Research On Recognition Of Abnormal Fish Behavior For Water Quality Monitoring

Biological monitoring is widely used in water quality monitoring and safety warning systems. It assesses the pollution of water quality using physical characteristics and behavior reaction of fish to the environment, where the hydrobios lives, and then provides scientific basis for the assessment of water quality from the biological point of view. It can reflect more intuitive, real-time the changes of water quality that because of the closely relation between the changes of hydrobios's behavior and the changes of their surrounding water environment. Therefore, the physiological characteristics and behavior reactions of hydrobios usually regarded as important research indicators in biological water quality monitoring system. How to analysis for the indicators fast and effectively and achieve anomaly detection will become a key issue between biological indicators and water quality, and a key part of water quality safety warning system.In this paper, we used the fish in the water as indicator biology, the behavior trajectory as carriers, and the behavior indicators as the research object, studying the methods of description and distinguish of the behavior of fish. The contributions of this paper are listed as follows:First, study the methods of description of fish. Using the behavior indicators as the definition elements of multiple element grid, Conversions the expression of coordinates set into the description of multiple element grid of fish behavior trajectory. The method can save the storage space and avoid the error of judging the behavior of fish using single indicator.Second, study the recognition methods for abnormal behavior of fish. Describes the behavior of fish using the multiple grid methods, trains and study the fish behavior pattern of normal for behavior detector using r-chunk negative selection algorithm learning from selection principle of biological immune system. And then we could distinguish the abnormal behavior from all of the fish behavior. The method proposed in this paper could distinguish lots of abnormal behavior using few of behavior pattern detectors.Third, analyzed the characteristics of behavioral indicators, and verified the indication for the behavior by experiments on fish.