结合分支特征和排斥损失的绝缘子检测研究

doi:10.3778/j.issn.1002-8331.2412-0061

摘要/Abstract

摘要： 针对高压线路中绝缘子检测中存在目标背景多样化、缺陷不规则、闪络缺陷不明显等问题，提出了结合多样化分支特征和排斥损失的YOLOv11绝缘子缺陷检测算法。设计了SEAM（spatially enhanced attention module）模块，将排斥损失（repulsion loss）融入模块中，并对C2PSA（C2PSA-SEAM）模块中的PSA注意力层进行了更新，提出C2PSA-S模块，来补偿被遮挡区域的响应损失，抑制背景多样化对特征图的干扰。并引入GELAN（generalized ELAN）思想，设计使用多分支卷积模块DBB-GC（diverse branches block-GELAN）来增强在特征提取时的空间感知能力，在基准模型中对C3K2内部结构进行重构，提升目标复合特征的质量。在此基础上，加入融合了多层次点采样策略的动态上采样器DySample（upsampling by dynamic），将上采样过程定义为点采样（point sampling），提升上采样的质量和效率。并使用参数化修正线性单元PReLU（parametric rectified linear unit）替换SiLU（sigmoid linear unit），利用参数传播使网络更好地拟合数据。实验结果表明，该模型的mAP@0.5和mAP@0.95分别达到了92.5%和63.0%，较基准模型mAP@0.5和mAP@0.95分别提升了3.4个百分点和2.2个百分点。同时，绝缘子破损、闪络缺陷的检测精度分别提高了5.0个百分点和4.9个百分点，满足背景多样化下绝缘子的实际检测需求。

关键词: 绝缘子检测, YOLOv11, 小目标, 排斥损失, 多分支卷积, 动态采样

Abstract: Aiming at the problems of diverse target backgrounds, irregular defects, and unclear flashover defects in insulator detection in high-voltage transmission lines, a YOLOv11 insulator defect detection algorithm combining diversified branch characteristics and repulsion loss is proposed. It designs the SEAM (spatially enhanced attention module) module, which incorporates repulsion loss into the module and updates the PSA attention layer in the C2PSA (C2PSA-SEAM) module. It proposes the C2PSA-S module to compensate for the response loss of occluded areas and suppress the interference of background diversity on feature maps. And it introduces the GELAN (generalized ELAN) concept, a multi-branch convolution module DBB-GC (diverse branches block-GELAN) is designed to enhance spatial perception ability during feature extraction. The internal structure of C3K2 is reconstructed in the benchmark model to improve the quality of target composite features. On this basis, a dynamic upsampling DySample (upsampling by dynamic) incorporating a multi-level point sampling strategy is added, defining the upsampling process as point sampling to improve the quality and efficiency of upsampling. And it uses parameterized rectified linear unit PReLU (parametric rectified linear unit) to replace SiLU (sigmoid linear unit), and uses parameter propagation to make the network better fit the data. The experimental results indicate that the model mAP@0.5 and mAP@0.95 reach 92.5% and 63.0% respectively, compared to the benchmark model mAP@0.5 and mAP@0.95, they have increased by 3.4 percentage points and 2.2 percentage points respectively. At the same time, the detection accuracy of insulator damage and flashover defects has increased by 5.0 percentage points and 4.9 percentage points respectively, meeting the actual detection needs of insulators under diverse backgrounds.

Key words: insulator defects, YOLOv11, small goals, repulsion loss, diverse branches block-GELAN, upsampling by dynamic

季星宇, 黄陈蓉, 姚军财, 王凯. 结合分支特征和排斥损失的绝缘子检测研究[J]. 计算机工程与应用, 2025, 61(7): 141-152.

JI Xingyu, HUANG Chenrong, YAO Juncai, WANG Kai. Research on Insulator Detection Based on Branch Characteristics and Repulsion Loss[J]. Computer Engineering and Applications, 2025, 61(7): 141-152.

参考文献

[1] ZHANG J R, WEI X, ZHANG L X, et al. YOLO v7-ECA-PConv-NWD detects defective insulators on transmission lines[J]. Electronics, 2023, 12(18): 3969-3978.
[2] SHENG W S, YU X F, LIN J Y, et al. Faster RCNN target detection algorithm integrating CBAM and FPN[J]. Applied Sciences, 2023, 13(12): 6913-6921.
[3] XIANG W B, SONG Z Y, ZHANG G X, et al. Birds detection in natural scenes based on improved faster RCNN[J]. Applied Sciences, 2022, 12(12): 6094-6102.
[4] 何湘杰, 宋晓宁. YOLOv4-Tiny的改进轻量级目标检测算法[J]. 计算机科学与探索, 2024, 18(1): 138-150.
HE X J, SONG X N. Improved YOLOv4-Tiny lightweight target detection algorithm[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(1): 138-150.
[5] HOU J L, CHE Y H, FANG Y R, et al. Early bruise detection in apple based on an improved faster RCNN model[J]. Horticulturae, 2024, 10(1): 100-107.
[6] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016: 779-788.
[7] WANG C Y, BOCHKOVSKIY A, LIAO H M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 7464-7475.
[8] TAN H C, ZHOU J, TIAN S J, et al. SFE-SSD: shallow feature enhancement SSD for small object detection[J]. Journal of Mathematical Research with Applications, 2019, 39(6): 733-744.
[9] MA L, LIU Y, ZHANG X L, et al. Deep learning in remote sensing applications: a meta-analysis and review[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2019, 152: 166-177.
[10] ZHOU M, WANG J, LI B. ARG-mask RCNN: an infrared insulator fault-detection network based on improved mask RCNN[J]. Sensors, 2022, 22(13): 4720-7433.
[11] ZHANG D, LI J J, XIONG L, et al. Cycle-consistent domain adaptive faster RCNN[J]. IEEE Access, 2019, 7: 123903-123911.
[12] HOU J W, YANG C, HE Y H, et al. Detecting diseases in apple tree leaves using FPN-ISResNet-Faster RCNN[J]. European Journal of Remote Sensing, 2023, 56(1): 623-631.
[13] 赵佰亭, 程瑞丰, 贾晓芬. 融合多尺度特征的YOLOv8裂缝缺陷检测算法[J]. 计算机工程与应用, 2024, 60(22): 261-270.
ZHAO B T, CHENG R F, JIA X F. YOLOv8 crack defect detection algorithm based on multi-scale features[J]. Computer Engineering and Applications, 2024, 60(22): 261-270.
[14] LIU W R, WANG S, GAO X J, et al. A tomato recognition and rapid sorting system based on improved YOLOv10[J]. Machines, 2024, 12(10): 689-696.
[15] QIU X Y, CHEN Y J, CAI W H, et al. LD-YOLOv10: a lightweight target detection algorithm for drone scenarios based on YOLOv10[J]. Electronics, 2024, 13(16): 3269-3276.
[16] SONG Q, YAO B D, XUE Y L, et al. MS-YOLO: a lightweight and high-precision YOLO model for drowning detection[J]. Sensors, 2024, 24(21): 6955-6964.
[17] 姚景丽, 程光, 万飞, 等. 改进YOLOv8的轻量化轴承缺陷检测算法[J]. 计算机工程与应用, 2024, 60(21): 205-214.
YAO J L, CHENG G, WAN F, et al. Improved lightweight bearing defect detection algorithm of YOLOv8[J]. Computer Engineering and Applications, 2024, 60(21): 205-214.
[18] 赵其昌, 吴一全, 苑玉彬. 光学遥感图像舰船目标检测与识别方法研究进展[J]. 航空学报, 2024, 45(8): 51-84.
ZHAO Q C, WU Y Q, YUAN Y B. Progress of ship detection and recognition methods in optical remote sensing images[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(8): 51-84.
[19] 王春梅, 刘欢. YOLOv8-VSC: 一种轻量级的带钢表面缺陷检测算法[J]. 计算机科学与探索, 2024, 18(1): 151-160.
WANG C M, LIU H. YOLOv8-VSC: lightweight algorithm for strip surface defect detection[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(1): 151-160.
[20] KHANAM R, HUSSAIN M. YOLOv11: an overview of the key architectural enhancements[J]. arXiv:2410.17725， 2024.
[21] CHEN Q, WANG Y M, YANG T, et al. You only look one-level feature[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 13034-13043.
[22] LIANG F T, ZHOU Y, CHEN X, et al. Review of target detection technology based on deep learning[C]//Proceedings of the 5th International Conference on Control Engineering and Artificial Intelligence. New York: ACM, 2021: 132-135.
[23] YU Z P, HUANG H B, CHEN W J, et al. YOLO-FaceV2: a scale and occlusion aware face detector[J]. Pattern Recognition, 2024, 155: 110714.
[24] WANG X L, XIAO T T, JIANG Y N, et al. Repulsion loss: detecting pedestrians in a crowd[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7774-7783.
[25] WEI J F, NI L Y, LUO L, et al. GFS-YOLO11: a maturity detection model for multi-variety tomato[J]. Agronomy, 2024, 14(11): 2644.
[26] 李琳, 靳志鑫, 俞晓磊, 等. Haar小波下采样优化YOLOv9的道路车辆和行人检测[J]. 计算机工程与应用, 2024, 60(20): 207-214.
LI L, JIN Z X, YU X L, et al. Road vehicle and pedestrian detection based on YOLOv9 for haar wavelet downsampling[J]. Computer Engineering and Applications, 2024, 60(20): 207-214.
[27] LIN Z, YUN B, ZHENG Y. LD-YOLO: a lightweight dynamic forest fire and smoke detection model with Dysample and spatial context awareness module[J]. Forests, 2024, 15(9): 1630-1642.
[28] LI Y G, LI Q, PAN J, et al. SOD-YOLO: small-object-detection algorithm based on improved YOLOv8 for UAV images[J]. Remote Sensing. 2024, 16(16): 3057-3068.
[29] WANG X Q, GAO H B, JIA Z M, et al. BL-YOLOv8: an improved road defect detection model based on YOLOv8[J]. Sensors, 2023, 23(20): 8361-9373.
[30] WU T Y, DONG Y K. YOLO-SE: improved YOLOv8 for remote sensing object detection and recognition[J]. Applied Sciences, 2023, 13(24): 12977-12986.
[31] MEI J H, ZHU W Q. BGF-YOLOv10: small object detection algorithm from unmanned aerial vehicle perspective based on improved YOLOv10[J]. Sensors, 2024, 24(21): 6911-6922.