Research On Fish's Movement Model Based On Machine Vision

Biological monitoring is widely used for environment quality evaluation based on the advantage of predictability, sensitivity and warning to the pollution of environment, compared to the chemical analysis. Biological water quality monitoring is very important in the water safety protection and emergency supervision. Especially, the use of fish as the main vector to the biology-line monitoring system has been developed and implicated to the actual projects. The mechanism of the existing water quality monitoring system is by catching and analyzing of fish respiration, weak pulse and other physiological parameters, or using of fish's movement parameters for warning the water quality. Therefore, quickly and accurately catching the fish's parameters is the key point to the biological water quality monitoring systems.However, the method of catching the fish's parameters of the existing systems is so complexity while the parameter is always too single. This paper propose the using of machine vision to build fish's movement model, then catching several fish's movement parameters by deeply analyzing this model. The new method well resolves the problem mentioned earlier, and provides accurate rich parameter which can be used to analysis the fish's exercise movement to the monitoring system.The research of this paper is to build a fish's movement model which based on the fish's skeleton line, use this model to catch the fish's movement parameters such as fish's tail beat frequency, velocity, trajectory and etc. The main research contents: fish's body image segmentation, extracting the center line of fish's skeleton, building and analyzing the fish's movement model to catch the fish's movement parameters and so on. The study can be summarized as follows:The first is studying the color image segmentation algorithm, proposing a new segmentation algorithm by combining the 2-D Otsu method based on differ of the gray-level histogram segmentation algorithm and the color clustering segmentation algorithm according to the fish's image features to get a good segmentation image of fish.The second is through analyzing the traditional characteristics of skeleton extraction algorithms, bringing forward a simple and rapid methods to extract fish's skeletons. This method is using the fish's boundary points to determine the fish's center point, and then using the center point to calculate the main axes, at last using the main axes to extract fish's skeletons.The third is to build a fish's movement model which based on the line of fish's skeleton. According to the features of fish's movement and on the based of accurately segmentation of fish's image beside the rapid extract fish's skeletons, proposing a fish's movement model , and analyzing this model to well catch the fish's movement parameters which mentioned above.By accurately segmenting the fish's image, rapid extracting fish's skeletons, correctly building fish's movement model and accurately analyzing the model to catch the movement parameters, the research is to provide the rich data for the analysis of fish's abnormalities movement of our water quality monitoring system.