Contourlet Transform-based Road Detection Methods With Remote Sensing Images Of The Active Contour Model

With the rapid development of remote sensing technology, we can obtain a large number of remote sensing images which play an important role in our practical applications. The spatial resolution of remote sensing images has been improved since the launch of high-resolution satellites such as IKONOS, Quick Bird and Spot-5. It becomes possible to detect targets from high-resolution remote sensing images, and thus provides a novel research route for target detection. At present, the application study of high-resolution remote sensing images is becoming the hot research focus in domestic and overseas remote sensing, GIS and military field.As the most basic and common man-made objects, road has the significant geographical and geometric feature. Thus combing road detection with the latest mathematical theories and image processing technology from high-resolution remote sensing images is not only of great theoretical significance, but also has the application value in the national economic construction. The road detection from remote sensing images can provide timely, accurate and effective data for setting up or updating the traffic geographic information database, and increase the spatial analysis abilities of GIS-T. It also greatly improves the timeliness of electronic map and sharply decreases its cost, and provides guidance on road traffic administration and planning which has important significance for the decision and planning of urban construction.During the past dozens of years, a number of road detection algorithms in remote sensing images have been developed, and certain progress was made under specific circumstances. The researching methods can be divided into automatic method and semiautomatic one according to its automation. On account of the complexity and variety of the remote sensing images, these methods possess respective merits and drawbacks, without universal adaptability. There aren't consummate methods so far.In this paper, we focus on the semiautomatic road detection from the high-resolution remote sensing images. Through consulting a large amount of relevant documents, analyzing and summarizing forerunners' working experience, to aim at the characteristic of the road in the high-resolution remote sensing image, the paper has put forward a kind of methods in order to achieve the goal of detecting road on the remote sensing image. The experimental results show the effectiveness of our proposed approaches.The main works are summarized as follows:(1) Introducing the background of research and the development trend of road detection, several main road detection algorithms are listed, and compared.(2) Introducing some traditional edge detection methods, and comparing their advantages and disadvantages.(3) The theory and implementation of Contourlet transform are studied in detail, and we point some limitations of Contourlet transform caused by the subsampling, such as the properties of shift variance and frequency scrambling. We give a brief introduction of Nonsubsampled Contourlet transform.(4) Discussing the traditional active contour models, as well as its improving ones, including the gradient vector flow active contour models and Ribbon snake models. We bring forward a new road detection method based on the Ribbon snakes, and extract the center line of road.(5) According to the T. Katherine's road detection method based on edge recognition, we divide the road detection process into four steps: edge detection, edge thinning, edge locating and edge tracking. The main processes are summarized as follows: First, the Nonsubsampled Contourlet transform is applied to the image edge detection. Then non-maxima suppression and mathematical morphology are used to thin the edge image. Finally, the gradient flow vector snake model is applied to track and connect the edges of road.