Ayeyarwady River Regions Detection and Extraction System from Google Earth Imagery

Nowadays environmental changes have frequently occurred all over the world. Among these changes, river changes are also important to be monitored. The accurate river region detection is the primary important step in river change detection. There is a lot of analysis in river regions detection and extraction based on Landsat and Synthetic Aperture Radar (SAR) images, but there are a few approaches for Google Earth RGB images which do not contain multispectral bands. Thus, this paper is proposed to detect and extract the river regions including both river and sandbank using Google earth RGB images. In this paper, the river regions and sandbank regions are firstly segmented using heuristic rules based on Sobel and thresholding methods. And then, HSV histogram features, color autocorrelogram features, color moments features, Gabor wavelet features and wavelet transform features are extracted from the segmented regions. The multi-class Support Vector Machine (SVM) is applied to categorize the river regions, sandbank regions, and others based on the derived features. According to the testing result, the proposed system obtained 94% overall accuracy for river regions detection.