Research Of Vehicle Monocular Measurement System Based On Computer Vision

With the rapid development of the modern social economy， the standard ofpeople’s material requirement is improving and the number of vehicles increasesgradually. This will bring about a consequence that traffic accidents occur frequently.Under the circumstances， Intelligent Transportation Systems and Intelligent Vehiclesare one of the major means to solve the traffic problem. They not only draw thecomprehensive attention of the society and the government， but also become asignificant research topic of public concern. As an important component of IntelligentTransportation Systems and Intelligent Vehicle， both vehicle detection and distancemeasurement technique have a broad application area.In this paper， related issues about vehicle detection and monocularmeasurement system are studied. Improved vehicle detection method and rangingmodel are given based on the summary and analysis of the existing methods andtechnologies.The main contents and results of this paper are as follows:(1) Vehicle detection based on computer vision. To reduce the system workload，region of interest in the image must be set at first. The preliminary location of theforward car can be achieved after finding the shadow portion as a possible vehicleregion in the ROI， and then the location of the license plate is detected. Thus therear of front car can be ascertained， and takes the middle of the rear of front car andits4-neighborhood points as reference points.(2) Construction of monocular measurement system. Pinhole imaging principle isthe basis of the theoretical ranging model. It needs to calibrate the camera accordingto the model in order to make sure the parameters of the camera. The actual model isdeduced by geometric method and it can build a conversion relationship betweenpixel coordinates and real world coordinates. So the distance between two cars canbe calculated by this relationship.(3) Design and implementation of the software system. Application of vehicledetection and ranging system is developed using open source computer vision libraryOpenCV in Visual C++6.0environment. It carries on error analysis for the staticexperiment data， and also corrects the data with the least square method. Experiments show that the method of this paper has certain effectiveness andfeasibility， and has higher efficiency and accuracy.