Pose estimation method based on combination of linear features and point cloud features

Abstract

The invention discloses a pose estimation method based on combination of linear features and point cloud features. The pose estimation method comprises the steps: (1) linear feature extraction fused with priori knowledge; (2) straight line matching in the binocular image; (3) three-dimensional reconstruction of linear features; and (4) pose calculation. The point cloud of the invention comes froman edge feature and has good anti-interference capability and accurate positioning capability, and the robustness advantage of point cloud matching can be brought into full play by using point cloud matching to replace line segment matching, and even if the line segment has the phenomena of length change or fracture, effective matching can still be realized after the line segment is changed into the point cloud; the number of the point clouds is limited, and the coverage area is a line segment set with limited space, so that the number of the point clouds is greatly reduced, and the matching speed is improved, and the point clouds come from the edge of an object with the maximum contribution to positioning, and the positioning precision is not obviously reduced; and in the linear feature extraction and matching process, compact depth field information is not needed, and the precision can be ensured for objects with complex textures and simple textures.

Description

technical field [0001] The invention relates to a pose estimation method, in particular to a pose estimation method based on the combination of straight line features and point cloud features, belonging to the technical field of image processing. Background technique [0002] The problem of pose estimation is an important problem in photogrammetry, computer vision, computer graphics and robotics. Calibration and many other core issues to be solved in engineering practice and theoretical research. [0003] A simple method is to directly extract some corner features of the object, and calculate the pose of the object relative to the visual system according to the position of the feature point set in the visual coordinate system and the position of the object itself, but the number of corner features of the object is limited, which affects The solution accuracy may even be unsolvable due to insufficient number of features. In recent years, the pose estimation method based on ...

AI Technical Summary

Problems solved by technology

[0003] A simple method is to directly extract some corner features of the object, and calculate the pose of the object relative to the visual system according to the position of the feature point set in the visual coordinate system and the position of the object itself, but the number of corner features of the object is limited, which affects solution accuracy, even the solution cannot be solved due to insufficient number of features

In recent years, the pose estimation method based on point cloud features has been popular. This method can locate objects with complex shapes and has good robustness. However, the method requires image acquisition equipment that can obtain high-precision and dense 3D point cloud information, and requires Realize complex operations such as point cloud segmentation and stitching between the measured object and the environment, and due to the large number of point clouds, the algorithm calculation process is time-consuming

[0004] Some methods consider positioning based on the straight line features of the object. The straight line features are more obvious than the corner features and have stronger anti-interference ability. The method is simple, the calculation amount is small, and there is no need to measure the image depth. However, the straight line features cannot be compared with the point features in the matching process. The same accurate positioning (especially in stereo vision), when the extracted straight line length changes or breaks due to occlusion or illumination, it is difficult to guarantee the accurate restoration of the three-dimensional pose of the object. Most of the existing methods focus on In the extraction and matching of straight line features, there is no in-depth discussion on the final positioning accuracy and reliability

Images