| In this paper, Qinghe reservoir water body is the research object. This paper combines Landsat 8 satellite OLI image data with the field sampling data, analysises the relationship between Qinghe reservoir water quality parameters and Landsat 8 remote sensing image data, chooses the best correlation band and band combination then respectively set up single band regression model, band combination regression model and the least squares support vector machine (LS-SVM) model to supervise Qinghe reservoir water quality situations. This article main research results are as follows:(1)Preprocessing Landsat 8 satellite OLI remote sensing image data in Qinghe reservoir, after the pretreatment of the image data in the study area we get 6 months’remote sensing image data which from June to November. And the image data cloud cover is 0% in June, July, October and November in the study area.The image data are part of could cover in August and September.This paper removed August 4 sets of data and September 9 sets of data in the process of inversion research for the cloud cover interference. Qinghe reservoir gets the water samples by ships field collection and uses chlorophyll meter tools to supervise the suspended solids and the chlorophyll a concentration of 20 samples of each month in the laboratory then acquires a total 120 groups data of the suspended solids and the chlorophyll a concentration in 6 months. In the determination of the suspended solids, the sample average level is the highest in June and the lowest is in November; In the determination of chlorophyll, the highest average concentration was in August and the average level is the lowest in November.(2)Using the data which is 56 sets of data from June to August and 51 sets of data from September to November analysises band sensitivity by the Pearson correlation analysis method, the results show that the correlation of the suspended solids, chlorophyll a and band reflectance value can be improved by the combination of band reflectance value in Qinghe reservoir. On the analysis of the correlation process of the suspended solids and the band reflectance in summer and autumn, (B2, B3 and B5) band combination shows the strongest correlation and the correlation respectively is 0.805 and 0.803 in order; On the analysis of the correlation process between chlorophyll a and band reflectance,(B4+B5)/2 and (B4-B5)/(B4+B5) band combination have showed the highest correlation with chlorophyll a in summer and autumn season and the correlation respectively is 0.771 and 0.805 in turn.(3)Establishing a single band band combination regression model, the band combination regression model and the LS-SVM model with the band reflectance values as the independent variable, the water quality parameters concentration value as the dependent variable to respectively does inversion quantitatively between the suspended solids and the chlorophyll a concentration in Qinghe reservoir in summer and autumn. Through the comparison of three kinds of models prediction results can be found that: the LS-SVM model predicted results are better than the prediction of the band combination regression model and the band combination regression model forecasting results are superior to the predicted results of single band regression model. In the suspended matter concentration prediction model in summer and autumn, the average relative error between predicted values and the measured values respectively is 8.012% and 2.498% in the LS-SVM model; In the chlorophyll a concentration prediction model in summer and autumn, the average relative error between the LS-SVM model predicted values and the measured values respectively is 5.022% and 5.993%. So we can come to the conclusion that:in the inversion process by Landsat 8 satellite OLI remote sensing image data in the Qinghe reservoir comparing with the band combination regression model and the single band regression model,the LS-SVM model is better for remote sensing monitoring on water quality situation in Qinghe reservoir. |
PDF Full Text Request