Algorithm and Experiment for Vision-Based Recognition of Road Surface Conditions Using Polarization and Wavelet Transform

Road surface conditions vary with snow, ice, wet and dry during the four seasons in Korea. The change of the road surface conditions can have significant influence on the safety of a running vehicle on the road. Nowadays, there are a lot of CCTVs that are installed to observe the road surface with the unaided eye, and it is common for an administrator to determine the road surface conditions after confirming visually. However, there is a limitation to quickly determine the road surface conditions manually. Thus there is a need for developing a technology that can automatically determine the road surface based on images. In this paper, a recognition algorithm was proposed to extract the state of road surface conditions based on images obtained by a vision camera, and empirical experiment was conducted. The road surface conditions were classified using histogram and wavelet filter based on the proposed algorithm. In order to verify the recognition algorithm, the asphalt road was recorded and a variety of road surface conditions such as dry, snow, ice and wet was utilized as the input data. First, the road surface condition was distinguished into dry and wet using polarization coefficient, and then the dry condition was distinguished into dry and ice using wavelet and frequency characteristic. As a result of experiment, the possibility of vision-based recognition algorithm of the road surface conditions was demonstrated. In the future, it is expected that a way of road surface control through CCTVs would change from manual to automati