摘要
针对城市燃气管网泄漏检测中存在的识别精度低、响应时效差等问题,文章提出了一种基于改进YOLOv8算法的智能检测系统。通过引入CBAM注意力机制与可变形大核注意力模块(D-LKA-Attention),该系统有效增强了对小尺度泄漏目标的检测能力,并融合地理信息系统(GIS)实现了泄漏点的精准定位与多级报警功能。实验结果表明,改进模型在燃气管道场景中平均精度(mAP@0.5)达到82.3%,相比原始YOLOv8提升3.1个百分点,检测速度达38 FPS,满足实时性需求。
In response to the problems of low accuracy and poor timeliness in leak detection of urban gas pipelines,an intelligent detection system based on the improved YOLOv8 algorithm is proposed.By integrating CBAM attention mechanism and deformable large kernel attention(D-LKA Attention),the small target detection capability is optimized,and combined with geographic information system(GIS)to achieve precise leak point localization and multi-level alarm.Experiments have shown that the improved model is effective in gas pipeline scenarios mAP@0.5 Reached 82.3%,an increase of 3.1 percentage points compared to the original YOLOv8,with a detection speed of 38 FPS,meeting real-time requirements.
作者
李敏
李宁宁
LI Min;LI Ningning(Chiping Hengyuan Gas Co.,Ltd.,Liaocheng,Shandong 252131,China)
