融合多尺度特征的高效网片缺陷检测算法

doi:10.3778/j.issn.1002-8331.2403-0126

摘要/Abstract

摘要： 针对人工目视排查大型旋转过滤设备网片缺陷时存在的效率低下，缺陷与背景之间边界模糊及网孔中的小水珠产生反光现象等问题，提出了一个基于多维特征融合的高效网片缺陷检测算法。引入了泊松图像增强技术，实现了缺陷目标与正常背景区域的平滑融合，增加了小样本缺陷数量的同时解决了缺陷数量分布不均匀的问题。在YOLOv8中融入轻量多维卷积改进的C2f_LWDC（C2f_lightweight multi-dimensional convolution）模块及加权多特征增强模块，既增强了网络对缺陷特征的提取又实现了各级特征的高效融合，提升了对多尺度缺陷样本的表征能力。采用EIOU（efficient intersection over union）定位损失函数，加速了对缺陷目标的准确定位。网片数据集检测结果表明，改进后的算法mAP（mean average precision）达到92%，相较于原始模型提升了16.8个百分点，能很好地完成缺陷目标的检测任务。

关键词: 网片缺陷, YOLOv8, 轻量多维卷积, 特征融合, 多尺度

Abstract: Addressing the inefficiency of manual visual inspection for mesh defects in large rotary filtration equipment, and considering the blurred boundaries between the defects and the background, and the reflection phenomenon caused by the small water droplets in the mesh holes, this paper proposes an efficient mesh defect detection algorithm based on multi-dimensional feature fusion. Firstly, the Poisson image enhancement technique is applied to achieve seamless fusion between defect targets and normal background areas thereby enhancing the visual coherence of the composite image. This approach effectively addresses the issue of uneven distribution of defect numbers by increasing the representation of small sample defects. Next, the paper integrates the lightweight multi-dimensional convolutional improved C2f_LWDC(C2f_lightweight multi-dimensional convolution) module and the weighted multi-feature enhancement module into YOLOv8. This integration not only enhances the extraction of defect features but also achieves efficient fusion of features at all levels. As a result, it improves the characterization ability of multi-scale defect samples. Lastly, the EIOU (expected intersection over union) localization loss function is employed to expedite the precise localization of the defect target. The detection results of mesh dataset show that the improved algorithm mAP(mean average precision) reaches 92%, which is 16.8 percentage points higher than that of the original model, and the improved algorithm can well complete the detection task of defect targets.

Key words: mesh defect, YOLOv8, lightweight multidimensional convolution, feature fusion, multiscale

何钢, 姚远, 韩征彤, 邹华涛, 王田田. 融合多尺度特征的高效网片缺陷检测算法[J]. 计算机工程与应用, 2025, 61(11): 316-324.

HE Gang, YAO Yuan, HAN Zhengtong, ZOU Huatao, WANG Tiantian. Efficient Mesh Defect Detection Algorithm Integrating Multi-Scale Features[J]. Computer Engineering and Applications, 2025, 61(11): 316-324.

参考文献

[1] 杨庆, 罗泳, 陈天朋, 等. 基于潮位的鼓网套筒智能冲洗方案研究[J]. 中国核电, 2023, 16(4): 599-604.
YANG Q, LUO Y, CHEN T P, et al. Study on the intelligent flushing scheme of the drum-net sleeve based on tide level[J]. China Nuclear Power, 2023, 16(4): 599-604.
[2] VILLARD P F, BOUDART M, ILEA I, et al. Anomaly detection on textured images with convolutional neural network for quality control of micrometric woven meshes[J]. Fluid Dynamic and Material Process, 2022, 18(6): 1639-1648.
[3] 俞文静, 刘航, 李梓瑞, 等. 基于图像增强和CNN的布匹瑕疵检测算法[J]. 计算机技术与发展, 2021, 31(5): 90-95.
YU W J, LIU H, LI Z R, et al. A fabric defect detec- tion algorithm based on image enhancement and CNN[J]. Computer Technology and Development, 2021, 31(5): 90-95.
[4] 彭晏飞, 袁晓龙, 陈炎康, 等. 改进YOLOv5s的带钢表面缺陷检测方法[J/OL]. 机械科学与技术: 1-9[2025-05-10]. https://doi.org/10.13433/j.cnki.1003-8728.20230254.
PENG Y F, YUAN X L, CHEN Y K, et al. Improved YOLOv5s strip surface defect detection method[J]. Mechanical Science and Technology for Aerospace Engineering: 1-9[2025-05-10]. https://doi.org/10.13433/j.cnki.1003-8728.20230254.
[5] 李斌, 曾筠婷. 基于多尺度上下文感知的绝缘子缺陷检测网络[J]. 高电压技术, 2022, 48(8): 2905-2914.
LI B, ZENG J T. Detection network for insulator defects based on multi-scale context awareness[J]. High Voltage Engineering, 2022, 48(8): 2905-2914.
[6] 裴少通, 张行远, 胡晨龙. 基于ER-YOLO算法的跨环境输电线路缺陷识别方法[J]. 电工技术学报, 2024, 39(9): 2825-2840.
PEI S T, ZHANG X Y, HU C L. The defect detection method for cross-environment power transmission line based on ER-YOLO algorithm[J]. Transactions of China Electrotechnical Society, 2024, 39(9): 2825-2840.
[7] ZHANG Y F, REN W, ZHANG Z, et al. Focal and effi cient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146-157.
[8] 赵崇林, 朱江. 融合注意力和多尺度特征的航空发动机缺陷检测[J]. 北京航空航天大学学报, 2023, 67(5): 1-14.
ZHAO C L, ZHU J. Aero-engine defect detection based on fusing attention and multi-scale features[J]. Journal of Beijing University of Aeronautics and Astronautics, 2023, 67(5): 1-14.
[9] YUN S, HAN D, OH S J, et al. CutMix: regularization strategy to train strong classifiers with localizable features[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019: 6023-6032.
[10] HAO W, ZHILI S. Improved Mosaic: algorithms for more complex images[J]. Journal of Physics: Conference Series, 2020, 1684(1): 012094.
[11] TAN M, PANG R, LE Q V. Efficientdet: scalable and efficient object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 10781-10790.
[12] CHOLLET F. Xception: deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 1251-1258.
[13] LI X, WANG W, HU X, et al. Selective kernel net works[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 510-519.
[14] 黄仁彬, 詹道桦. 基于加权多尺度特征融合的带钢表面缺陷检测算法[J]. 计算机集成制造系统, 2024, 30(3): 1-17.
HANG R B, ZHAN D H. Surface defect detection algo rithm for strip steel based on weighted multi-scale feature fusion[J]. Computer Integrated Manufacturing Systems, 2024, 30(3): 1-17.
[15] ROY A M, BOSE R, BHADURI J. A fast accurate fine-grain object detection model based on YOLOv4 deep neural network[J]. Neural Computing and Applications, 2022, 34(5): 3895-3921.
[16] ZHANG J, LIN S, DING L, et al. Multi-scale context aggregation for semantic segmentation of remote sensing images[J]. Remote Sensing, 2020, 12(4): 701.
[17] ZHENG Z, WANG P, LIU W, et al. Distance-IoU loss: faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 12993-13000.
[18] LI Y, MAO H, GIRSHICK R, et al. Exploring plain vi sion transformer backbones for object detection[C]//Proceedings of the European Conference on Computer Vision, 2022: 280-296.
[19 KIRILLOV A, MINTUN E, RAVI N, et al. Segment any thing[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023: 4015-4026.
[20] WANG C Y, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464-7475.
[21] ZUO C, QIAN J, FENG S, et al. Deep learning in optical metrology: a review[J]. Light: Science & Applications, 2022, 11(1): 1-54.