基于 Yolov7-Tiny 的车身漆面损伤检测

Surface Damage Detection of Vehicle Paint Based on Yolov7-Tiny

摘要

针对 Yolov7-Tiny(You Only Look Once v7 Tiny)在车身漆面微小损伤以及不明显损伤检测中存在特征提取能力不足、 细节信息丢失和计算效率低的问题, 提出一种改进的 Yolov7-Tiny 算法。 首先在主干网络中引入RepNCSPELAN4(Re-parameterizable Cross Stage Partial Efficient Layer Aggregation Networks) 模块替换 ELAN-Tiny(Efficient Layer Aggregation Networks Tiny)模块增强模型的特征提取能力; 其次引入 Swish 激活函数以提升模型的非线性表达能力; 最后在颈部网络引入空间到深度卷积 SPDConv(Space to Depth Convolution), 结合改进的SPDConv_Res(Space to Depth Convolution_Residual)检测头以保留微小目标的细节信息并平衡参数负担; 并在颈部网络引入分布偏移卷积 DSConv(Distribution Shift Convolution), 保证缺陷特征信息充分融合的同时提升计算效率。 在自建数据集上的实验结果表明, 改进后的模型 mAP@ 50 达到了 86. 0% , 相比基线模型提升 7. 4% ,计算量降低 3. 3×10⁹ 次。 在提升精度的同时, 一定程度上减少了计算资源的需求。

Abstract

Aiming at the problems of insufficient feature extraction ability, loss of detail information, and low computational efficiency in the detection of minor damages and unobvious damages on the car body paint surface by Yolov7-Tiny(You Only Look Once v7 Tiny), an improved Yolov7-Tiny algorithm is proposed. Firstly, the RepNCSPELAN4 (Re-parameterizable Cross Stage Partial Efficient Layer Aggregation Networks) module is introduced into the backbone network to replace the ELAN-Tiny(Eficienta Layer Aggregation Networks Tiny) module, to enhance the feature extraction ability of the model. Secondly, the Swish activation function is introduced to improve the nonlinear expression ability of the model. Finally, the SPDConv (Space to Depth Convolution) is introduced into the neck network. Combined with the improved SPDConv_Res(Space to Depth Convolution _ Residual) detection head, it retains the detail information of small targets and balances the parameter burden. The DSConv(Distribution Shift Convolution) is introduced into the neck network, ensuring the full fusion of defect feature information and improving the computational efficiency simultaneously. Experimental results on the self-built dataset show that the mAP@ 50 of the improved model reaches 86. 0% , which is 7. 4% higher than that of the baseline model, and the computational amount is reduced by 3. 3×10 ⁹ times. While improving the accuracy, it also reduces the demand for computational resources to a certain extent.

关键词

Yolov7-Tiny 目标检测算法 / 缺陷检测 / 漆面检测 / 空间到深度卷积 / 分布偏移卷积

Key words

You only look once v7 tiny detection algorithm(YOLOv7-Tiny Object Detection Algorithm) / defect detection / paint inspection / space to depth convolution(SPDConv) / distribution shift convolution(DSConv)

参考文献

[1] 邹彦艳, 曹衍芬, 张馨月, 等. 基于 YOLOv8-DSG 的钢铁表面缺陷检测算法[J]. 吉林大学学报(信息科学版), 2025, 43(1): 116-125.

[2] ZOU Y Y, CAO Y F, ZHANG X Y, et al. Algorithm for Defect Detection of Steel Surface Based on YOLOv8-DSG[J]. Journal of Jilin University (Information Science), 2025, 43(1): 116-125.

[3] 程晓雪, 张亚磊, 毕静, 等. 漆面外观缺陷智能检测系统技术应用[J]. 汽车实用技术, 2024, 49(24): 78-82.

[4] CHENG X X, ZHANG Y L, BI J, et al. Application of Intelligent Detection System Technology for Paint Surface Appearance Defects[J]. Automobile Applied Technology, 2024, 49(24): 78-82.

[5] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York, NY, USA : IEEE, 2014: 580-587.

[6] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single Shot Multibox Detector[C] // Computer Vision-ECCV 2016: 14th European Conference. Amsterdam, The Netherlands: Springer International Publishing, 2016: 21-37.

[7] REDMON J, DIVVALA S, GIRSHICK R, et al. You Only Look Once: Unified, Real-Time Object Detection[C] // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA: IEEE, 2016: 779-788.

[8] 张利巍, 杨万帅. 基于改进 Yolov3-Tiny 的加油站目标检测算法研究[J]. 吉林大学学报(信息科学版), 2024, 42(3): 559-566.

[9] ZHANG L W, YANG W S. Research on Gas Station Target Detection Algorithm Based on Improved Yolov3-Tiny[J]. Journal of Jilin University (Information Science), 2024, 42(3): 559-566.

[10] WANG C Y, BOCHKOVSKIY A, 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. Vancouver, Canada: IEEE, 2023: 7464-7475.

[11] 徐威, 李为相, 方志, 等. 基于改进 Yolov7-tiny 的光伏电池缺陷检测算法[J]. 计算机工程与应用, 2024, 60(15): 336-343.

[12] XU W, LI W X, FANG Z, et al. Solar Cell Defect Detection Algorithm Based on Improved Yolov7-Tiny[J]. Computer Engineering and Applications, 2024, 60(15): 336-343.

[13] 阳丽莎, 李茂军, 胡建文, 等. 基于改进 Yolov7-Tiny 的带钢表面缺陷检测算法[J]. 计算机工程, 2025, 51(1): 208-215.

[14] YANG L S, LI M J, HU J W, et al. Strip Steel Surface Defect Detection Algorithm Based on Improved Yolov7-Tiny[J]. Computer Engineering, 2025, 51(1): 208-215.

[15] 周世睿, 陶楚青, 费敏学. 基于多传感器融合的隧道运营风险感知方法[J]. 吉林大学学报(信息科学版), 2025, 43(2): 347-354.

[16] ZHOU S R, TAO C Q, FEI M X. Multi-Sensor Based Method for State Perception of Tunnel Operation and Maintenance[J]. Journal of Jilin University (Information Science), 2025, 43(2): 347-354.

[17] 罗箫瑜, 张志. 基于改进 YOLOX 的变电站设备缺陷检测方法[J]. 吉林大学学报(信息科学版), 2023, 41(5): 848-857.

[18] LUO X Y, ZHANG Z. Defect Detection for Substation Based on Improved YOLOX[J]. Journal of Jilin University (Information Science), 2023, 41(5): 848-857.

[19] WANG C Y, YEH I H, MARK LIAO H Y. Yolov9: Learning What You Want to Learn Using Programmable Gradient Information[C]// European Conference on Computer Vision. Berlin: Springer Nature Switzerland, 2024: 1-21.

[20] 袁硕, 刘玉敏, 安志伟, 等. 基于改进 ShuffleNetV2 网络的岩石图像识别[J]. 吉林大学学报(信息科学版), 2023, 41(3): 450-458.

[21] YUAN S, LIU Y M, AN Z W, et al. Rock Image Recognition Based on Improved ShuffleNetV2 Network[J]. Journal of Jilin University (Information Science), 2023, 41(3): 450-458.

[22] SUNKARA R, LUO T. No More Strided Convolutions or Pooling: A New CNN Building Block for Low-Resolution Images and Small Objects[C] // Proceedings of the 2022 European Conference on Machine Learning and Knowledge Discovery in Databases. Grenoble: Springer, 2022: 443-459.

[23] 江晟, 张仲义, 汪宗洋, 等. 基于改进 Yolov7 的交通路口目标识别算法[J]. 吉林大学学报(理学版), 2024, 62(3): 665-673.

[24] JIANG S, ZHANG Z Y, WANG Z Y, et al. Target Recognition Algorithm of Traffic Intersection Based on Improved Yolov7[J]. Journal of Jilin University (Science Edition), 2024, 62(3): 665-673.

[25] 欧阳继红, 王梓明, 刘思光. 改进多尺度特征的 Yolo_v4 目标检测方法[J]. 吉林大学学报(理学版), 2022, 60(6): 1349-1355.

[26] OUYANG J H, WANG Z M, LIU S G. Yolo_v4 Object Detection Method with Improved Multi-Scale Features[J]. Journal of Jilin University (Science Edition), 2022, 60(6): 1349-1355.

[27] NASCIMENTO M G, FAWCETT R, PRISACARIU V A. Dsconv: Efficient Convolution Operator[C] // Proceedings of the IEEE/ CVF International Conference on Computer Vision. Seoul, Republic of Korea: IEEE, 2019: 5148-5157.

[28] 何乐华, 谢光珍, 刘柯翔, 等. 基于 Yolov5 的倒地检测[J]. 吉林大学学报(信息科学版), 2024, 42(2): 378-386.

[29] HE L H, XIE G Z, LIU K X, et al. Ground Detection Based on Yolov5[J]. Journal of Jilin University (Information Science), 2024, 42(2): 378-386.

基金资助

河北省自然科学基金资助项目(D2022107001)