A supervised segmentation algorithm for crop classification based on histograms using satellite images

Abstract

Recognizing different types of crops trough satellite imagery is an important application of Digital Image Processing in Agriculture. A supervised algorithm for identifying different types of crops is proposed. In the training stage, the studied images are preprocessed using a bilateral filter, and then the histogram of intensity levels is constructed for every crop class. The segmentation stage begins with the assignment of the likelihood of each pixel to belong to each class, which is based on the histogram information. Finally the segmentation is obtained using Gauss-Markov Measure Field. For this research Landsat-5 TM satellite images are used. The experimental work included synthetic and real images. In the case of the real image, the ground truth image was given by an expert. The results of the proposed algorithm were compared with other methods such as Maximum likelihood, Fisher linear likelihood, and Minimum Euclidean distance, among others.