Anomaly Water Quality Monitoring Based On Fish Movement

Water quality and environmental issues have focus the attention of China's society,in recent years a variety of environmental pollution incidents,seriously affecting people's normal social activities and economic activities.At present,for the study of water quality monitoring and early warning,many scholars use biological water quality monitoring method to monitor the water quality according to the changes of physiological characteristics and behavior characteristics of aquatic organisms.However,due to the existence of multi-factor,non-linear,high-dimensional and other complex relationship between the evaluation factor and water quality,the effect of monitoring and early warning is not ideal.In order to carry out comprehensive and scientific on-line monitoring of the water environment,this paper studies the problem of abnormal water quality monitoring based on the characteristics of fish movement.The relationship between water quality and fish movement characteristics is studied in many aspects.The paper main contents of the study include:(1)Based on the computer vision theory,the fish culture was preliminarily applied by the bilateral filtering method,and the spatial distribution of the fish was detected,segmented and tracked.And trajectory of movement,lay the foundation for the analysis of fish movement behavior.(2)According to the Delauny triangulation method,the fish center center of gravity algorithm is improved and the relevant parameters of the fish movement dispersion are introduced: the variance,the distribution area,and the center index of the trajectory related parameters.At the same time,the curvature,The traditional characteristic parameter velocity,the acceleration and the distance are eight characteristic parameters,and the water quality identification feature data set is established.(3)According to the size of the experimental device,the appropriate number of fish groups were selected,and several groups of different pollutants and different concentrations of abnormal water quality were compared.The characteristics of XGBoostclassifier are studied,and the parameters are optimized by grid search method.The water quality monitoring scheme based on XGBoost classifier is designed,and the recognition results of support vector machine classifier are compared and analyzed.The experimental results show that the model based on variance,distribution area and center distance index has high efficiency and robust recognition effect as the model of water quality evaluation.