Remote Sensing Image Road Network Separation Method based on Edge Feature Selection

摘要

Aiming at the feature selection issues in road network recognition process of remote sensing image, a road network selection method of remote sending image based on the ant colony algorithm is proposed. The results of simulation comparison experiment indicate that the ant colony algorithm can select the feature subset with high relational degree to the road network recognition results of remote sensing mage, which can achieve dimensionality reduction on the feature effectively, improve the global convergence of the algorithm and road-network recognition rate of the remotes-sensing image, and be the reference on how to give consideration to accuracy rate and instantaneity at the same time while recognizing the remote sensing image network separation method. The feature of this method shall be: extract key points of multidate remote sensing image road network as the initial seed point and feature points for image registration; it can improve the efficiency of change detection by combing the image registration and change detection. A strategy of energy minimization shall be applied in this method to extracting the opacity of pixel, which can detect changes in extensiveness, including some imperceptible changes. The simulation indicates the effectiveness of this method.