Research on River Segmentation Method of Remote Sensing Image

Traditional change detection methods based on spectral information are difficult to achieve ideal results. Accurate extraction of river water shapes is a key issue for remote sensing image interpretation and recognition. High-resolution remote sensing images contain a large amount of surface information, and it is difficult for a single method to effectively segment the target area. When segmenting the edge of the remote sensing image water area, there will be problems such as various types of surface water area and complex and changeable water area texture structure. Therefore, there are higher requirements for the accuracy of segmentation in the simulation process. Aiming at the problems of river flow in remote sensing images that easily cause image noise and the existence of a large amount of regional information similar to the river gray value in remote sensing images or the complex environment where waters are blocked by a small area, this paper proposes a multi-scale and multi-structural element Morphological reconstruction edge detection combined with improved region growing algorithm to achieve remote sensing image water segmentation, both accuracy and processing time are improved.