Geometry target positioning method based on multi-layer depth and color visual information

Abstract

The invention discloses a geometry target positioning method based on multi-layer depth and color visual information. The method comprises the following steps: correcting internal parameters and distortion coefficients of a color camera and a depth camera, and determining a conversion matrix relationship between the images of the two cameras of the depth camera and the color camera; acquiring color and depth image information of the target object; preprocessing the color image in an HSV color space according to the intuitive image characteristics of the color, and segmenting a binary image ofthe target object; eliminating the background of the depth image through background subtraction, carrying out the multi-level threshold binarization operation, and segmenting an object through the multi-level depth image); carrying out edge contour detection based on the multi-level binary image, and identifying a target geometry through parameter screening; and calculating a central coordinate and a rotation angle of the target geometry through the Hu invariant moment and converting the central coordinate and the rotation angle into a world coordinate system of the visual system. According tothe invention, the identification effect on the texture-free geometry is good, and the anti-interference capability for different light environments is strong; the principle of object recognition issimple, the calculated amount is small, and the requirement for real-time recognition can be met.

Description

technical field [0001] The invention relates to a geometric object positioning method based on multi-layer depth and color vision information, which belongs to the technical field of computer vision. Background technique [0002] Since the birth of the first industrial robot 150 years ago, people have been committed to using robots to replace human heavy work. Among them, autonomous grasping ability has always been the key to detecting the level of intelligence of a robot, while robot vision has become the main way for robots to obtain external information with low cost and high information content. [0003] Traditional monocular vision or binocular vision positioning is seriously affected by the current light, and the amount of calculation is relatively large. However, monocular vision captures limited object information in a static state, which is not enough to provide sufficient three-dimensional coordinate information of the target object. Therefore, using an infrared ...

AI Technical Summary

Problems solved by technology

[0005] The present invention proposes a geometric object positioning method, the purpose of which is to overcome the problem that existing matching algorithms cannot effectively identify regular geometric objects without texture features

Images