Volume  Number

previous article

next article

Hyperspectral remote sensing estimation of soil nutrients in the black soil region based on computer vision model

Li Ma^a, Anqi Li^a, Helong Yu^a,b, Guifen Chen^c,*

 
ABSTRACT:     : With the continuous development of modern agriculture, precision fertilization is based on the profit and loss of soil moisture and nutrients, scientific irrigation and formula fertilization, to achieve full resource efficiency. Hyperspectral remote sensing and computer vision technology are very important tools in precision fertilization. The purpose of this study was to estimate soil nutrients in the black soil region using computer vision technology combined with hyperspectral remote sensing. Soil samples (n=163) were collected from northwestern China. The content of soil organic matter (SOM), total nitrogen (N), phosphorus (P), potassium (K), and pH were measured. In the study, measured nutrient component spectrum conversion data was selected to compare global and local spatial autocorrelation of the soil nutrient elements. Multi-step regression analysis was used to estimate soil nutrients. Finally, the estimated value was compared with the actual value, and the percentage error was calculated to evaluate accuracy. The results showed that the prediction model of soil total nitrogen content was the most correct, with a prediction accuracy of 78.16% and a relative error of 21.84%, It is concluded that computer vision hyperspectral remote sensing has high estimation accuracy and effective reflection. This study provides a basis for the determination of soil nutrients estimated by remote sensing and can supplement the implementation of precise fertilization.

Download PDF

239 Downloads 1341 Views

* Corresponding author, E-mail: chenguifen@jlau.edu.cn

Received 26 Feb 2021, Accepted 23 Nov 2021