摘要：The quality and nutrient content of soil are important factors in determining the soil’s usability. Quantitative analysis is a highly uscftil tool for establishing the trends of these properties in different areas. A neural-network-based method is proposed herein. The soil nutrient indicator was taken as the input of the neural network and the soil nutrient grade as the output, and Matlab was utilized to establish the comprehensive assessment model for soil nutrients in a back-propagation (BP) neural network with eight hidden layer nodes and three layers of networks. Different grades of assessment standards for soil nutrient indicators were taken as training samples and inspection samples of the model to train and inspect the model. The results show that the BP neutral network is consistent in analog and expected outputs of the inspection samples. The correlation coefficient for the analysis of the soil was 0.999. The trained model was utilized to perform a comprehensive appraisal of soil nutrients in the Heilongjiang Land Reclamation Jiansanjiang Administration Bureau. The results show (hat all the Jiansanjiang farms have a soil nutrient grade of 2, except for Erdaohe Farm and Chuangyc Farm, which aie grade 1.
