Two-stage hough transform algorithm for lane detection system based on TMS320DM6437

摘要

This paper presents a novel two-stage Hough transform (HT) algorithm and how to use it on a lane detection system for vehicles and industrial robot based on TMS320DM6437. As is known to all, white and yellow are the basic color of lane mark, which can be easily extracted through YCbCr color space. In order to improve detection performance, the two-stage HT is proposed to act on a binary image which is produced by setting threshold for `Y' component and `Cb' component. As lane detection algorithm usually does, HT is used on a edge image first. Then the two-stage Hough transform use the same accumulator with the first time. Finally, an Enhanced Hough Transform Graph (EHTG) is generated. The EHTG contains the information both edge and color. Thus, we can utilize it to visualize lane and transplant the algorithm on DSP. After a series of optimization in programming, the result shows that the proposed algorithm give an outstanding performance on DM6437 platform despite of lane mark embedded in severe noise environment.