Automatic Road Network Extraction Based On Color And Texture

Automatic road extraction from panchromatic or multi-spectral aerial images is one of the most active research fields, which is significant for city planning, traffic control, monitoring hazards and etc.The early aerial data can but be gray image for technical limitation, and so most of the semi-automatic and automatic road extraction methods hadn't include color and texture into their solutions. And for the nature of semi-automatic of most road extraction solutions, the computer technology didn't play a leading actor in cartography practice. So, unsupervised road extraction is also the key point of research.After studying many literatures in this field, an automatic method of road network extraction based on color and texture is brought forward. The object of this method only is a single true color image, and the content of it consists of four modules:1. Many initial road points are picked up automatically, which resolve this key point of automatic system of road extraction, but also can avoid some road area fail to be detected.2. Studying TPA in detailed. After modification, it is applied to determine single pixel, which resolve the key point of road research.3. Gradual road overspreading is brought forward, which can avoid the defect of traditional methods had to deal with all the pixels in the aerial images, and also decrease the fake extracted result.4. Finally, a prototype system is developed for validate the correction of the solution mentioned above.