A rotatable battery recognition method based on improved YOLOv5 Online publication date: Mon, 03-Jun-2024
by Wenming Chen; Dongtai Liang; Wenhui Ding; Meng Wang; Zizhen Chen
International Journal of Sensor Networks (IJSNET), Vol. 45, No. 2, 2024
Abstract: To realise end-to-end visual identification, positioning, and angle detection of cylindrical batteries, a rotated object recognition method based on YOLOv5 is proposed. Firstly, aiming at the problems of battery appearance scale variation and surface reflection, a recursive gated convolution and feature fusion module was added to the neck network to enhance the multi-scale feature extraction. Secondly, considering the boundary problem of angle range, a circular smooth label was introduced after the prediction network, and the logistic regression cross-entropy was used to realise rotation angle classification. Finally, a SIoU intersection ratio model was used to introduce an angle vector penalty index. The experimental results show that the parameters of rotated object detection model are reasonably optimised on the cylindrical battery dataset. The model accuracy reaches 98.6%, the recall rate reaches 97.2%, and the inference speed of single frame image reaches 10.5 ms, which meets the performance requirements of practical applications.
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Sensor Networks (IJSNET):
Login with your Inderscience username and password:
Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.
If you still need assistance, please email subs@inderscience.com
Our Blog 
Follow us on Twitter 
Visit us on Facebook