Stacked object recognition method and device and machine sorting system

Abstract

The invention provides a stacked object recognition method and device and a machine sorting system. The method comprises the steps: obtaining a depth map of stacked objects; determining K segmentation positions of a top material layer of the stacked object according to the depth map, and extracting K top material layer depth maps from the depth map according to the K segmentation positions, wherein K is an integer which is greater than or equal to 2; carrying out the object segmentation of K top material layer depth maps, so as to segment candidate objects from the top material layer; carrying out the fusion of the candidate objects in K object segmentation results, so as to obtain a candidate object after fusion; and recognizing the object in the top material layer from the candidate objects after fusion according to the pre-determined size and / or shape of the stacked objects. The method, device and system can achieve the effective segmentation of objects which are pasted together in an irregular manner, accurately recognize a plurality of objects located at the top layer of the stacked objects, and greatly improve the recognition and grabbing efficiency of objects from the disordered stacked objects.

Description

technical field [0001] The present disclosure generally relates to the field of image and video processing, and specifically relates to a recognition method, device and machine sorting system for stacked objects. Background technique [0002] With the development of vision-based automation technology, a large number of automation systems and equipment have begun to be used in many fields, such as fruit quality sorting, tablet detection and identification, cell analysis, etc. In particular, the technology of robot grasping has gradually stepped out of the laboratory and been applied to industrial product lines, freeing laborers from complex and heavy work. Items that can usually be grasped by the robot's arm include electromechanical components, food and agricultural products, etc. [0003] In order to enable the robot to recognize each individual object from the messy pile of objects, the conventional methods are template matching method, shape detection method, etc., but t...

AI Technical Summary

Problems solved by technology

However, in the existing layering techniques, a single global threshold is used to obtain the topmost image, and this single threshold layering method is often ineffective in dealing with the adhesion problem between objects, especially when the upper layer It is difficult to accurately segment the top layer of the stacked object when the object and the lower object are closely connected

This is because the objects in the stacked objects are stacked freely instead of being placed one by one, layer by layer, and the size and orientation of each individual object are also different, so it is determined by a single threshold Layering Difficult to correctly segment top layer objects and identify individual objects within

Images