多成像融合的带电作业区域预测算法研究OA
Research on Prediction Algorithms for Live Working Areas Based on Multi-Imaging Fusion
为保障带电作业人员的安全,提出基于改进 YOLO 算法模型(you only look once version 8,YOLOv8)的可见光与红外成像相融合的带电作业区域预测算法,通过多模态数据互补方法提高了带电作业边界的精确性与可靠性.首先,在 YOLOv8 中引入相似度注意力模块(simple parameter-free attention module,SimAM),提高特征提取能力;其次,结合自适应可变形卷积(adaptive deformable convolution v2,A-DCNv2),智能选择使用常规卷积或可变形卷积来提高小目标检测精度;最后,将可见光检测结果与红外温度数据相融合,采用自适应加权融合算法来改善和优化边界判定,确保高低温区域与视觉目标的位置相匹配.试验结果表明,在不增加计算量的条件下,该算法提升了目标检测的鲁棒性.
To ensure the safety of live working operaters,it proposes a live working area prediction algorithm based on the fusion of visible light and infrared imaging and the improved you only look once version 8(YOLOv8)algorithm model.It improves the accuracy and reliability of the live working boundary through the multimodal data complementarity method.Firstly,among these improvements,it introduces the simple parameter-free attention module(SimAM)attention mechanism into YOLOv8 to enhance feature extraction ca-pabilities.Secondly,it incorporates adaptive deformable convolution adaptive deformable convolution v2(A-DCNv2)to improve the detection accuracy of small targets.Finally,it fuses the visible light detection results with infrared temperature data using an adaptive weighted fusion algorithm to optimize boundary determination,ensuring precise positional matching between high/low-temperature areas and visual targets.Experimental results demonstrate that the proposed algorithm enhances the model's robustness for target detection with almost no increase in computational load.
孙学斌;王家峰;刘世健;孙绍郁;刘永辉
国网辽宁省电力有限公司鞍山供电公司,辽宁 鞍山 114002国网辽宁省电力有限公司鞍山供电公司,辽宁 鞍山 114002国网辽宁省电力有限公司鞍山供电公司,辽宁 鞍山 114002国网辽宁省电力有限公司鞍山供电公司,辽宁 鞍山 114002国网辽宁省电力有限公司鞍山供电公司,辽宁 鞍山 114002
信息技术与安全科学
YOLOv8SimAM注意力机制多模态融合带电作业
YOLOv8SimAM attention mechanismmultimodal fusionlive working
《东北电力技术》 2026 (9)
33-37,5
评论