The inspection of vehicle profile is one of the important researches for the test ofvehicle passability and driving safety. The inspection of vehicle profile detects the vehicleprofile informations in use in time. It also automatically identifies whether a vehicle isoverloaded or illegally modified. The inspection of vehicle profile provides a significantbasis for vehicle dimensions detection, intelligent recognition and classification of vehicle,deformation detection of vehicle in use. Therefore, the inspection of vehicle profile basedon active vision has important significance to reduce the incidence of vehicle accidents,protect people’s lives and property safety, and improve and perfect vehicle performance testtechnology.A reconstruction method for vehicle profile information is proposed based on3Dvision measurement of line structured light. The method is aimed at the research status ofexisting vehicle profile inspection. In order to realize the3D reconstruction of the vehicleprofile, we design a laser scaning system of vehicle profile inspection drived by a steppingmotor. The electronic control system and the software control system have been researched.It is realized the control of the laser scaning plane. In the3D active vision measurementsystem of structured light, the key point of detection accuracy is the precise2D coordinatesof the laser stripe center in the image. Hence, an extraction method of laser stripe center hasbeen conducted a thorough research. First, the laser stripes are separated from the complexbackground using difference image method in the images. The images contain theintersections of the autobody and the laser stripes. Second, the distribution characteristics ofthe grayscale of the feature points to be elected in the laser stripe center are describedaccording to the distribution characteristics of the grayscale of the laser light center in theimage. The eigenvalues and eigenvectors of the Hessian matrix are adopted in thisdescription. Third, Sine-Gaussian function normalized formula and Sigmoid-Gaussianfunction normalized formula are established according to the characteristic of the Hessianmatrix eigenvalues of the pixels in laser stripe center. An active vision method for extractinglaser line centers based on Sine-Gaussian function standardized identification model and anactive vision method for extracting laser line centers based on Sigmoid-Gaussian functionstandardized identification model are proposed. Finally, according to the character that the the sub-pixel light stripe center, the position of the extremum points are obtained bycalculating the Taylor expansion formula along the direction of eigenvector. The sub-pixelextraction of laser stripe center of vehicle profile inspection system is achieved. In order toimprove the running speed of vehicle profile inspection system, we use the Kalman filteringmethod to track the laser stripe and predict the area of laser stripe in the video. The laserstripe tracking model based on Kalman filtering is constructed accoding to the motion lawof the laser stripe. The area of laser stripe is predicted by the laser stripe tracking modelbased on Kalman filtering. The laser stripe centers are extracted in the local area thatcontains the laser stripe. It is realized the sub-pixel extraction and tracking of theintersections of the autobody and the laser stripes. |