Detection method for vehicles in tunnels based on surveillance images
Lili Dong, Li Qin, Wenhai Xu, Haitao Jiang · 2017
A “vehicle in, light brightens; vehicle out, light darkens” energy-saving control system is proposed to address electrical energy waste in tunnel lighting. In this system, timely and accurate detection of vehicles in tunnels is guaranteed through the use of a tunnel surveillance application. Under different illumination conditions, with no vehicles on the road, we obtain images of each tunnel segment with each surveillance camera and use them as background images before detection begins. Regions of interest (ROIs) are then extracted to reduce the subsequent processing area. For the detection, we use the background subtraction method, edge detection method, and block division processing method. Then, the nonzero point (NZP) number in specific blocks is calculated to determine if vehicles are in the tunnel based on the NZP threshold. This method takes into account practical constraints including real-time operation and poor image conditions. Our main contribution is a simple and practical method to avoid the omission ratio of vehicle detection under the special tunnel environment, which is essential for tunnel security. Experimental results show the proposed approach achieves a 0% emission ratio (existence is detected as nonexistence) and a false detection ratio (nonexistence is detected as existence) of less than 5%, which ensures the safe operation of the system in the tunnel.