| In recent years,the ownership of automobile has gradually increased with the improvement of the people's material standard of living.As the people's main travel tool,automobile has changed people's living conditions and lifestyles,provided great convenience to people,but also brought a greater threat to people's lives and property safety.Therefore,it is particularly important to regularly detect and maintain the technical condition of automobile by means of the technology.Machine vision technology is gradually developing as an important mean to improve the accuracy and efficiency of automobile inspection.Therefore,the vehicle profile is an important factor that affects the safety and passability of the vehicles.The vehicle profile inspection based on machine vision can automatically identify whether the vehicle is overloaded or not,and provide important basis for vehicle classification,vehicle size parameters,acquisition and reconstruction of the vehicle information.Therefore,the research on active vision inspection method of the vehicle profile based on the support platform without the motion constraint has a great significance to enrich the methods of vehicle performance inspection,ensure the safety of people's lives and property,and realize the automation inspection of the vehicle.An active-vision inspection method of the vehicle profile is studied based on the support platform without the motion constraint,according to the current research of the vehicle profile inspection.The closed-loop consistency model of an active-vision inspection method of the vehicle profile without the motion constraint is established to gain the solution of the transform relation between the internal camera and the support platform.The closed-loop consistency model consists of the 3D target,the support platform and the external camera solving loop,3D target,the support platform and the internal camera solving loop.According to the solving model,the coordinates extracted from the image are taken as initial values.The optimization function aiming at minimizing the re-projection error is established by combining the re-projection coordinates of the feature points of the coordinate system of the support platform in the image.The transform relation between the support platform and the external camera is optimized by using the bundle adjustment algorithm.In addition,the calibration model of the laser plane of an inspection method is established based on the 3D target.Moreover,the calibration of laser plane is realized by adopting the SVD decomposition.In order to reconstruct the 3D feature points of the vehicle body,the LED feature points are firstly tracked and identified by the combination of the frame difference method and Canny edge detection algorithm to improve the extraction efficiency of feature points of the support platform with different poses in the visual field of the external camera.Then the centroid position of LED can be extracted by calculating the contour moment.Secondly,the reconstruction model of the active vision inspection system without motion constraint is established according to the inherent pose parameters of the detection system obtained by the closed-loop system and the projection relationship the intersection point of the laser projective points and vehicle body.The effectiveness of the system is evaluated by the standard cylinder.Then the 3D reconstruction of feature points of the vehicle body is realized under the external camera by using the transform relation of the laser projection points of the vehicle body in the internal camera,combined with the closed-loop consistency model.Finally,the analysis model of reconstruction error of the detection system is established in order to study the influence of various influencing factors on system reconstruction.The influence of camera internal parameters,external parameters and image coordinates are analyzed theoretically on the reconstruction error of the detection system. |
PDF Full Text Request