Construction and Monitoring of a Quantitative Inversion Model for Conductivity Using Unmanned Aerial Vehicle Remote Sensing Based on ELM Algorithm

Abstract

Currently, unmanned aerial vehicle remote sensing technique has wide application in agricultural soil salinization monitoring. However, there are still issues such as high costs and complex data preprocessing. Based on this, this study uses unmanned aerial vehicle multispectral images and ground synchronous collection of soil conductivity as the data source. The spectral and texture features of unmanned aerial vehicle multispectral images were extracted using the gray-level co-occurrence matrix and Gabor two-dimensional filters. A quantitative inversion model for soil conductivity using unmanned aerial vehicle remote sensing was established by combining extreme learning algorithms. A quantitative inversion model for soil conductivity based on extreme learning machine algorithm was constructed by integrating spectral and texture feature information. The determination coefficients of its modeling set were 0.77, I0.82, and 0.86, respectively, and the prediction set was 0.73, 0.77, and 0.80, respectively, with the highest prediction accuracy. Therefore, by integrating texture feature spectrum index information and combining extreme learning machine algorithm, an unmanned aerial vehicle remote sensing monitoring model for soil conductivity is constructed, which can achieve high-precision monitoring of soil salinization. This study has certain technical guidance value for unmanned aerial vehicle remote sensing monitoring of soil salinization.