Aiming at the technical problems of axle-type recognition under the condition of non-uniform vehicle speed,such as ultra-close distance,large area,and low distortion imaging,a method of ultra-close imaging correction and mosaic for vehicle chassis with speed-adaptive ability is proposed.The ultra-wide-angle image sequences of a moving vehicle chassis are collected by a fish-eye camera,and a method of circle fitting by multi-sample points is proposed to extract the region of interest(ROI).The method for image distortion correction is then performed using an isometric projection model based on the 3D imaging geometry model.Aiming at the problem of vehicle's non-uniform speed,which leads to large range of the overlap area between video's adjacent frames,a multi-step layered key frame extraction is carried out with image content overlap analysis technology,avoiding the influence of uneven image distribution on mosaic results.The experimental results show that the proposed method can obtain clear and complete vehicle chassis'images with lower distortion at the distance of 20~50 cm under the condition of non-uniform vehicle speed,with a 5%average measuring error of the vehicle axles distance.