Research On River Extraction Method From Remote Sensing Image

Rivers play an important role in the landform of China.The monitoring of river waters is conducive to the better utilization of water resources,the reduction of flood disasters and the regulation of water conservancy and hydropower.With the development of remote sensing technology,the surface information can be collected and analyzed.River conditions,areas and river channel changes can be monitored effectively by remote sensing image river extraction.Based on the river features in remote sensing images,this paper proposes a regional growth algorithm based on k-means clustering and applies color HSV image segmentation to remote sensing image river extraction by studying the application of regional growth algorithm and color HSV image segmentation.Aiming at the low efficiency and low accuracy caused by manual selection of initial seed points for regional growth,a regional growth algorithm based on k-means clustering was proposed.In this algorithm,the river boundary frame is obtained by edge segmentation of remote sensing image.Morphological treatment was used to fill the river frame,and k-means clustering method was used to obtain river particles.The particle is used as the starting point of the regional growth algorithm to obtain the river target.In order to solve the problem of information loss and difficulty in extracting disconnected waters caused by the need of grayscale conversion in remote sensing image processing,a method for river extraction of remote sensing image based on HSV color image was proposed.The algorithm converts color RGB image into color HSV image.The influence of H and S components on image segmentation was removed by dot product.By empirical method,V threshold value is set as 0.4,and the threshold image is constructed with H and S component dot product graph.The threshold image is divided by multiplying with color RGB image.Check by shape to get river water unless river water is obtained.This paper studies the river extraction of remote sensing image and demonstratesthe proposed algorithm.Experiments show that the improved regional growth algorithm can effectively improve the accuracy and efficiency of river extraction.HSV color image segmentation can effectively extract all waters in remote sensing images.