| As the core part of the intelligent traffic management, Auto Vehicle License PlateRecognition System(LPRS) is an important research topic, based on computer vision,integrating multi-technology of different fields, such as digit image processing, patternrecognition, artificial intelligent and so on. Meanwhile, LPRS itself is an intelligentsystem as a whole. Generally, the vehicle images are captured from actual scenes byhigh-definition cameras and all the recognition work is done automatically, withoutaffecting the traffic order. It is this character that makes LPRS has a huge applicationmarket in the field of intelligent traffic management. At present, part of LPRS has beensuccessfully applied to the actual scenes both at home and abroad, such as highwaycharge management, bridge charge management, vehicle access control of some specialsectors, intelligent parking management and the detection of drive against trafficregulations.LPRS mainly consists of five functional modules: image acquisition, license platelocation, character segmentation and character recognition. The research focus of thispaper is license plate location technology, which has a direct impact on the quality of allthe following parts, thus affects the performance of the entire system. At present, mostof the license plate location technologies just do well for good lighting conditions,ignoring the fact that vehicle images are captured from the outdoor actual scenes.However, in a natural scene, both the weather change and the light intensity changefrom morning to night bring a lot of trouble to license plate location. To solve theproblem, this paper has systematic research and discussion on license plate locationtechnology in complex weather conditions.Firstly, I have a research on single image defogging technology in this paper. Infoggy weather condition, due to the atmospheric scattering, the images captured bycameras are with low contrast and low sharpness, which has a huge negative impact onlicense plate location. To solve this problem, this paper presents a simple method basedon dark image to determine whether the image is interfered by the fog. At same time,based on the development of He’s single image defogging algorithm base on darkchannel prior, a more effective defogging method is proposed in this paper, which notonly maintains the good defogging effect, but also reduce the time consumption to onethirtieth of the original method. In addition, a new license plate location algorithm based on wavelet decompositionis proposed in this paper, which takes advantage of the texture features, color andgeometric features of Chinese license plate. The new algorithm is divided into two steps.Firstly, the candidate plates are initially detected from the original images using waveletdecomposition base on texture features and geometric features. In this step, both thelow-frequency sub-band and the vertical high-frequency sub-band are used to detectcandidate plates, in order to overcome the bad influence of the complex weatherconditions. Meanwhile, according to the different characteristics of the two sub-bands,two different binarization methods are applied to the two sub-bands respectively. Then,with the help of the candidate plates’ position information provided by both sub-bands,the geometry characteristics are used to realize candidate plate location. However, alittle amount of non-plate regions are inevitably included in these candidate plates,because of the similar texture features of license plates and car lights. Therefore, themain work of the second step of the proposed method is to exclude the interference ofthe car lights and other non-plate regions taking advantage of color characteristics.Experiments show that the proposed license plate location algorithm in this paper issimple, effective, with good location results for different weather conditions. Meanwhile,it also meets the real-time, accuracy and versatility requirements of actual LPRS, whichhas a certain value in practice. |
PDF Full Text Request