Analysis Of Object Pose And Motion Trajectory In Image Sequences

Visual detection technology is an advanced measurement technology with broadapplications in recent years, it has the advantages of non-contact, high efficiency, moderateaccuracy. Object pose measurement and motion trajectory calculation based on vision is animportant part of visual detection, it can be used in aerospace and wheel alignment, etc.This paper comes from a car wheel alignment visual inspection system, the researchobject is planar target, it uses monocular vision technology to study the method of objectpose measurement and motion trajectory calculation in the sequence images, thus provideevidence for the follow wheel alignment parameters’calculation.First, the paper analyzed the measurement of objects space pose measurement, thebasic theory of motion estimation, the extraction technology of visual image feature andcorresponding fields at home and abroad. Describes the imaging model of the visualsystem and the traditional calculate method of spatial plane pose based on vector, and onthis basis, given the positioning method of plane pose based on characteristics of elliptic.The method make the problem of determine the posture of space plane translate to theproblem of determine the posture of ellipse by using circular perspective projectionfeatures. It determines the spatial plane pose by building relationships with spatial planepose and target image. Extract trajectory of target plane from sequences by using segmentfitting of least square method,the track is curve fitting in a plane.Secondly, this paper proposes a fast algorithm for extracting the ellipse characteristicparameters in sequence images. The method scans the target image and collects the edgepoints, then it uses the elliptical nature to get other edge points and calculate the desiredellipse. After the raster conversion process of the ellipse, we get the correct parameters. Ituses the corner points to create multiple coordinate systems, thus achieves rapid topologyposition labeling. On this basis, the target detection of sequence frames is based on theresults of the previous frame. It can improve the detection efficiency and avoid repeatedsegmentation, boundary scanning, and automatic labeling. Finally, the paper has made a verification analysis on the computing model andalgorithm through experiments. It shows that the computing model is correct and effective,the algorithm can extract characteristic parameters quickly and accurately.