3D mosaic method in monocular vision measurement system for large-scale equipment

Abstract

In order to enlarge the measurement range of the monocular vision measurement system, enhance its measurement precision in depth direction, and realize the accurate measurement of large-scale equipement, a 3D mosaic method based on assistant target is proposed in this paper. By using the image of an assistant target in two adjacent measuring positions, on the basis of matching the images of feature points and epipolar constraint, it quickly computes the orientation and translation transform between the two adjacent measuring positions through linear algorithm and LM iteration method, and the scale factor of the translation is calculated from the distances between the feature points on the assistant target. At last, the measurement data in the local coordinate systems of different measuring positions can be transformed to the global measurement coordinate system with the transformation between the two adjacent measuring positions, and the three-dimensional mosaic of the measurement data is realized. This method overcomes the problem that the precision of the feature point three-dimensional coordinates in different coordinate systems seriously affects the mosaic accuracy. The experiment result proves that the proposed method is flexible and valid, and not only enlarges the measurement scope, but also raises the measurement precision for large-scale equipement.