An inverse projective mapping-based approach for robust rail track extraction
| Autor: | Zhenhui Huang, Zhongli Wang, Guiguo Wang, Yan Yan, Baigen Cai, Chunxiao Jia, Xian Wu, Tianbai Zhang |
|---|---|
| Rok vydání: | 2015 |
| Předmět: |
Computer science
business.industry ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Image processing Image segmentation Image (mathematics) Hough transform law.invention Line segment Robustness (computer science) law Shadow Computer vision Artificial intelligence business Feature detection (computer vision) |
| Zdroj: | 2015 8th International Congress on Image and Signal Processing (CISP). |
| DOI: | 10.1109/cisp.2015.7408003 |
| Popis: | Rail track extraction from the image can be used to determine the train localization, or vision-based inspection of the rail, and some other driver supporting systems in railway. For the vision-based inspection, it is one of the most important tasks. But visual approach suffers from some common difficulties, the illumination change, shadow, weather effects, etc, which make the image processing for the successive procedures a very tough work, even in a sample application. In this paper, with a calibrated camera and based on inverse projective mapping (IPM), we can get a bird-view image or IPM image first. Then the IPM image is converted to an edge map with modified binary method. In order to extract the rail track, a Hough transformation is used detect the line segments in the edge map. Combined with some geometric constraints from the prior knowledge, the rail track can be robustly extracted. Some experiments with on-site image show the proposed method can works well. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|