Analysis Of Feeding Behavior Of Fish Based On Computer Vision

Our nation is the biggest aquaculture country in the world.However,China's traditional aquaculture farming methods still plays an important role.This will lead to a waste of hunman resources and land resources.In recent years,more than 90% of the fish farming workers prefer to apply RAS(recirulating aquaculture systems).Because RAS(recirulating aquaculture systems)not only create a good rearing conditions and feed utilization for the good growth of the shoal,but also provide a necessary prerequisite for the sustainable development of the aquaculture.Although RAS has a total of more convenience,nevertheless,But it still has a lot of problems: the over-feeding events continues to affect the welfare of shoal,the operation of systems also influence the growth and survival of shoal,in addition,the water quality of the system increasing the risk of diseases and other problems.The most direct and effective way to solve this problem is to determine the amount of feeding activities according to the appetite of the shoal.It is of great significance to explore the feeding behavior of fish schools in order to improve the economic benefits of aquaculture.In actual operation,manual observation is an effective method.But,on the one hand,this way requires fish-farming workers to increase labor and time costs.It is of great significance to explore the feeding behavior of fish schools in order to improve the economic benefits of aquaculture industry,collective behavior is one of the important behaviors of fish.Exploring the dynamic behavior in fish school is the basic content in the fish ethology and has shown great significance for understanding the collective behavior mechanism and protection of fishery resources.This thesis presents the methods of fish school video motion tracking,the main research contents and results are shown as follow,In this paper,based on the precise feeding of fish,a method of fish feeding behavior analysis and evaluation based on computer vision technology was proposed.This paper studies the following three aspects:(1)The first collection of fish swim state and feeding state images using gray differential statistics,gray level co-occurrence matrix method and Gauss Markov random field model based on texture feature extraction of fish,and then use principal component analysis dimensionality reduction PCA algorithm to extract fish texture information,more feature information using support vector machine(SVM)algorithm based on the recognition of fish feeding state;(2)secondly,the fish feeding pictures were processed using the mean background modeling method to get background no fish pictures,extract the texture features and the prospect of fish,gray level co-occurrence matrix to extract fish were compared.The feeding intensity assessment of fish,and compared with the traditional method of area method,by contrast to the gray level co-occurrence matrix Correlation between fish feeding activity intensity characterization of feeding activity intensity and area was characterized by fish to reach 0.8942;(3)Finally,the paper attempts to apply adaptive neural fuzzy inference system(ANFIS)model is applied to the fish feeding quantity,and makes a preliminary exploration of the model,to expect in the next stage of the this model is applied to practice.