Detection of Miscellaneous Bicycle Lane Based on K-Means Algorithm

Abstract

The precise detection of bicycle lanes is quite supportive to cyclist tracking in either Advanced Driver Assistance System (ADAS) or intelligent connected vehicles. This paper proposes the bicycle lane detection approach based on the k-means algorithm. Firstly, the pre-processing steps are performed to obtain the promising region of interest, and then the k-means algorithm is used to classify the data. Afterwards, with the help of the classification index of each data point, the image is re-filled with color to segment the lane lines. The implementation of the algorithm employs Intel OpenCV library. Finally, the nonlocal means denoising is applied to remove the noise and obtain the desired lane lines.