A Calculation Method of 3D Position and Normal Vector of Planar Parts Based on Depth Map

Abstract

The invention discloses a method for calculating the three-dimensional position and normal vector of a plane part based on a depth map, which uses shape matching to identify the part, and then uses a binocular stereo vision system to obtain a dense depth map of the region of interest where the part is located; Non-uniform sampling and statistical analysis to remove outliers; then use a robust random sample consensus algorithm to fit the plane in the dense depth map; calculate the 3D position and normal vector of the part according to the plane equation, and give the camera The conversion method between the coordinate system and the world coordinate system. The invention has strong versatility, high positioning accuracy and good stability, and is suitable for planar parts.

Description

technical field [0001] The invention relates to the field of three-dimensional positioning of visual industrial robot parts, in particular to a method for calculating the three-dimensional position and normal vector of a plane part based on a depth map. Background technique [0002] Visual industrial robots are often used in fields such as welding, palletizing, assembly, processing, testing, logistics handling, and spraying. Parts need to be identified and three-dimensionally positioned during the application process. Three-dimensional positioning of parts is an important prerequisite for visual industrial robots to accurately perform operations such as grasping and assembly. The accuracy of three-dimensional positioning directly affects the quality of manufacturing. In the industrial unstructured environment, the poses of the parts are different, and the parts are often disturbed by factors such as nonlinear lighting changes, messy backgrounds and occlusions, so that the tr...

AI Technical Summary

Problems solved by technology

In the industrial unstructured environment, the poses of the parts are different, and the parts are often disturbed by factors such as nonlinear lighting changes, messy backgrounds and occlusions, so that the traditional 3D positioning method based on sparse feature points cannot obtain stable and accurate positioning. result

In addition, a lot of research has been done in related fields at home and abroad. The existing visual three-dimensional positioning system is mainly developed according to the shape and size of specific parts, and its versatility and robustness are not strong. The promotion of visual three-dimensional positioning software is limited limit

Images