基于光度立体和深度学习的电池缺陷检测方法

doi:10.13229/j.cnki.jdxbgxb.20230130

摘要/Abstract

摘要：

针对电池缺陷检测易受黑色外观干扰，导致仅通过一张单光源下观测图像的局限视觉，无法实现缺陷的有效识别的问题，提出了一种端到端的光度立体视觉缺陷检测模型。首先，利用光度立体特征生成模块生成法线特征，获取物体表面细节信息；然后，采用通道协同注意力机制，探讨特征信道间的相互关系，充分挖掘特征间的关联以自适应地增强全局表示，进一步提升信息表达能力；最后，利用特征金字塔和空间金字塔池化实现多尺度预测，提升分类准确率。在自建Battery 101数据集上的实验结果表明：与其他算法相比，本文算法在检测精度和推理速度上都取得较好效果。此外，消融实验也进一步验证了模型中各个模块的有效性。

关键词: 信号与信息处理, 电池缺陷检测, 光度立体视觉, 通道协同注意力机制

Abstract:

Aiming at the problem that the battery defect detection is susceptible to black appearance interference， which leads to the limited detect identification of the observation image under a single light source， an end-to-end deep-learning photometric stereo network （DPSNet） is proposed. Firstly， the photometric stereo feature generator （PSFG） enables model to facilitate transformation from multi-input features to normal feature so that better excel in defect signals. Then channel co-attention （CCA） module explores the channel interactions between each input and surface normal for informative representations， and fuses modify features to coordinately enhance global representation. Finally， spatial pyramid pooling （SPP） and feature pyramid networks （FPN） achieve multi-scale prediction. The experimental results on the self-built Battery101 dataset show that the proposed method achieves better effect. In addition， the ablation experiment further verifies the effectiveness of each module in the model.

Key words: signal and information processing, battery defect detection, photometric stereo, channel co-attention

中图分类号:

TP391.4

苏育挺,景梦瑶,井佩光,刘先燚. 基于光度立体和深度学习的电池缺陷检测方法[J]. 吉林大学学报(工学版), 2024, 54(12): 3653-3659.

Yu-ting SU,Meng-yao JING,Pei-guang JING,Xian-yi LIU. Deep photometric stereo learning framework for battery defect detection[J]. Journal of Jilin University(Engineering and Technology Edition), 2024, 54(12): 3653-3659.

图/表 6

图1

图2

表1

不同λ参数值对模型效果的影响"

评估指标	$λ$
评估指标	0.01	0.1	1	10	100
mAP/%	87.2	87.6	89.6	86.4	84.3
方差	2.08	1.61	1.44	1.47	5.94

表1

图3

表2

表3

参考文献 34

1	Burges C J C. A tutorial on support vector machines for pattern recognition[J]. Data Mining and Knowledge Discovery, 1998, 2(2): 121-167.
2	Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, San Diego, USA, 2005: 886-893.
3	时亚涛,戴芳,杨畅民. 太阳能光伏电池缺陷检测[J]. 电子测量与仪器学报,2020,34(4):157-164.
	Shi Ya-tao, Dai Fang, Yang Chang-min. Defect detection in solar photovoltaic cells[J]. Journal of Electronic Measurement and Instrumentation, 2020, 34(4): 157-164.
4	刘磊,王冲,赵树旺,等. 基于机器视觉的太阳能电池片缺陷检测技术的研究[J]. 电子测量与仪器学报,2018,32(10):47-52.
	Liu Lei, Wang Chong, Zhao Shu-wang, et al. Research on machine vision-based defect detection technology for solar cells[J]. Journal of Electronic Measurement and Instrumentation, 2018, 32(10): 47-52.
5	Liu W, Anguelov D, Erhan D, et al. SSD: single shot multibox detector[C]∥Proceedings of the European Conference on Computer Vision, Amsterdam, Netherlands, 2016: 21-37.
6	Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 7263-7271.
7	Redmon J, Farhadi A. YOLOv3: an incremental improvement[DB/OL]. [2023-01-08]. .
8	Girshick R. Fast R-CNN[C]∥Proceedings of the IEEE International Conference on Computer Vision, Santiago, Chile, 2015: 1440-1448.
9	Ren S, He K, Girshick R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 39(6): 1137-1149.
10	He K, Gkioxari G, Dollar P, et al. Mask R-CNN[C]∥Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 2017: 2961-2969.
11	郝刚, 金涛. 基于多尺度数据融合的锂电池健康状态评估[J]. 江苏大学学报: 自然科学版, 2023, 44(5): 524-529.
	Hao Gang, Jin Tao. Lithium battery health evaluation based on multi-scale data fusion[J]. Journal of Jiangsu University (Natural Science Edition), 2023, 44(5): 524-529.
12	桂久琪,李林升,毛晓,等. 基于改进YOLOv4的锂电池缺陷检测方法[J]. 电子测量技术,2022,45(15):144-150.
	Gui Jiu-qi, Li Lin-sheng, Mao Xiao, et al. An improved YOLOv4-based defect detection method for lithium batteries[J]. Electronic Measurement Technology, 2022, 45(15): 144-150.
13	Liu L, Zhu Y, Rahman M R U, et al. Surface defect detection of solar cells based on feature pyramid network and GA-Faster-RCNN[C]∥Proceedings of the 2nd China Symposium on Cognitive Computing and Hybrid Intelligence, Xi'an, China, 2019: 292-297.
14	Silver W M. Determining shape and reflectance using multiple images[M]. Massachusetts: Massachusetts Institute of Technology, 1980.
15	Woodham R J. Photometric method for determining surface orientation from multiple images[J]. Optical Engineering, 1980, 19(1): 139-144.
16	Mukaigawa Y, Ishii Y, Shakunaga T. Analysis of photometric factors based on photometiric linearization[J]. Optical Society, 2007, 24(10): 3326-3334.
17	Wu L, Ganesh A, Shi B, et al. Robust photometric stereo via low-rank matrix completion and recovery[C]∥Proceedings of the 10th Asian Conference on Computer Vision, Queenstown, New Zealand, 2010: 703-717.
18	Ikehata S, Wipf D, Matsushita Y, et al. Robust photometric stereo using sparse regression[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Providence, USA, 2012: 318-325.
19	Torrance K E, Sparrow E M. Theory for off-specular reflection from roughened surfaces[J]. Optical Society, 1967, 57: 1105-1114.
20	Chung H S, Jia J. Efficient photometric stereo on glossy surfaces with wide specular lobes[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Anchorage, USA, 2008: No.4587771.
21	Cook R L, Torrance K E. A reflectance model for computer graphics[J]. ACM Transactions on Graphics, 1988, 1(1): 7-24.
22	Santo H, Samejima M, Sugano Y, et al. Deep photometric stereo network[C]∥Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 2017: 501-509.
23	Chen G, Han K, Shi B, et al. Deep photometric stereo for non-lambertian surfaces[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(1): 129-142.
24	刘根,蔡念,肖盼,等. 基于光度立体和图像显著性的皮革缺陷检测[J]. 计算机工程与应用,2019,55(8):215-219.
	Liu Gen, Cai Nian, Xiao Pan, et al. Leather defect detection based on photometric stereo and image saliency[J]. Computer Engineering and Applications, 2019, 55(8): 215-219.
25	Cao Y, Ding B, Chen J, et al. Photometric stereo-based defect detection system for metal parts[J]. Sensors, 2022, 22(21): No.8374.
26	He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.
27	Lin T Y, Dollar P, Girshick R, et al. Feature pyramid networks for object detection[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, 2017: 2117-2125.
28	Lin T Y, Maire M, Belongie S, et al. Microsoft COCO: common objects in context[C]∥Proceedings of the European Conference on Computer Vision(ECCV), Zurich, Switzerland, 2014: 740-755.
29	Lin T Y, Goyal P, Girshick R, et al. Focal loss for dense object detection[C]∥Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 2017: 2980-2988.
30	Nguyen C H, Nguyen T C, Tang T N, et al. Improving object detection by label assignment distillation[C]∥Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision(WACV), Waikoloa, USA, 2022: 1005-1014.
31	Ge Z, Liu S, Wang F, et al. YOLOX: exceeding YOLO series in 2021[DB/OL]. [2023-01-15].
32	Zhang H, Wang Y, Dayoub F, et al. VarifocalNet: an IoU-aware dense object Detector[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Online, 2021: 8510-8519.
33	Zheng Z H, Ye R G, Hou Q B, et al. Localization distillation for dense object detection[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, USA, 2022: 9407-9416.
34	Cheng B, Misra I, Schwing A G, et al. Masked-attention Mask transformer for universal image segmentation[C]∥Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, New Orleans, USA, 2022: 1290-1299.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

方法	骨干网络	缺陷类别						mAP/%	FPS/（f·s^-1）
方法	骨干网络	耳朵	气泡	标签	褶皱	变形	划痕	mAP/%	FPS/（f·s^-1）
SSD300	ResNet-50	76.9	95.8	99.9	53.9	100.0	0	71.1	173.4
YOLOv3	Darknet-53	89.4	97.6	99.9	85.1	100.0	3.2	79.6	193.6
LAD	ResNet-50	78.5	94.4	99.9	64.5	99.9	40.4	79.6	258.3
VFNet	ResNet-50	79.8	97.4	99.9	71.1	100.0	32.6	80.1	254.2
LD	ResNet-50	77.5	100.0	99.9	77.9	100.0	27.7	80.5	182.4
YOLOX	Darknet-53	85.6	100.0	99.9	81.7	100.0	22.8	81.0	252.0
RetinaNet	ResNet-50	92.2	100.0	99.9	56.6	94.4	40.1	82.2	535.7
Mask2Former	ResNet-50	78.2	100.0	99.9	85.1	100.0	39.6	83.8	103.3
Faster RCNN	ResNet-50	92.2	97.6	99.9	93.6	100.0	43.4	87.8	37.0
本文	Darknet-53	99.2	100.0	99.9	93.1	100.0	45.6	89.6	42.6

PSFG	CCA	SPP	mAP
		√	83.6
√			82.7
√	√		84.0
√		√	86.2
√	√	√	89.6

[1]	王海涛,刘慧卓,张学永,韦健,郭校源,肖俊哲. 基于单目视觉的车辆屏显式封闭驾驶舱前视视野重现[J]. 吉林大学学报(工学版), 2024, 54(5): 1435-1442.
[2]	窦慧晶,谢东旭,郭威,邢路阳. 基于改进的正交匹配跟踪算法的波达方向估计[J]. 吉林大学学报(工学版), 2024, 54(12): 3568-3576.
[3]	王春阳,丘文乾,刘雪莲,肖博,施春皓. 基于平面拟合的地面点云精确分割方法[J]. 吉林大学学报(工学版), 2023, 53(3): 933-940.
[4]	李雪梅,王春阳,刘雪莲,施春浩,李国瑞. 基于超体素双向最近邻距离比的点云配准方法[J]. 吉林大学学报(工学版), 2022, 52(8): 1918-1925.
[5]	李雪梅,王春阳,刘雪莲,谢达. 基于SESTH的线性调频连续波激光雷达信号时延估计[J]. 吉林大学学报(工学版), 2022, 52(4): 950-958.
[6]	林乐平,卢增通,欧阳宁. 面向非配合场景的人脸重建及识别方法[J]. 吉林大学学报(工学版), 2022, 52(12): 2941-2946.
[7]	窦慧晶,丁钢,高佳,梁霄. 基于压缩感知理论的宽带信号波达方向估计[J]. 吉林大学学报(工学版), 2021, 51(6): 2237-2245.
[8]	金心宇,谢慕寒,孙斌. 基于半张量积压缩感知的粮情信息采集[J]. 吉林大学学报(工学版), 2021, 51(1): 379-385.
[9]	郭立民,陈鑫,陈涛. 基于AlexNet模型的雷达信号调制类型识别[J]. 吉林大学学报(工学版), 2019, 49(3): 1000-1008.
[10]	林欣堂，李艳东，吴攀超. 新的闭式通用浮雕变换解算法在三维表面检测中的应用[J]. 吉林大学学报(工学版), 2015, 45(6): 1987-1993.