Measurement Of Tunnel Luminance And Vehicle Detection Based On Video Stream And Its Application

With the rapid development of highway construction in China,the number of tunnels and the total mileage are increasing rapidly,and new requirements are put forward for highway tunnel lighting,it considers not only the safety and comfort of the driving,but also the energy saving to reduce the cost.The energy-efficient tunnel lighting is a combination of the brightness outside the tunnel,traffic flow and the lighting intensity in the inside the tunnel.The dissertation detects the brightness and vehicle of outside the tunnel,and realizes the goal of "lighting the car come" and "lighting on demand" on the basis of video streaming.The brightness measurement method based on video stream has the advantages of wide measurement area and high accuracy,we use the OV5640 camera that calibrated with an imaging brightness meter to measure the brightness outside the tunnel.The vehicle detection method based on video stream has outstanding the advantages of large area of detection,easy expansion of function and convenience of remote monitoring and so on,we use DS-2ZMN2006 network camera,and combine Codebook and Perceptual Hash algorithm to complete detect vehicle detection.The main works of this dissertation are as follows:1.First,in order to simplify and visualize the operation steps of calibrate camera,we design a IFIX based on MFC as a framework,combined with OpenCV,libvlc library and socket.Then we specifically calibrate the OV5640 camera based on the camera's optical principle and digital image processing technology.Finally,we propose the method of segmented measurement to improve the accuracy of camera measurement luminance.2.Using the calibrated parameters and camera image brightness calculation methods,we develop a linux-based brightness measurement system,it is responsible for image acquisition and sending,grayscale information extraction,target area brightness calculation and transmission.Finally,we test the system in the indoor and outdoor environment.The results show that the measurement error of the system is less than 3%,which can meet the requirement of detection accuracy of tunnel.3.In order to solve the problem that the camera has a slight displacement caused by the natural environment in the measurement of the external brightness of the tunnel,which causes the deviation of the target measurement area(the tunnel hole in circular region center),we propose the brightness measurement system use the image processing technology such as Sobel edge detection,non-maximum value suppression and Hough gradient algorithm to complete the target measurement area selection automatically.4.We propose a foreground detection algorithm based on Codebook and combined it with Perceptual hash algorithm to realize vehicle detection and traffic statistics.And write a IFIX to test the accuracy of this algorithm for vehicle detection and traffic statistics under typical weather conditions at night and during the day.By testing the sample video,the results show that the detection accuracy rate of the algorithm is 100% in the presence or absence of the vehicle passing detection area in the daytime,and the detection accuracy rate of the traffic flow is more than 97%.If the camera is properly installed,it will eliminate the multiple detection caused by the lights in the evening,and it can also achieve the accuracy of the day.Finally,this algorithm is transplanted to the Linux-based video vehicle detection system.