Fish Behavior Based On Computer Vision Testing And Research Structural Features

Along with the fast development of the computer technology, the computervision technology has become an important method to study the group behaviorof fish. Research on behavior detection and structure characteristics of fishschool based on computer vision technology are very important to improve theinformatization level in modern aquaculture.In this paper, by means of computer vision and image processingtechnology, based on a statistical method of motion features, the anomalydetection of fish school behavior is studied. The zebra fish is selected as thestudy object in this paper. Firstly, based on the foreground object detectionmethod with threshold value method， the backgrounds are removed from theoriginal video images to reduce the influences of noises. Secondly, by theLucas-Kanade optical flow method, which is based on the local deferencemethod and has better performance, the vectors of motion behavior can beobtained in different temporal and spatial conditions. Thirdly, from these data,the joint histograms and joint probability distributions of turning angles andvelocities are calculated. Since from the practical point of view, the anomalybehaviors of fish school mainly include the change of the moving velocity andthe chaos of the moving direction, this is the reason to select turning angle andvelocity as the characters to analyze. At last, the NMI and the LDOF methodsare applied to study the anomaly detection of fish school behavior. By choosingproper threshold values, the NMI and the LDOF methods can implement thedetection to the behaviors of zebra fish school. The experiments showed that theNMI method and the LDOF method for anomaly detection of fish schoolbehavior can achieve high accuracy, which implies that all the two methods havebetter effects.The structure property of fish school is also studied in this paper. Bymaking use of the target tracking algorithm, we can obtain the individual fish’sposition at every moment in the video. Based on these data, we first study the space relationship of individual fish. The research result shows that the distanceof nearest neighbor is about one fish body length. Moreover, the correlationanalysis of the relative position of individual fish is discussed. Then, we alsostudy the social network property of the fish schools. For the fish schools withdifferent group sizes, we establish social networks, respectively, and study therelationship between certain social network indexes and group sizes.