High-resolution Remote Sensing Images Road Extraction Study

Owing to successful launch of major satellites, the epoch of remote sensing is coming quietly. (?) can get lots of ground information from high resolution remote sensing images through remote (?)sing technology. It is a great practical work that extracting the object from high resolution remote (?)sing image by using image processing technology, at the same time, because of the characteristic mplexity presenting in high resolution remote sensing image, this work is full of challenging. This (?)ject becomes an exploring direction for multitudinous researchers.The road is very important in our daily life and the military. We can't ignore the effects of road. (?)ad extraction form high resolution remote sensing image has the vital significance. It can provide (?)erence for city planning and update the data of electronic map. It also provides the basis for the (?)d resources management etc. Aiming at high spatial resolution remote sensing image this paper (?)sents a new road extraction algorithm to complete road edge and centerline extraction based on (?)thematical morphological method and Snakes model.(1) According to the initial position sensitivity of active contour model, this paper is studied in (?)d initial contour setting method:first, extracting the roughly road area from the high resolution note sensing image based on mathematical morphology method directly. According to the gray (?)tribution and morphological characteristics of object we can get roughly area of road by retaining mary shape of the road and delete other non-road shapes by using proper structural elements and (?)thematical morphological operation. Then use the Contour_following operator to get contour line labeling the road edge. Experimental results show that this method is more efficient than ditional regional growth for initial contour setting method. This method can extraction road (?)tures more quickly and more accurate. This method is not only effective on single road but also (?)ective on road with intersections and multiple parallel roads. It is adaptive and robust, but the ti-interference ability is not strong. The shade of cars, trees etc causes deviation. The edge of road ists jag.(2) This paper presents a road centerline extraction algorithm. Based on the initial contour tracted in method (1), use skeleton method to extract the road centerline, and do some st-processing to make the centerline more accurate. Experiments showed that this method is faster d more accurate than the method using skeleton method from original road image. The result of nterline algorithm this paper presents is close to the practical road centerline.(3) Road extraction from high resolution remote sensing image based on Mathematic Morphology and Snakes model. Namely:the first step consists in extracting the road area from the high resolution remote sensing image based on morphology. In a second step, get the contour line of the area as the initial contour of Snake to track the road. Experiments turn out that the algorithm is extremely fast in image division and effective in extracting roads in high resolution remote sensing images after few iterative. The edge of road extracted is continuously and smooth.