Research on Visual Inspection for Defects in Lithium Battery Packaging Based on Faster R-CNN
- Resource Type
- Conference
- Authors
- Li, Lina; Sun, Hongchang; Msalika, Othman P.
- Source
- 2023 4th International Symposium on Computer Engineering and Intelligent Communications (ISCEIC) Computer Engineering and Intelligent Communications (ISCEIC), 2023 4th International Symposium on. :202-206 Aug, 2023
- Subject
- Communication, Networking and Broadcast Technologies
Computing and Processing
Signal Processing and Analysis
Training
Visualization
Computational modeling
Machine vision
Fitting
Manuals
Packaging
Packaging defects
Faster RCNN
Classification
Deep learning
loss-function
LabView
TensorFlow APIs
- Language
In response to the low efficiency of manual detection methods and techniques for defects in lithium battery packaging, an automated method based on faster region convolutional neural network (RCNN) machine vision algorithm was studied. This article proposes a feature extraction and loss-function based on the Faster RCNN structure, which reduces the amount of data loss through extensive training to improve detection accuracy. This model is deployed using LabView and TensorFlow APIs to get better performance. The actual results indicate that the application of this faster RCNN structure achieves with accuracy of over 95% for all types of packaging defects used in training.