The efficiency of manual wheel hub classification is low and the error rate is high. The machine vision scheme has the advantages of simple structure, high recognition rate, high precision and strong anti-interference ability, which can meet the requirements of automatic production.
Share
一. Project background:
Requirement description:
1. After wheel production and transportation, different kinds of wheels need to be classified for subsequent sorting and packaging.
2. manual detection method identification efficiency is low, high labor cost, low accuracy.
Technical requirements:
1. Recognition accuracy: >99.9%
2. wheel identification type: >200 kinds
二. Solution architecture:
The hub type detection system uses a 1.3 million high-resolution camera of Haitianxiang, combined with a large area of medium-hole surface light source, and uses deep learning classification tools to realize automatic identification and classification of a variety of wheels in the state of random mixed flow on the conveyor belt of the production line.
三. Program advantages:
1. accurate recognition: the initial model training of each type of samples collected 80-100, the comprehensive recognition rate can reach more than 99%, through the field iteration, the final recognition accuracy of more than 99.99%.
2.Large recognition capacity: Deep learning classification algorithms take advantage of convolutional neural networks and big data samples to improve the recognition capacity.
EN
