Water Quality Monitoring Algorithm Optimization And System Integration Research In Dynamic Scenarios

In this paper,the application of Gaussian-Hermite moments in image analysis and computer vision is studied.The moments are also applied to the water-quality monitoring in dynamic scenes.First of all,relevant researches on the monitoring of water quality,mainly water surface monitoring at home and abroad are analyzed and summarized,their merits and demerits are compared,with orthogonal Gaussian-Hermite moments as the tool of analysis.Secondly,the nature,algorithm and efficiency of orthogonal Gaussian-Hermite moments are analyzed in a detailed way,dividing the region according to the characteristics of the water surface region and the targets of interest.Orthogonal Gaussian-Hermite moments are used in local optimization analysis,meanwhile moment template is used to improve the efficiency of computation.On the time dimension,moving floating objects can be tested,on the spatial dimension,the condition of the current frame can be tested,meanwhile reducing the interruption of noise and strengthening the regional target of interest.Thirdly,through color difference analysis in the HSI color space,by integrating Gaussian-Hermite moments,the floating object can be obtained accurately.Through analysis on the images of region of the water surface of floating objects,it is concluded that in H component graph of color tones,the floating objects are scattered in area of low color tones,in I component graph of intensity,the floating objects are scattered in areas of high intensity.In S component graph of saturation,the floating objects are mainly scattered in areas with high saturation degree.The target area can be obtained by integrating the three independent components of H,S and I with threshold segmentation.Finally,as orthogonal Gaussian-Hermite moments have excellent image reconstruction ability,take the orthogonal Gaussian-Hermite moments as an importantcharacteristic vector,compare RGB space and HSI space in the space and color,compare characteristic values extracted from the characteristics of moments in the past such as the wavelet transform energy decomposition coefficient and Hu invariant moments,and conduct training and identification with the same categorizer to support vector machine.The experimental results show that HSI space is superior to RGB space,the characteristics of Gaussian-Hermite moments are superior to wavelet transform energy decomposition coefficient,the invariant moments of Gaussian-Hermite are superior to Hu invariant moments,the identification rate is improved.Making use of hybrid programming of MATLAB?SQL,and Net in systematic integration,achieved part of the functions of the modules.