Real time lane detection on embedded platform for advanced driver assistance system /
Vision-based Lane Departure Warning (LDW) system is an important module in Advanced Driver Assistance System (ADAS) to promote safe driving. In practice, it is exceptionally hard to accurately and efficiently detect lanes due to a variety of complex noise such as environmental variability. However,...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Subjects: | |
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/5171 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Vision-based Lane Departure Warning (LDW) system is an important module in Advanced Driver Assistance System (ADAS) to promote safe driving. In practice, it is exceptionally hard to accurately and efficiently detect lanes due to a variety of complex noise such as environmental variability. However, image processing techniques have shown promising and reliable outcome in detecting lanes during non-ideal conditions such as faded lane marker. Although there are various researches on LDW system, they either suffer from computational complexity or not robust in degraded conditions. In this study, research and problem related to lane detection algorithm are highlighted and implementation of algorithm on embedded platform for real-time application is proposed. General process of lane detection can be divided into the following stages; pre-processing stage, colour conversion, selection of a region of interest and feature extraction. In pre-processing phase of algorithm, the input image was resized, filtered by the bilateral filter, and region of interest was selected to obtain the bird's eye view image. After that the top down view image is converted to Hue-Saturation-Value (HSV) colour space for the process of computing the binary image. As for lane feature extraction, from the binary image obtained, sliding window method was used and the lane is fitted by using 2nd degree of polynomial equation. In regard to the lane departure, the decision is made from lane measurement and observe the host vehicle is near to lane boundary or not. Then, a warning is sent to the driver visually. After the algorithm development, the algorithm was deployed to NXP development board with 64-bit Quad ARM Cortex-A53 @ 1000 MHz and 32-bit ARM Cortex-M4 @ 133 MHz and Renesas R-Car H2 with Quad-core ARM Cortex-A15 up to 1.4GHz for real-time application. The proposed algorithm is programmed in C++ language, which is compatible across multiple platforms. Optimization performance was performed to increase the frame rate of the algorithm while maintaining the percentage of detection accuracy. The result shows accuracy of detection is 90% when tested with Caltech dataset and the frame rate when running on embedded platform is 16 frame per second for NXP development board and 14 frame per second for Renesas R-Car H2 on images with the resolution of 256 by 144. In conclusion, the objectives of the research have been met. The algorithm for the LDW system has been developed and be ported to the embedded platform for real time implementation. |
|---|---|
| Item Description: | Abstracts in English and Arabic. "A thesis submitted in fulfilment of the requirement for the degree of Master of Science in (Mechatronics Engineering)." --On title page. |
| Physical Description: | xiii, 77 leaves : colour illustrations ; 30cm. |
| Bibliography: | Includes bibliographical references (leaves 72-76). |