Kinect-based pallet dynamic identification and location method and system, and medium

Abstract

The invention provides a Kinect-based pallet dynamic identification and location method and system, and a medium. The method comprises the steps of first, calculating extrinsic camera parameters; second, acquiring color and depth images, aligning the color and depth images, screening out effective points according to gray-scale and depth data and converting the effective points into a point cloud;third, pre-processing the point cloud obtained in the second step; fourth, performing Euclidean clustering on the point cloud processed in the third step; fifth, processing a cluster obtained in thefourth step, and extracting a characteristic plane; sixth, matching a characteristic plane point cloud with a front end surface of a pallet; and seventh, performing geometry calculation, and determining a position and a deflection angle of the pallet relative to a world coordinate system origin. The method and system have the beneficial effects that the pallet is identified mainly based on the characteristic of the front end surface of the pallet as well as the three-dimensional size of the pallet, the robustness is high, a set target point in the pallet can be identified accurately and a geometric center of the pallet can be calculated when only a half end surface is seen, the dynamic performance is high, and accurate location is realized.

Description

technical field [0001] The invention relates to the technical field of computer vision recognition, in particular to a Kinect-based pallet dynamic recognition and positioning method, system and medium. Background technique [0002] In recent years, with the rapid development of the logistics industry, automated logistics and warehousing systems have gained more and more attention. Among them, intelligent forklifts that can independently complete cargo handling and placement occupy an important position, and forklifts are the key to completing automatic handling tasks. It is whether the pallet can be accurately identified and positioned. [0003] In terms of algorithms, the current pallet recognition algorithm can be divided into two types: adding manual marks to pallets and not adding them. The method of adding manual marks refers to pasting two-dimensional codes, reflective stickers and other features that are easy to identify and locate on the end of the pallet. Marking, ...

AI Technical Summary

Problems solved by technology

[0003] In terms of algorithms, the current pallet recognition algorithm can be divided into two types: adding manual marks to pallets and not adding them. The method of adding manual marks refers to pasting two-dimensional codes, reflective stickers and other features that are easy to identify and locate on the end of the pallet. Marking, through the camera installed on the forklift to identify the mark and then locate the pallet, but this method needs to be modified for each pallet, the cost of labor and time is high, and the mark is easily damaged during use, resulting in The recognition is unstable; the method of not adding marks usually uses a horizontally installed lidar to detect the end face of the pallet, but from an economic point of view, the price of the lidar is high, which is not conducive to reducing the cost of industrial products

[0004] In terms of sensors, the CCD / CMOS cameras that are often used for object recognition have high requirements on ambient lighting conditions, and it is difficult to maintain the accuracy of recognition in environments with uneven or frequently changing lighting. Moreover, it is difficult for monocular cameras to directly obtain depth information. The camera needs to consume a lot of computing resources

Images