A Study on Weed Discrimination through Wavelet Transform, Texture Feature Extraction and Classification

Texture based weed classification has played an important role in agricultural applications. In the recent years weed classification based on wavelet transform is an effective method. But the feature extraction is main issue for proper classification of weed species. In this paper, the issue of statistical and texture classification based on wavelet transform has been analysed. The efficient texture feature extraction methods are developed for weed discrimination. Three group feature vector can be constructed by the mean and standard deviation of the wavelet statistical features (WSF), Texture feature as Contrast, Cluster Shade, Cluster Prominence and Local Homogeneity (WCSPH) and Energy, Correlation, Cluster Shade, Cluster Prominence and Entropy features (WECSPE) which are derived from the sub-bands of the wavelet decomposition and are used for classification. Experimental results show that Rbio33 Wavelet with WECSPE texture feature obtaining high degree of success rate in classification.