Soil Salinity Inversion Model Based on BPNN Optimization Algorithm for UAV Multispectral Remote Sensing

Rapid and accurate inversion of soil salinity is a key scientific problem that needs to be solved urgently. Due the accuracy UAV multispectral remote sensing based on back propagation neural network (BPNN) low, in this study, used image field measurements 60 surface as data sources, 16 indexs were constructed using extracted spectral reflectance, performed gray relation analysis screen index features after applying film removal construct BPNN model. Particle swarm optimization (PSO), thinking evolutionary algorithm (MEA), genetic (GA) applied optimize inverse model, respectively, capabilities four algorithms compared evaluated best algorithm. The results showed GRA variable screening can effectively remove redundant information parameters reduce complexity model; PSO, MEA, GA improve robusticity has effect terms model effect, followed by MEA PSO algorithms; PSO-BPNN, MEA-BPNN, GA-BPNN models are better than which achieves R 2 0.6659, RMSE 0.0751, RPD 2.0211. This approach solve monitoring issues inversion.