Integration of Hough Transform and Inter-Frame Clustering for Road Lane Detection and Tracking

This paper proposes a lane detection and tracking algorithm using vision-based measurement techniques. A framework consisting of color space conversion, region of interest (ROI), edge detection, Hough transform (HT), slope filtering and inter-frame clustering are used in the lane detection algorithm. An efficient ROI selection is obtained through vanishing point determination. Any conventional edge detection algorithm can be used in this algorithm with no trade-off in functional performance. HT detects all lines in the road images, but the potential lane candidate selection is achieved with slope filtering and inter-frame clustering. Further increasing the robustness of the lane detection algorithm, Kalman filter-based estimation technique is employed for lane tracking. The proposed algorithm is tested on the Caltech dataset, and its performance parameters are evaluated in terms of correct rate and false rate. It has shown a promising results in scenario of shadow, illumination variation, multiple lane markings, markings on the road, no lane markings, with road curb, and multiple lanes. The proposed algorithm is implemented on National Instruments compact Reconfigurable input-output (NI cRIO) real-time processor.