重庆科技大学学报（自然科学版）2025，Vol.27Issue(2)：70-79,10.DOI:10.19406/j.issn.2097-4531.2025.02.009
基于改进YOLOv8的化工泄漏检测方法
A Chemical Leakage Detection Method Based on the Improved YOLOv8
摘要
Abstract
It is crucial to strengthen the chemical leakage detection to ensure production safety,reduce economic losses and protect the environment,traditional sensor-based leak detection methods have limitations such as limited monitoring range,low real-time performance,and susceptibility to environmental factors.To address these issues,a chemical leakage detection method based on the improved YOLOV8 model is proposed,and three optimization strategies are employed to upgrade the basic YOLOv8 model.The Efficient Vision Transformer(EfficientViT)is utilized to update the backbone network to enhance the model′s capability of capturing the global information.A su-per lightweight dynamic upsampling module is introduced to heighten the model′s ability to restore texture details and edge information of chemical leak features.C2f module is replaced by C2f_DCNv2 module,enhancing the model′s detection capability for large-scale chemical leaks.The experiment result shows as follows,Compared to YOLOv5 and YOLOv8 models,the improved model′s accuracy is increased by 5.4 percentage points and 3.5 per-centage points respectively,and its mean accuracy is respectively increased by 10.3 percentage points and 3.5 per-centage points.关键词
化工泄漏检测/YOLOv8 模型/EfficientViT模块/DySample模块/C2f_DCNv2 模块Key words
chemical leak detection/YOLOv8 model/EfficientViT module/Dysample module/C2f_DCNv2 module分类
计算机与自动化引用本文复制引用
王爽,欧阳泽,祝皓轩,殷毅超,周帝宏,王祺,李永豪,王海洋..基于改进YOLOv8的化工泄漏检测方法[J].重庆科技大学学报（自然科学版）,2025,27(2):70-79,10.基金项目
重庆市自然科学基金项目"基于加速寿命试验和不精确概率理论的工业机器人关键件可靠性评估研究"(2022NSCQ-MSX1911) （2022NSCQ-MSX1911）
重庆市教育委员会科学技术研究项目"加速寿命试验下谐波减速器可靠性建模及寿命预测"(KJQN202101539) （KJQN202101539）
