Soil texture classification using wavelet transform and maximum likelihood approach

In this paper, a wavelet-based soil texture classification system is proposed for identifying soil with different textures. The wavelet transform is used for feature extraction. The wavelet is a systematic and powerful tool for signal and image analysis due to its multiresolution characteristic. The maximum likelihood (ML) classifier is designed using a set of training samples. The ML parameter estimation method has been shown to give out optimal results. During the process of training and classification, the Fisher´s Linear Discrimination Analysis (FLDA) is incorporated for feature vector dimension reduction and optimization. Three different soil texture images, i.e., sand, silt, and clay are used for training and classification. Experimental results and discussion are presented.