In this paper, a motion-based vehicle detection is proposed to detect the vehicles in Hsuehshan Tunnel. Camera vision based vehicle detection in the long tunnel is a challenging problem. The video quality emerging from camera is affected by dynamic illumination environment, time variant, camera resolution, camera aging, camera position, camera view angle, heterogeneous camera, and vehicle speed. Besides, the color, shape, size, and appearance of vehicles are very variable. The proposed method provides illumination compensation, moving objects detection, and vehicles identification. Experimental results show that the proposed method can be effective to detect vehicles in Hsuehshan Tunnel with illumination compensation.