Vehicle Identification Research In Complex Scenes

The vehicle automatic detection and recognition plays an important role in intelligenttraffic system and parking automatic management system and other systems, themanagement of car wash shop is gradually becoming automatic and intelligent. The carwashing service charges of car wash shop is mainly according to the shape of vehicle andthe owner of the vehicle is the member of shop or not. So, researching the vehicleidentification in car wash shop has important practical significance. The scene of car washshop is different from outdoor road and community and other places, car wash shops arevaried, the scene of car wash shop is unusually complex, these problems bring greatchallenges to the research work. The complex scenes of car wash shop mainly include:narrow shop space, internal of shop are dark, but the doorway has intense light, in winter itis easy to produce thick fog, and so on, all these makes the vehicle detection and recognitionbecomes difficult.For the complex scenes of car wash shop, this paper makes depth study at imagepreprocessing. At first, we use image enhancement algorithms to improve image contrast,after a lot of experiments, Retinex enhance algorithm has performance prominent, and thenthis paper introduces wipe the image fog methods, in many image defogging methods, weachieve the purpose of defogging using based the shadow channel priori image defoggingalgorithm by adjusting the degree of demisters. The task of vehicle detection is detectionand recognition vehicles, eliminate the target which not belong to vehicle at the same time.This dissertation puts forward a kind of vehicle detection algorithm which based on imagehistogram matching, the algorithm first set a ROI(Region of Interest) in car wash shopground, then through matching histograms which come from ROI image at two differenttime to determine whether there is a car coming。For the vehicle identification，due to thelimit of the car wash shop space, this paper studies vehicle recognition in monocular vision.At first, we extract the vehicle body by means of digital image processing technology, thenby calculating the length and width of the vehicle body to identify and classify the vehicle.The experimental results show that, the vehicle automatic detection algorithm which based on image histogram matching has much advantage, such as simple operation, smallamount of calculation and high real-time characteristics, can accurately detection andrecognition the vehicle targets and eliminate targets which not belong to vehicle at the sametime, almost meet the actual demand. For the algorithm of car shape identify, white, silver,etc of light color vehicle identification effect is good, for deep color of vehicle recognitioneffect is not satisfactory, also needs to be improved.