Evaluation Of Geochemical Quantitative Inversion Results Based On Gradient Field Of Remote Sensing Images

The quality of remote sensing inversion results is related to the application and development of remote sensing,and its authenticity evaluation is of great significance to improve the accuracy of quantitative remote sensing inversion and prospecting efficiency.At present,in the test of quantitative remote sensing inversion results,the mainstream evaluation indexes such as RMSE and R~2 are based on the principle of mathematical statistical analysis.However,this evaluation method focuses more on the matching degree of points.When evaluating the inversion results of remote sensing prospecting,the determination coefficient(R~2)and correlation coefficient(R)are not high,which are inconsistent with the actual results.As a result,the inversion results that can correctly express the surface environment information lack credibility because they do not meet the accuracy requirements.Image gradient algorithm is very sensitive to the change speed and direction of mineralization anomaly in image space pixel,so this paper proposes an idea based on two-dimensional image gradient field.From the perspective of spatial similarity,this paper takes Qishitan mining area as an example to explore the evaluation of geochemical quantitative inversion results based on remote sensing image gradient field.The main work and conclusions of the paper are as follows:(1)Using WorldView-2,ASTER remote sensing data,based on multiple linear regression,BP neural network inversion method for quantitative remote sensing inversion modeling and visualization operations,through different inversion elements,different inversion methods,different remote sensing data sources,the obtained inversion image constructs the authenticity evaluation data set,and the anomaly map of the ground geochemical exploration data is used as the reference image to prepare the data for the image similarity judgment for the authenticity evaluation of the quantitative remote sensing inversion product.(2)In addition,according to the perspective of comparing the spatial area,spatial location and spatial distribution of mineralization anomalies in remote sensing inversion image and geochemical anomaly map,we calculates the superposition rate of positive anomaly response area,and obtains the macro judgment of whether the remote sensing inversion product is similar to the reference image in space by using the subjective evaluation method;Combined with the mathematical statistical evaluation indexes(RMSE,R,R~2),the basic test results of remote sensing product accuracy are obtained.Provide reference and comparison for evaluation methods.(3)The final evaluation is obtained by calculating the image gradient structure similarity,gradient amplitude similarity and image similarity combined with scale-invariant features as the three factors of authenticity evaluation and combining the results of basic inspection.In this study,the traditional point-to-point accuracy test evaluation method is upgraded to image-to-image test,and the method pays more attention to the relationship between pixel and pixel,local scale and local scale,which has a good enhancement effect on image features.In addition,the similarity evaluation method based on image gradient field can effectively enhance the contrast between mineralized anomaly areas and background areas,which is conducive to improving the accuracy of evaluation.It is also applicable to the evaluation of inversion results of different elements,and has a broad application prospect.