Analysis And Optimization Of Road Extraction Algorithm For High Resolution Remote Sensing Images

The road is one of the important data information of remote sensing images,image processing has been a hot and difficult research field coincidence,real-time updating of city traffic management,city planning,automatic vehicle navigation,geographic information system database updates are dependent on the road of the high precision automatic identification and extraction of road network and information.But because of the current computer,target recognition,artificial intelligence and other technical limitations,road extraction has not a perfect system can realize the intelligent and automatic extraction of road information,so the research of high resolution remote sensing image road extraction algorithm has important significance.The road can be divided into single road and road network extraction algorithm can be divided into supervised and unsupervised classification,based on the theory and method of system analysis of the current status of remote sensing image segmentation techniques,this paper introduces three basic algorithms for extracting road information from remote sensing images.(a)this paper studies the watershed algorithm,aiming at the shortcomings of traditional watershed algorithm,proposed a watershed algorithm based on threshold marker,experiments show that this algorithm can effectively overcome the influence of over segmentation,extraction accuracy of single road information can reach 78%.(b)the watershed algorithm is non supervised category,in this paper the support vector machine supervised algorithm,study the basic principle of support vector machine classification algorithm,this paper selected the image features,the image is divided into land,grassland,residential area,playground,lawn,road six,selects six samples for training the experimental results,the image is divided into blue,green,white,black,purple,yellow,six different colors,including yellow road information,using pixel features can be well extracted road information,the algorithm extracts the correctness of single path information reached 95%.(c)the extraction of the two algorithms in remote sensing images is one way of information,real life road information from remote sensing images more performance for the road network,fuzzy C means(FCM)clustering algorithm can extract the information of road network,but the extraction accuracy is only 91%,road and other objects together to separation,therefore studies one kind of algorithm is very important,which can effectively extract the remote sensing image of the road network,is proposed in this paper by support vector machine and fuzzy C means(FCM)clustering algorithm to extract road network information,the correctness of the algorithm for extracting road network information is 94%.In summary,road extraction from remote sensing images of the classical algorithm and the improved algorithm,the extraction effect is different;the different attributes of the image to be extracted to obtain better extraction effect,the application and improvement of factors,the need to consider the properties of image and target,difficulty and complexity of the algorithm and the application background and demand etc..