Dynamic Model Of Fish Behavior And Its Application In Water Quality Monitoring

With increasingly rapid economic development, pollution caused by industrial and agricultural spills into water supplies has become common in recent years. Polluted water has serious direct and indirect impacts on human health. In order to improve the safety of drinking water, a series of monitoring systems are available to evaluate water quality with high cost. Therefore, it is highly desirable to provide a biochemical monitoring system able to respond quickly and accurately at low cost. Fish are rapidly becoming favored as convenient sentinels for behavioral assays of toxic chemical exposure. So, it is crucial to quantify fish behavior effectively and build a model mapping fish behavior to water quality.In this paper, Red Crucian Carp is used as a biological indicator. We analyzed changes in behavior using a toxicology experiment. Furthermore, we also studied the shoaling behavior and found that it could serve as tools for environmental risk assessment and analysis of toxicological impact. The main works are summarized as follows:First, we studied the dynamics model of fish behaviors that include a single fish and population of fish. The fish contour was obtained from the raw images on basis of a segmentation method and an edge-detection algorithm. In addition, we identified the centroid, tail point and head point of the fish in the raw images. Behavior of was quantified by measuring tail beat frequency and wall-hitting rate (frequency of collision using the head with the wall of the tank), both which are quantitative evaluations of escape behavior.Second, we put forward a TBF-based analysis to detect responses of red crucian carp to the toxins NaOH and three concentrations glyphosate. First, we examined the effect on tail-beat frequency (TBF) of exposure to two chemicals, and evaluated the ability of the novel behavioral parameter to accurately monitor water quality. To further refine our approach, the wall-hitting rate (WHR) was used to characterize behavioral avoidance after exposure. Overall, exposure to test chemicals at different levels induced significant increase in both behavioral parameters of the red crucian carp during1-hour exposure periods. Moreover, the TBF achieved better performance as an indicator when it was calculated in cases where the fish hit the tank wall. Collectively, this study further demonstrates the capacity of the TBF of fish to assess water quality in a reliable manner.Third, we implement an algorithm to track individual in a group fish. Tracking and segmentation of fish group targets were implemented on the basis of Kalman filtering. Furthermore, several common measures of group cohesion of fish were examined and used to monitor water quality. Our results suggest that the nearest neighbor distance is suitable measure for water monitoring.