No abstract

Based on lane-marking tracking with fuzzy adjustable vanishing point mechanism, this paper presents robustness forward vehicle detection system. Compared to most of the detection systems with a large curvature of road trend, which are not effective for the routes to detection and marking. Therefore, follow the current image frame, the proposed system calculate the error between lane detection point and regressed lane-marking lines, and then fuzzy inference system derive and update the proper vanishing point in next image frame. The algorithm proposed in this paper can be carried out for different road situations and adaptive tuning of the car camera has been successfully tested and proven for the highway to the robustness of the system.

Based on the lane marking tracking, this study presents a robust tracking approach for vision-based forward vehicle detection. Following vanishing point of lane which is according to lane detection point and linear regression analysis, five scan lines created as region of interest (ROI) for tracking is adopted to reduce the computational cost in detection process. The proposed algorithm implemented on TI DM648 DSP platform, and get very well results for various road environments.

An image based distance self-calibrating method for automotive electronic system is proposed to ease the difficulty in setup procedure and offset the possible changes caused by moving the camera that used to capture image for distance measurement. By process the information from vertical edge of dashed road line, the lookup table is established to translate the pixel distance to real distance. The only parameter needs to be provided by user is the practical distance that bottom row of captured image stand for. In addition, a noise-insensitive filter is designed to find the edge of dashed road line. As a result, the error is within 1 meter inside the range of 30 meter.