Title: Application of deepLabV3+ network model in garbage detection and classification

Authors: Lixiang Shi; Guofang Liu

Addresses: School of Computer Engineering, Chengdu Technological University, Chengdu, Sichuan Province, China ' School of Computer Science, Sichuan University Jinjiang College, Meishan, Sichuan Province, China

Abstract: In order to optimise the cleaning performance of the intelligent sweeping robot, the problem of garbage area segmentation and garbage detection and classification of the sweeping robot is studied and proposed by using the deep LabV3+ network and DBN model. The results show that the semantic segmentation accuracy of the deepLabV3+ model is 91.7%, and the semantic segmentation effect of the model image is good. The average detection and recognition accuracy of DBN deep LabV3+ model for various types of indoor garbage is 92.48%. The detection and recognition effect of garbage with regular shape is better. The detection and recognition rates of plastic bottles and battery garbage are 96.21% and 94.79% respectively. The proposal of the deepLabV3+ model provides a new research idea for the improvement of the intelligent level of the sweeping robot, and has certain reference value for the construction of the garbage detection, identification and classification system.

Keywords: computer vision; deepLabV3+; refuse classification; sweeping robot; area detection.

DOI: 10.1504/IJWMC.2023.131328

International Journal of Wireless and Mobile Computing, 2023 Vol.24 No.3/4, pp.380 - 389

Received: 29 Apr 2022
Received in revised form: 07 Dec 2022
Accepted: 11 Dec 2022
Published online: 06 Jun 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article