.Mental disorders have bothered many people in their lives,therefore,accelerating the development progress for the drugs against infantile autism,alzheimer disease,depressive disorder and other mental disorders has always been an urgent problem that needs to be solved in our country.In order to achieve this target,we use mice,dog and other animal models to conduct the research,as the mental disorder would be reflected through behaviors,it is necessary to study animal internal behavior to analyze the related neural circuit.Meanwhile,since change of pupil diameter is closely related certain mental disorders,pupil diameter change often serves as the diagnosis index of mental disorder,thus,it is necessary to study the pupil diameter change to further analyze the cognitive state of animals.However,the traditional method has shortcomings in the following three aspects:(1)The limitation in real-time analysis and feedback regulation of animal behavior;(2)Unable to carry out accurate calculation for pupil diameter under multiple eye movement types;(3)Lack of appropriate algorithm when analyzing the social behavior of dogs.As a result,the traditional method is powerless and inefficient due to these shortcomings.The development of machine vision provides new solutions for analysis of animal internal behavior and recognition state,and studying internal behavior and cognitive stateof animals is of certain application value and practical significance.This thesis intends to explore an accurate analysis system for internal behavior and cognitive state of animals,therefore,in this thesis,a system and two algorithms were put forward,namely feedback animal behavior analysis system,eye movement detection algorithm and analysis algorithm of dog social behavior respectively.The main contents in this thesis included the following aspects:1.Developing feedback animal behavior analysis system based on Bonsai open-source framework: Feedback animal behavior analysis was realized based on Bonsai opensource framework,image processing algorithm was employed to carry out real-time analysis for mice behavior,and the feedback regulation was performed when the behavior feature variables of the mice were up to the threshold value,which optimized the shortcomings of traditional method in real-time analysis and feedback adjustment.2.Developing the eye movement detection system based on neural network: Deep Lab Cut tool kit was invoked for this algorithm,adopting neural network could accurately identify the pupil features under multiple eye movement types,and then based on Matlab platform,ellipse fitting and average method was employed to calculate the pupil diameter.By doing this,it optimized the shortcomings of traditional method in unable to perform accurate calculation for mobile eye movement and pupil diameter when the eye movement was shielded,and improved the calculation accuracy.3.Proposing the analysis system for dog social behavior: developing dog social behavior analysis algorithm based on neural network,image processing and SVM could achieve accurate analysis of dog social behavior,it overcame the shortcomings of lack of dog social algorithm in the market,heavy workload for artificial statistics of dog social behavior and low efficiency,and enhanced the analysis efficiency.The common target for analysis of internal behavior and cognitive state of animals is on how to perform real-time analysis and feedback regulation for animal behaviors;how to realize accurate calculation for the pupil diameter under multiple eye movement types;and how to achieve effective analysis towards social behaviors of dogs.In this thesis,the shortcomings of traditional method in the analysis of animal behavior and pupil diameter was studied,and optimization method for corresponding algorithm was put forward.Besides,in virtue of Bonsai open-source framework,image processing,neural network and SVM,the ability in analyzing internal behavior and cognitive state of animals was greatly enhanced.