In the gas insulated substation (GIS) intelligent visualization monitoring system, factors such as internal complex structures and fully enclosed spaces affect the quality of the collected images, leading to issues like uneven lighting and blurred details, which result in missed and false detection. An improved model based on YOLOv8 is proposed to enhance the efficiency and accuracy of instance segmentation for GIS isolation switch images, thereby obtaining more precise on-off judgment. Firstly, the original C2f module is replaced with the RepNCSPELAN4 module, which splits the input channels into two parts for separate processing before merging the results, making more efficient use of channel information and enhancing the model's ability to recognize targets of different sizes. Secondly, the proposed SPPELAN_LSKA model is composed of the SPPELAN module fused with the LSKA attention mechanism, further improving the model's feature extraction capabilities in images with uneven lighting and blurred details. Lastly, the redesigned segmentation head, Segment_EfficientHead, significantly reduces the number of parameters, enhancing the model's segmentation efficiency and accuracy. Experimental results show that on the self-construction GIS isolation switch data-set, the GIS-YOLO model outperforms the YOLOv8 model in terms of precision, mask@0.5, and mask@0.5∶0.95 by 8.1, 7.8, and 16.8 percentage points, respectively, while reducing the model's parameter count and computational load by 37.3% and 13.3%, respectively. The improved model not only has higher segmentation performance but also possesses a more lightweight structure, addressing the requirements for online instance segmentation tasks of GIS isolation switch images.
PANGXianhai, GUChaomin, LUShijie, et al. Comparative study on propagation characteristics of optical and electrical signals during partial discharge in GIS[J]. High Voltage Apparatus, 2023, 59(8): 146-153. (in Chinese)
WUXiaohu, DongmingLÜ, GAOMeihua, et al. Structure types of three-position disconnector and earthing switches for high voltage GIS[J]. High Voltage Apparatus, 2024, 60(1): 197-204. (in Chinese)
FANGZhou, ZHANGWei, LIUHui, et al. Pattern recognition algorithm and monitoring system of GIS partial discharge based on image morphological features[J]. Electrical Engineering, 2024, 25(10): 48-54. (in Chinese)
HAOJian, LIUQingsong, PENGHuadong, et al. Characteristic difference analysis and identification method of typical mechanical defects for GIS three-position switches[J]. High Voltage Engineering, 2024, 50(6): 2567-2579. (in Chinese)
XUYuan, LIUWeidong, CHENWeijiang, et al. High-sensitivity measurement method and application of GIS spacer partial discharge[J]. Proceedings of the CSEE, 2020, 40(5): 1703-1713. (in Chinese)
LIWei, ZHANGLiangen, LILeying, et al. Analysis on UHF signal intermittency and effective detection rate of partial discharge of metal particles on GIS insulator surface[J]. Insulating Materials, 2024, 57(8): 122-129. (in Chinese)
LIUZhao, LIXiaoxi, GONGYanpeng, et al. A partial discharge defect caused by cracks in GIS internal insulator based on acoustic and optic joint[J]. Guangdong Electric Power, 2023, 36(9): 100-106. (in Chinese)
ZHANGGuozhi, HUXukun, DENGGuangyu, et al. Experimental study on substrate compatibility of SF6 and SF6 faultdecomposing gases with partial discharge flexible uhf antenna sensors[J]. Transactions of China Electrotechnical Society, 2023, 38(15): 4050-4062. (in Chinese)
LIANGBoyuan, GAOJun, LIUHongliang, et al. Fault detection and location of gas insulated switchgear based on vibration characteristic estimation[J]. Science Technology and Engineering, 2020, 20(31): 12836-12842. (in Chinese)
ZHANGYongqiang, GUOCheng, ZHANGHaojun, et al. GIS contact position monitoring system based on image recognition technology[J]. Electrical Engineering, 2019, 20(5): 51-54. (in Chinese)
[21]
LONGJ, SHELHAMERE, DARRELLT. Fully convolutional networks for semantic segmentation[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015: 3431-3440.
[22]
CHENL, PAPANDREOUG, KOKKINOSI, et al. DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834-848.
[23]
RONNEBERGERO, FISCHERP, BROXT. U-Net: convolutional networks for biomedical image segmentation[M]//Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015. Cham: Springer, 2015: 234-241.
[24]
RENS, HEK, GIRSHICKR, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[25]
HEK, GKIOXARIG, DOLLÁRP, et al. Mask R-CNN[C]//2017 IEEE International Conference on Computer Vision (ICCV), 2017: 2980-2988.
LIAOYanghang, DIANSongyi, ZHONGYuzhong, et al. Foreign object detection method in GIS based on improved YOLOv3[J]. Computer Simulation, 2023, 40(3): 529-535. (in Chinese)
[30]
WANGB, DONGM, RENM, et al. Automatic fault diagnosis of infrared insulator images based on image instance segmentation and temperature analysis[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(8): 5345-5355.
LIUYunpeng, ZHANGZhe, PEIShaotong, et al. Faulty insulator segmentation method in infrared image based on deep learning[J]. Electrical Measurement & Instrumentation, 2022, 59(9): 63-68. (in Chinese)
ZHAOWeida, CHENHaiwen, GUOLuyang, et al. Substation meter readings and dial information identification method based on YOLO-E and enhanced OCRNet image segmentation[J]. Electric Power Construction, 2023, 44(11): 75-85. (in Chinese)
WANGFuqi, WANGZhifeng, JINJiancheng, et al. Detection method for gangue mixed ratio in fully mechanized caving faces based on the GSL-YOLO model[J]. Journal of Mine Automation, 2024, 50(9): 59-65. (in Chinese)
LUZice, LIUXiaofang, WANGDewei. Semantic segmentation method for PCB solder joint based on improved YOLOv8[J]. Radio Engineering, 2024, 54(7): 1614-1621. (in Chinese)
[44]
LAUK W, POL M, REHMANY A U. Large separable kernel attention: rethinking the large kernel attention design in CNN[J]. Expert Systems with Applications, 2024, 236: 121352.
[45]
WANGC Y, MARK LIAOH Y, WUY H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2020: 1571-1580.
[46]
ZHANGX, ZENGH, GUOS, et al. Efficient long-range attention network for image super-resolution[M]//Computer Vision-ECCV 2022. Cham: Springer Nature Switzerland, 2022: 649-667.
[47]
DINGX, ZHANGX, MAN, et al. RepVGG: making VGG-style ConvNets great again[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021: 13728-13737.
[48]
LIX, WANGW, HUX, et al. Generalized Focal Loss V2: learning qualified and distributed bounding boxes for dense object detection[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021: 11627-11636.
[49]
WANGX, ZHANGR, SHENC, et al. SOLO: a simple framework for instance segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(11): 8587-8601.
[50]
CHENGB, MISRAI, SCHWINGA G, et al. Masked-attention mask transformer for universal image segmentation[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022: 1280-1289.