Arrow Marking Detection And Recognition Algorithms

In the field of intelligent transportation, arrow marking detection and recognition is very important. For the purpose of fast and accurate recognizing arrow markings drawn on the surface of road, many researchers introduced many technologies of Computer Vision and obtained good effect. In order to improve the speed and effectiveness of recognition, a thorough research on the existing algorithms related to arrow marking recognition and image matching is carried out in this paper, and some algorithms are improved to implement more efficient detection and recognition. In this paper, main work is as follows:(1) An arrow marking detection and recognition algorithm based on the improved Chamfer matching is proposed in this paper. Firstly, an edge image is represented with a collection of line segments. These line segments are quantified and grouped by directions. When distance transformation is adopted on each group of edge lines, the direction differences between the matching points is introduced to the computation of matching cost, which is different from the classical Chamfer matching algorithm. So the algorithm can improve the accuracy of matching. In order to improve the computing speed of the algorithm, the integral image is used to calculate the cost between the template image and the testing image. The experimental results show the proposed algorithm has a better recognition performance.(2) An arrow marking detection and recognition algorithm based on point pairs matching and geometric structure matching. This algorithm firstly extracts local multi-scale HOG feature from the neighborhood of edge points in the test image and the template image. And then, the point pairs matching performs between the test image and the template image by the feature. At last, a geometric structure matching algorithm is proposed for analyzing and comparing the geometric distribution of the selected edge points so that the category of the test image can be determined. A large number of experiments show that the proposed algorithm can overcome the influence of occlusion, scaling and rotation, and recognize the arrow markings accurately.