基于作物光谱的耕地质量反演.

【Objective】 Research on remote sensing evaluation of cultivated land quality is of great significance to realize real-time and effective management and protection of cultivated land quality, promote the development of modern and efficient agriculture and ensure national food security.【Method】 The present paper took Conghua area as the study area, the spectral parameters were calculated and obtained from GF-1 image data by using various transformation technologies. And the correlation coefficient and variance inflation factor were used to select the spectral response of the cultivated land quality indicators. The remote sensing spectral response models of cultivated land quality in Conghua district were constructed based on linear（ PLSR） and nonlinear （BPNN, GA-BPNN） models, and the optimal model was used to realize the rapid inversion and spatial mapping of cultivated land quality.【Result】 The crop spectral response indexes of cultivated land quality were FD-3, SD-4, EVI and LR-4, and the correlation coefficients with cultivated land quality were 0.661, 0.663, 0.714 and-0.674, respectively, indicating that red and near infrared in GF-1 image were the crop spectral response bands of cultivated land quality. The PLSR, BPNN and GA-BPNN models were used to retrieve the cultivated land quality respectively. Compared with the inversion accuracy, it was found that the GA-BPNN model could quickly realize the inversion of cultivated land quality（R~2=0.64,NRMSE=10.75%）. Spatial mapping and validation of cultivated land quality using optimal model were conducted, and the results showed that the crop spectral indicators extracted from GF-1 satellite image data based on GA-BPNN model could quickly and accurately retrieve the cultivated land quality at regional scale. 【Conclusion】 The use of remote sensing technology is helpful to quickly understand the quality of cultivated land at the regional scale, improve the existing discipline system of cultivated land quality evaluation, promote the development of the discipline, and provide a theoretical basis for the subsequent spectral inversion of cultivated land quality. [ABSTRACT FROM AUTHOR]