The Technology Of Curvilinear Structure Extraction And Its Application

Geometric characteristics are important attributes of object recognition in computer vision systems. When the object’s position, direction or scale changes, the system can still describe and distinguish the object relying on geometric characteristics. Linear feature is a very important middle-level descriptor, which can reflect a lot of high-level information of images. This paper studies one kind of the linear feature, called curvilinear structure, which s line or curve with width. Curvilinear structure contains more information and can better express the object. Its application contains many aspects, so it is always the hot research issues.We focus on the two application of curvilinear structure detection in this paper:the lane marking detection in intelligent transportation system and the road extraction in aerial images. Both the lane marking and road is elongated and bright structure in images. So we solve the related problem based on the research of curvilinear structure detection.Roads location and recognition can be used to acquiring and updating the GIS data, identifying the moving target and planning the transportation. This paper analyzes the necessity and feasibility of the multiscale analysis in road extraction. We successfully detect road feature based on automatic scale selection, and we use the roads’attributes in the method to reduce the computation complexity with scale space iteration. This method can detect roads with arbitrary width.As a part of intelligent transportation system, the lane marking detection can warn the drivers when the car is off track, then it can reduce the traffic accidents. In this paper, we design a brightness difference feature based on that the lane marking is brighter than its surroundings. Besides, we study the high-dimension RANSAC curve fitting, and use this technology to describe the lane markings’geometric information. The experiments show that this method can satisfy the real time requirements. At the same time it is robust to illumination changes and can well eliminate interferences.