Inversion of free ferric oxide content in surface soil based on HJ-1A hyperspectral images

Abstract: The real-time monitoring of soil free ferric oxide content is very important for obtaining regional soil physical and chemical characteristics data, monitoring soil environment, and implementing regional precision agriculture. It took HJ-1A hyperspectral remote sensing image as the basis to establish an inversion model of free ferric oxide contents in surface soil of Nong'an county, which also had combined with test data obtained from soil sampling in the study area. The whole process was divided into three steps: first of all, soil samples were taken in the research area from April 29th to April 30th, 2017, and 82 soil samples were collected. At the same time, the longitude and latitude of the sampling points were measured by hand-held GPS receivers to determine the position of the sampling points in remote sensing images, and the free ferric oxide content of the soil samples was tested in the laboratory. Moreover, the HJ-1A hyperspectral remote sensing image covering the study area, which was synchronized with the soil sampling time, was selected, and the FLAASH atmospheric correction model was used to carry out an atmospheric correction on the remote sensing image to eliminate the influence of water vapor, ozone, dust and the like on the imaging process in the imaging process of the sensor. Erdas software was used to carry out a geometric precise correction on the remote sensing image, and the corrected error was less than 1 pixel. Erdas software was used to draw the boundary of the research area, and the remote sensing image was cut and spliced; Finally, the reflectivity was subjected to the mathematical transformation such as reciprocal, logarithmic, exponential, power function, reciprocal of the logarithm, first-order differential of reflectivity, first-order differential of reciprocal, first-order differential of logarithmic, first-order differential of exponential, first-order differential of logarithmic reciprocal, etc. The correlation analysis between the reflectance and its mathematical transformation data and the content of soil free ferric oxide was carried out to obtain the sensitive band of ferric oxide. The inversion model of soil free ferric oxide content in the study area was established by using the method of multiple linear regression analysis, which was used to invert the content of soil free ferric oxide in the study area and draw the spatial distribution map. The results showed that the reflectivity of HJ-1A hyperspectral remote sensing image was negatively correlated with the content of free ferric oxide in the soil in the study area, and reached the peak value at the 104th band, and the correlation coefficient was -0.455. The correlation between the reflectivity and the content of soil free ferric oxide could be significantly improved by first-order differential, reciprocal first-order differential, logarithmic first-order differential, exponential first-order differential, and other mathematical transformation. Among them, the reflectivity index of first-order differential transform and free soil ferric oxide content in the 79th band correlation was the best-reaching and the correlation coefficient was 0.908. A multivariate regression model was established by using reflectance index transformation. The determination coefficient of the model was 0.837 and the root mean square error was 1.59 g/kg. According to the testing results of the influence of organic matter in soil on the inversion accuracy of free ferric oxide contents, it was obvious that the organic matter in the soil of the study area could produce minimal impacts on the inversion accuracy of free ferric oxide contents. This experiment could use for reference for the spectral analysis of soil free ferric oxide and provide data support for the monitoring of the regional soil ecological environment.