Research On Discrimination Method Of Urban River Water Quality Level Based On UAV Hyperspectral Remote Sensing

The problem of water environmental pollution is becoming more and more serious,there are many rivers in the city,the water network is intertwined,and the ecosystem is fragile,so it is a place with high incidence of urban water pollution.The cost of traditional water sampling and testing methods is high.With the development of science and technology,remote sensing technology can be used as a supplement to existing monitoring methods due to its non-contact,global monitoring technical advantages,and improve the monitoring of the existing urban environment.However,due to the limitation of the temporal and spatial resolution of the existing satellite remote sensing detectors,it can only monitor a large area of water,which is difficult to meet the water quality monitoring needs of small and medium rivers in cities.In view of the fact that the existing satellite remote sensing is difficult to meet the application requirements of urban water quality monitoring of small and medium rivers,this paper applies UAV hyperspectral technology to the water environment monitoring application of urban small and medium rivers.First,we select the Jiading District and Qingpu District of Shanghai as the research areas to carry out field synchronization water quality spectral data collection experiments to obtain water quality sample data of various grades.Then,on this basis,a remote sensing method for urban river water quality classification based on the second-order differential fluctuation index of the spectrum was proposed.In order to apply the model to UAV hyperspectral remote sensing images,a technical process scheme based on UAV hyperspectral remote sensing water environment monitoring was designed,and three key points were reflectance inversion,water surface shadow correction,and flare processing.Technology for research.Finally,the feasibility and applicability of the water quality classification of small and medium rivers in the city are tested by the flying experiments of the Jiading and Qingpu District of Shanghai.The main work and innovations of this paper are as follow1)A remote sensing method for urban river water quality classification based on the second-order differential fluctuation index of the spectrum is proposed.Different from the existing remote sensing method for distinguishing water quality grades based on parameter inversion,this paper analyzes the spectral characteristics of each grade of water directly,and extracts the local extreme points of the second-order differential of the water spectrum to form the upper and lower envelopes of the fluctuation range of the second-order differential.The upper and lower envelopes are then used to calculate the fluctuation index reflecting the range of the second-order differential vibration.After analyzing the fluctuation indexes of each grade of water,it is found that the second-order differential fluctuation index of the water spectrum has a positive correlation with the water quality grade.The higher the level,the worse the water quality,and the higher the second-order differential fluctuation index of the water spectrum.Then the average fluctuation index of the characteristic bands with large differences in each level is selected as the feature vector to construct a remote sensing discrimination model for water quality level.After testing,the method proposed in this article The recognition accuracy of various water quality grades of the effect is better than the water quality grade based on parameter inversion.Subsequently,simulation tests were performed on the basis of the original data to test the effects of different spectral resolutions and signal-to-noise ratios on the model,and to analyze the effects of spectral resolution and signal-to-noise ratios on the water quality level identification method based on parameter inversion and the method proposed in this paper.2)For the pollution types of such polluted water bodies inferior to Level V,the spectral characteristics of three different types of water bodies inferior to Level V,including total phosphorus exceeding standard,ammonia nitrogen exceeding standard,and total phosphorus ammonia nitrogen exceeding standard simultaneously exceeded,were analyzed,and the differences in characteristics were analyzed.The foundation puts forward a spectral characteristic morphology index,and a remote sensing method for distinguishing water pollution types inferior to Level V based on the morphology index.After testing,the accuracy of the pollution type identification reached 75%.3)A water surface shading correction method based on neighborhood linear correction is proposed.Based on the full investigation of existing remote sensing image shadow correction methods,this paper proposes a shadow correction method based on neighborhood linear correction,which is transmitted through the relationship between the visible light and near-infrared reflectivity of normal water bodies in the neighborhood,and calculates the shadow of the water body in the shadow area.Compensation amount.Finally,the shadow area correction is completed according to the spatial filtering to eliminate the boundary of the area.The verification experiment is carried out.The method in this paper is compared with the traditional HSI-based color space and the inner and outer contour matching method.The results show that the method can effectively perform the shadow area.Correction,the ability to maintain texture features is equivalent to that of traditional methods,but the ability to maintain spectral features is better than that of classic methods.4)Propose a method for recovering water surface glint based on discrete cosine transform.This method first constructs a global optimization function from two aspects of fidelity and continuity,and then uses discrete cosine transform to iteratively solve the optimal pixel value of the area to be restored,and then combines the method in this paper with CDD based on partial differential equations and block filling based on The Criminisi method of Criminisi is used to conduct comparative experiments on UAV remote sensing images with different proportions of the area of the glint area.The results show that all three methods can recover the small area of the glint area very well.As the area of the glint area increases,CDD The method draws on the principle of thermal diffusion in physics.When the area of the glint area is larger,the phenomenon of regional blocking occurs,and the larger the area of the glint area,the more obvious the phenomenon of blocking;while the Criminisi method adopts the principle of similar block matching,which is easy to appear.For the phenomenon of mismatch,the method in this paper is better than solving the area to be processed directly,and it is less affected by the area of the glint area,and the ability to maintain the spectral characteristics is also better than the other two classical methods.