The Research And Application Of Road Detection Algorithm Based On Spot Image

With the rapid development of the space technology and sensor technology , remote sensing information has become one of the main sources for geographical data. But our knowledge and application about it are way behind with collecting of the information sources of the space and aerial system. Road extraction has been an important part of the extraction of remote sensing information and got wide attention, which has a matter of great significance in regional planning, traffic navigation,video matching and military spying, especially for the applications of remote sensing imformaton science and geographic information system(GIS), the remote sensing images can be extracted road information to update GIS data and provide favorable conditions for GIS application.According to the theory of Marr,aiming at the SPOT remote sensing images'characteristics, this paper proposes a method of detecting road in SPOT image. And this paper reaches the objective of acquiring road middle lines.And the work in this paper focuses on the following aspects:(1) the research of preprocessing method for smoothing road imagesAt the base of studying and analyzing the algorithm of anisotropic diffusion smoothing filter, this paper studys a method of self-adapting image smoothing,which improves on original algorithm,on aspects that diffusion coefficient is how to self-adaptly acquire and stopping time is how to decide. This paper's method extends nonlinear diffusion method to solve the paradox of keeping structure information and smoothing homogenous areas.Through contracting salt and pepper noise with edge,this paper improves standard median filter and proposes a selective filter method, which can remove salt and pepper noise in the same of keeping edge. Experiments also argue this papper's methods is effective.(2) the research of getting line-support regions through choosing seed in road regionAiming at the characters of road pixel and road direction, at the base of fuzzy connectedness, this paper proposes an algorithm of image segment based on direction restrict, which includes optimal path's calculation, directional condition's establishment and road seeds'choosion. The algorithm acquires the degree of pixels belonging to the seed,in way of choosing seeds and calculating optimal path between pixel and seed. Through contracting with original method, this paper's algorithm is more better in robustness and producing more less redundance.(3) the research of detecting road middle line in way of road's characterThis paper introduces line degree, improves the algorithm of thread tracking, proposes a new method of tracking based on line degree, which can choose the direction in the course of tracking according to road's linearity. This paper's method can avoid dividing road into some segments and appearing false connection. Some experiments improve the algorithm is effective in getting integrate road line.(4) the vectorization of road and the application in MapStarWith the purpose of producing digital line graphic(DLG), digital raster graphic(DRG) needs to be vectorized. Because of the disadvantage of traditional vectorizing methods, this paper improves variable-step-length vectorization method based on the idea of variable-step-length vectorization method. In the end, the result is showed in MapStar, which perfects the map collection model, forms a simple extracting frame and realizes the purpose of reducing manual work in production of MapStar.