Vision based tunnel detection for driver assitance system

Knowing the environmental changes are essential in vision based driving assistance system. Detection these changes prior to algorithm selection will help to handle different situation separately and in turn enhance the performance of ADAS (advance driver assistance system). In particular, this paper focuses on detecting tunnel using images captured from monocular camera mounted on the ego vehicle. In this paper, we propose detection of tunnel using the pattern and shape of lights available inside the tunnel. Tunnel detection using entry and exit information fails for different shaped tunnel and night condition. Also the process is computationally expensive. The proposed method is simple to implement using basic image processing tools and not computationally intensive. As per observation, tunnel light follow a straight line pattern. Line fitting is used to analyze the pattern and the magnitude of error in line fitting decides whether the image has a tunnel or not. Region of interest and brightness parameter is calculated to avoid unnecessary processing of normal day image. The proposed algorithm has been tested on both day and night images captured from grayscale monocular camera mounted on the ego vehicle.