Research On Water Quality Monitoring System Based On The Behavior Of Fish School

Drinking water safety is an important and concerned issue of the society. When the water pollution occurs, the key to prevent polluted water into drinking water is to quickly alert. With the biological water quality monitoring system concept putting forward, more and more scholars begin working on biological water quality monitoring technology. Use aquatic organisms for water quality early warning, because aquatic organisms are very sensitive to environmental changes of water. When the water is contaminated, the aquatic organisms’ behavior, body color or morphology will present significant changes. By using computer vision technology to capture these changes and make quantitative analysis, we can easily find out the relationship between pollution water and the behavior of aquatic life. Finally achieve the goal of water quality warning.This research use red carp as an indicator for water monitor. We use a video camera to shoot the swimming behavior of multiple red carps. Then we calculate the 3d coordinate of each fish by combining computer vision technique with the image processing technique. After getting the 3d coordinate, we use Kalman filtering algorithm to track the fishes. Then we extract and quantify the eigenvalues of fish’s behavior, such as activity, vergence, average height etc. By combining these characteristic values with fish behavior to get the semantic model, and finally the semantic model is applied to water quality monitoring platform. The main research works are summarized as follow:First, study the relationship between fish school behavior and water quality. By studying the behavior of fish, such as physiological behavior, chemotactic behavior, social behavior, we understand the basic form of fish behavior. By reviewing the researches on the change of the fish school’s behavior in a variety of water quality, such as metal compounds water quality, hot water quality, anoxic water quality, acidic and slightly alkaline water quality, proposed the relationship between water quality and fish behaviors. In addition we extract the motion features of fishes.Second, study the three-dimensional tracking of fishes based on Kalman filter algorithm. The behavioral characteristics of fish that obtained from the traditional two-dimensional space lost a lot of real and useful information. After extended to three-dimensional space, you can get more comprehensive fish behavioral characteristics, such as the average height of fishes in the water, the fish’s activity, the spatial distribution of fishes. In this thesis, we use an auxiliary plane mirror with a monocular camera to capture the three-dimensional video data. Then based on the imaging principle of the object under water combined with camera imaging principle, plane mirror imaging principle and computer vision technology, we deduce the formulas of three-dimensional coordinates of the centroid of fish. At last, we use the Kalman filter algorithm for real-time three-dimensional coordinate prediction and tracking of fish’s centroid, and then extract the average speed, the average amount of activity, the average height and group vergence of fish schools. Also, we analyze the applicability of these typical fishes’ behavior parameters in the application of the water quality monitoring system.Third, this paper presents a semantic model of fish behavior based on multi parameter. After thresholding of average activity, average height,fish swimming vergence and other parameters, the range of each parameter can be obtain. Then the value of different parameters at the same time are combined to form an output list which is corresponding to the social behavior of fish, establishing the relationship between the characteristics and behavior to analyze the semantic meaning of each relation. Finally, semantic model is obtained.Last, this paper study the water quality monitoring system based on fish social behavior, and preliminary design a water quality monitoring system based on semantic model of fish behavior. By characteristic value combinations of fish activity, fish distribution and average height, semantic output is obtained after input through the semantic model, determining whether the current fish behavior is normal, thereby the water quality conditions is estimated.