A New Methodology of Soil Salinization Degree Classification by Probability Neural Network Model Based on Centroid of Fractional Lorenz Chaos Self-Synchronization Error Dynamics

Abstract

In this article, a new methodology for the centroid variation of chaos self-synchronization error dynamics was used to determine soil salinization degree based on probability neural network (PNN). This was done to overcome the difficulties involved in the handling of a large amount of spectroscopy data, as well as many spectral bands and a low determination rate caused by bands redundancy. The spectral reflectance of saline soils in Xinjiang Uygur Autonomous Region was used as data sources. The results showed that salinization in the area was severe. The proportion of saline, moderately saline, and severely saline soil accounted for 67.3% of the total sample. Fractional-order master/slave chaotic analysis was carried out on the characteristics of soil spectroscopy data with different salinization degrees. The differences between integer order and fractional order of chaotic dynamic error were compared. Simulation results showed that changes in the 0.6 order chaos dynamic error were the most significant and so these were used as PNN input vectors. The PNN model was used to identify the nonlinear hyperspectral signal of soil salinization degrees after chaotic system conversion. The input vector was normalized after insertion into the PNN model input layer and was added to the hidden layer for Gaussian operations. Finally, the hidden layer results were used in the summation layer to calculate the correlation. The verification set classification result was 93.5%. The studies showed that the method proposed in this article could serve as a new way for classifying soil salinization, which has a classification accuracy of 93.5%, and the soil salinization degree can be rapidly determined.