Abstract:

A vision-based guardrail detection method was proposed and used on the cleaning vehicle deviation warning system to obtain the accurate distance information between the guardrail cleaning tanker and the guardrail, and to avoid the distracted driving and fatigue driving which was caused by the driver simultaneously
observing the position of the guardrail and operating the tanker. The binocular stereo vision was used to obtain the exact distance value of the guardrail. By combining the camera calibration and marking watershed algorithm to segment the area where the guardrail was located, and by using the Hough transform line detection algorithm, the line features was extracted in the guardrail segmentation area. The guardrail detection was realized by constraining the line detection parameters and the number of line detections. The position of the guardrail was tracked with the Kalman filter under the condition of constant speed model and the feature detection information and position tracking information were combined for decision making and warning. The results show that the average warning accuracy of the system is 89.40%, and this system has certain practical value in engineering sanitation vehicles.

Key words: active safety , guardrail cleaning vehicle , deviation warning system ,  stereo vision ,  watershed ,  Hough transform linear detection ,  Kalman filter