Research On Vehicle Shape Measurement Technology Based On Image Processing

The automatic vehicle type recognition technology is one of the key technologies of ITS system,and the 3D information of the vehicle shape can provide a reliable and effective data base for vehicle type recognition.The traditional method of vehicle shape measurement mainly depends on the laser radar,which has high measurement accuracy,but is susceptible to weather and atmosphere,and the cost is high,which is not conducive to large-scale practical application.Therefore,this paper studies a vehicle shape measurement method based on image processing,through the image processing technology,to achieve the vehicle shape length,width and height of the measurement.In this paper,two kinds of realization schemes based on planar image and depthbased image are proposed for different application scenarios.These two schemes use the camera close installation method to obtain a clear vehicle image,and then through some prior knowledge and image processing methods to obtain 3D information on the outer surface of the vehicle.And then according to the sequence image in the process of vehicle movement,the complete vehicle image is concatenated by image registration method,and finally to achieve the vehicle shape of the 3D measurement.The main contents of this paper are as follows:1.Based on the traffic application scenario,this paper studies the calibration algorithm based on the vanishing point which can effectively use the field information,realizes the off-site calibration of the camera,and verifies the accuracy of the method by experiment.2.This paper studies the method of obtaining 3D coordinates of vehicle based on image.In the scheme based on the plane image,the 3D coordinates of the vehicle side are obtained by the reverse projection method by using the horizontal distance of the vehicle to the camera and the characteristics of the large and medium-sized vehicles whose sides are perpendicular to the ground.In the depth image-based scheme,the 3D point cloud of the outer surface of the vehicle is obtained through the depth information and the geometric relationship of camera imaging,and the error and accuracy of the obtained 3D coordinate are analyzed experimentally.3.For the problem that the camera can not get a complete image through a frame image for a few meters,or even ten meters of the vehicle,according to the sequence image of the vehicle during the process of motion,this paper studies the registration method of the vehicle image and proposes a constraint method based on the consistency of the rigid body motion,which effectively eliminates the wrong match points,to improve the result of the vehicle image registration,and achieve a complete stitching of the vehicle image.Finally,for the complete vehicle image with 3D coordinates,this paper introduces the concrete method of vehicle shape measurement,and tests the measurement result of the method and analyzes the main factors of measurement error.