Multi road marking detection system for autonomous car using hybrid- based method
For at least two decades, the development of autonomous systems has led to the development of embedded applications allowing to improve the driving comfort and safety. One of the embedded systems that received great attention is road detection system, that operates using road markings detection algo...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/76074/1/FK%202018%20154%20-%20IR.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | For at least two decades, the development of autonomous systems has led to the development of embedded applications allowing to improve the driving comfort and safety. One of the embedded systems that received great attention is road detection system, that operates using road markings detection algorithm. To date, the issue on detecting road markings under various imaging conditions has not been tackled yet. Generally, the road markings detection is performed on road images extracted from videos that were recorded using a camera, which was placed inside a vehicle at a fixed position. In this thesis, a road markings detection system that tackle the problems of detecting road markings under various weather and illumination conditions is proposed. The proposed system consists of a combination of Inverse Perspective Transform method, an image enhancement method and edge detection method. The Inverse Perspective Transform method was used to convert images, which were extracted from the recorded videos to bird’s-eye view images, while an image enhancement method, namely Contrast Limited Adaptive Histogram Equalization (CLAHE) was used to tackle various illumination conditions and Sobel edge detection method for detecting the road markings. Experimented on Large Variability Road Images database (LVRI) that consists of 22,500 road images, which were extracted from videos recorded around Selangor and Kuala Lumpur and T. Wu dataset that consist of 1208 road images, which were extracted from videos recorded around California, the proposed algorithm performed satisfactorily. With an accuracy of 96.53% using LVRI and 99.33% using the T. Wu datasets, the proposed algorithm able to detect almost all types of road markings. The types of road markings available in the LVRI and T. Wu datasets are forward arrow, left-side arrow, right-side arrow, lanes and signs printed on the road that are under various imaging conditions, including complex background and occlusion. In addition, the proposed algorithm outperformed the algorithm introduced by T. Wu. However, the algorithm has difficulty in detecting road markings painted in soft yellow color. Hence, in future, the algorithm will be improved by incorporating HSI color analysis with the aim of tackling the problem of detecting road markings that are painted in soft yellow color. |
|---|