Road Segmentation From QuickBird Imagery Based On Mathematical Morphology And Texture Characteristic

With the development of remote sensing technology in recent years, remote sensing resolution improved a lot. At present, the highest spatial resolution taken by commercial satellite is 0.61m, achieved by QuickBird panchromatic band. High-resolution remote sensing provided massive data and richer information. It brought a new challenge to identify and extract information fleetly and automatically. Image segmentation was the important step from image process to analysis. As the premise of imagery information extraction, the study on imagery division technology attracted increasing attentions, which has great significance to improve the application of high-resolution remote sensing.The research aim of this paper was to enhance division precision of road information on QuickBird imagery. Base on the summarization of methodology and theory, the imagery segmentation technology was analyzed, combining the improved mathematics morphology algorithm and texture characteristic of QuickBird image and the road information is extracted automatically.After the division experiment to the typical experimental region and precision estimate, it was tested that the average precision of road and other linear goal extracted by mathematics morphology algorithm was 89.06, while the average precision by image texture characteristic was 89.30. Both methods get similar high precision and utilizing the image texture characteristic in road information extraction from QuickBird image achieved higher precision.The paper was divided into five parts. In the first chapter, the background of this research was expatiated. The study status and development tendency of road feature extraction from high-resolution remote sensing imagery both in domestic and oversea was analyzed, furthermore, the imagery segmentation methods was evaluated. The structure of the paper was introduced in first chapter as well. The Second chapter mainly introduced mathematics morphology based imagery segmentation method for QuickBird imagery. After the depiction of basic theory and method of mathematics morphology, the controlled mathematics morphology was utilized for segmentation experiment for road extraction and the result was evaluated. In the third chapter, the road segmentation technology and Markov random field model based on the texture characteristic of QuickBird imagery was illuminate and employed for the road segmentation, and the result was evaluated subsequently. The fifth chapter was the conclusion and expectation.