The Fish Early-Warning Technique Of Water Quality Based On Computer Vision

With the development of modern industry and agriculture, human living standards havemarkedly improved, but the attendant environmental pollution is increasingly serious, beginning toreceive much attention. Frequent water pollution problem is one of the most serious problems andis urgent to be solved. Therefore, how to ensure water safety is becomming a hot topic of concernboth at home and abroad in recent years. Compared with the traditional physical and chemicalanalysis, the biological monitoring method is widely used because of its advantages and themethod combined with computer vision technology can provide an easy, fast and sensitive way forreflection of water quality. After the information on biological characteristics which reflect thewater quality has been acquired, how to make classification of large amounts of data is a difficultproblem. The development of artificial intelligence and pattern recognition technology provides agood idea.This topic focuses on the research of fish early-warning technique based on computer vision.The main contents of this paper are shown as follows:1. The paper elaborates two kinds of water quality monitoring and early warning methods:physical and chemical analysis and biological monitoring methods. The principle and advantagesof biological monitoring are described and the problems and shortcomings of the existing methodare analyzed.2. Combining computer vision technology with biological monitoring and taking accout ofthe stable environment and the color of fish body, a fish moving object detection method based onRGB component difference is used. The method is simple and easy to realize. The foregroundobject could be detected clearly. Then some school parameters quantification methods such asaverage swimming velocity and position coordinates are proposed, and used in quantifying featureparameters of fish school. These parameters could comprehensively reflect changes in fishbehavior from multiple aspects, thus to evaluate water quality instead of the tranditionalphysicochemical parameters.3. For this study, the experimental system is designed to simulate realistic water pollutionincidents and acute fish toxicity tests are finished with it. The zebrafish and several kinds of heavymetal toxicity solutions are used for acute fish toxicity tests to verify the effectiveness of the proposed method. The result shows that the fish have obvious acute stress responses in theenvironment of toxic substances and the quantified data of fish school from normal and abnormalstate have significant difference.4. In this paper, the theory of Support Vector Machine (SVM) is applied to the biologicalmonitoring field and a method for water quality assessment using multiclass Support VectorMachine is developed. Then a Support Vector Machine method based on mixed kernel functionand multi-parameters optimization is presented. In the study, the experimental data are processedand the classification accuracy of each experiment is shown. For three different degree of toxicity,the SVM classifier performs well. According to the different classification results, different alarmsignals can be sent out and different emergency measures can be taken to control the pollution. It’sof great practical value.