Research On Image Correction And Stitching Technology For Vehicle Axle-type Recognition

Currently,overloading of trucks is a prominent problem in the field of highway transportation.Truck axle-type recognition is of great significance for accurately determining the weight limit of trucks and implementing refined management.However,the axle-type recognition instrument currently under test is still difficult to meet the practical engineering needs in terms of technical maturity and recognition stability.It is difficult to accurately judge the axle-type of trucks directly from the side of the vehicles.In addition,from the perspective of law enforcement,clear and complete chassis image data is an indispensable resource for industry management.In response to the above-mentioned engineering requirements,this paper conducts technical research from the perspective of improving the imaging quality and integrity of the truck chassis under complex application conditions,and uses the generated truck chassis image to quickly distinguish the axle-type information of the truck.The specific research work is as follows:In order to solve the problem of imaging difficulty caused by the large size of the truck chassis and the short imaging distance,a technical idea of using a fish-eye camera to build a vehicle chassis imaging system is proposed.Based on the prior knowledge of truck chassis height,width and speed,etc.,the frame rate of video acquisition is calculated,and the video data acquisition scheme is designed.Aiming at the shortcomings of the traditional fish-eye image acquisition algorithm,a method of using multi-point fitting circle to obtain the effective area of the fish-eye image of the vehicle chassis is proposed.This method uses multiple point coordinates on the boundary of the effective area of the fish-eye image to complete the calculation of the fitting circle,as to obtain the radius and center of the effective area.The experiment proves that the method has better adaptability.The equidistant projection model of the fish-eye lens is used and the reverse mapping method is used to complete the correction of the fish-eye image of the vehicle chassis.Compared with the cylindrical surface projection method and the latitude and longitude mapping method,the calibration results are verified by qualitative and quantitative experimental analysis methods respectively.In the quantitative analysis,three parameters,SSE,RMSE and R-square,are used to assess the corrected image results.The experimental results verify the proposed algorithm.To address the influence of uncontrollable truck speed on the result of sequence image stitching,a multi-step hierarchical filtering method of image key frame extraction is proposed.This method uses interval sampling to quickly extract images,and then uses the overlap relationship between images to fine-screen images.In the process of image registration,the combination of ORB registration algorithm and template matching algorithm is used to eliminate the mismatched points effectively,so that the registration accuracy is improved.The simulation experiment is carried out by using a car with varying speed,and the experimental results show that this method can effectively filter images and eliminate the influence of vehicle speed changes on the final result.The actual truck chassis data is used for experimental verification,and the results show that the proposed method can output clear and complete truck chassis images.