AI Depaletizing

Masterless, various judgments compatible with transformation, order control

Solution Overview

This system uses deep learning to recognize workpieces from 2D data acquired by a 3D vision sensor, and then determines the workpiece’s position and orientation from the 3D data of the recognition unit. It is effective for recognizing bags and boxes of arbitrary sizes.

Target work

Bags
箱
Aluminum ingot

The technologies used

Arm mounting
Fixed vision
off-line
Automatic Calibration
3D point cloud generation
Deep Learning
Point cloud processing technology
FA specifications
Implementation support
SIer collaboration
After-sales service

Features

Learning Model

By using a pre-trained model specifically designed for bags, you can immediately recognize bags. The recognition rate can be improved by adding photos of actual workpieces (approximately 30-50 photos) using the following procedure.

AI bag recognition software

Recognizing bags with uneven filling or overlapping bags, which was previously difficult, is now possible. By combining 3D robot vision and deep learning, it is now possible to reliably recognize irregularly shaped bags that are stacked on a pallet and overlapping. Furthermore, instead of relying solely on deep learning recognition, by using 3D and 2D data in combination, it outputs bag recognition and bag position/orientation recognition results that can be used in the field. ■ Prevention of misrecognition through dimension checks (size, aspect ratio, area ratio) ■ Detection of only the top bag to prevent picking products from the next lower layer first ■ Ability to set conditions such as the picking order of multiple products on the same layer

specification

System configuration and operation flow