Researching | Improved Faster R-CNN-Based Contact Lens Surface Defect Detection

Laser & Optoelectronics Progress, Volume. 60, Issue 20, 2015004(2023)

Improved Faster R-CNN-Based Contact Lens Surface Defect Detection

Hao Wang¹, Tao Zha¹, Lingmei Nie¹, Jun Zhang², Yuxi Tang², and Youquan Zhao^1、*

Author Affiliations

¹School of Precision Instrument and Opto-Electronics Engineering, Tianjin University, Tianjin 300072, China

²Gansu Constar Technology Group Co., Ltd., Baiyin 730900, Gansu , China

show less

Abstract Get PDF(in Chinese)

Figures&Tables (8)

References (14)

Paper Information

Figures & Tables(8)

Fig. 1. Structure of Faster R-CNN with FPN

Download full size

Fig. 2. Detection platform. (a) Structure of detection platform; (b) physical image of detection platform

Download full size

Fig. 3. Several defects of contact lens. (a) Without defect; (b) bubble; (c) turning point; (d) scratch; (e) mold point; (f) mold edge

Download full size

Fig. 4. Comparison of train loss and precision. (a) Loss function curve; (b) precision curve

Download full size

Fig. 5. K-means++ clustering result

Download full size

Fig. 6. Results of contact lens defect detection obtained by improved Faster R-CNN algorithm

Download full size

Table 1. Result of defect detection statistics

Table 1. Result of defect detection statistics

Defect	Total number of images	Number pf correct detected images	Number pf false detected images	Number pf missed detected images	Accuracy /%
Bubble	60	60	0	0	100
Turning point	42	42	0	0	100
Scratch	8	8	0	0	100
Mold point	11	11	0	0	100
Mold edge	16	16	0	0	100

Table 2. Performance comparison of improved Faster R-CNN

Table 2. Performance comparison of improved Faster R-CNN

Model	Backbone	AP /%					mAP /%	Time /ms
Model	Backbone	Bubble	Turning point	Scratch	Mold point	Mold edge	mAP /%	Time /ms
Faster R-CNN	MobileNetv2	42.08	56.62	23.56	44.64	32.55	39.89	14.33
Faster R-CNN	VGG16	79.64	61.55	8.14	63.63	77.44	58.08	25.38
Faster R-CNN	ResNet50	88.24	79.21	22.80	100.00	97.18	77.50	16.71
Faster R-CNN+FPN	ResNet50	96.47	69.08	54.17	99.17	98.91	83.56	23.70
Faster R-CNN+FPN+K-means++	ResNet50	93.43	75.94	66.15	100.00	99.23	86.95	24.05
YOLOv3	Darknet53	82.91	51.44	30.93	82.11	71.46	63.77	11.61
SSD	ResNet50	83.87	47.55	9.51	81.19	87.43	64.91	10.14

Tools

Get Citation

Copy Citation Text

Hao Wang, Tao Zha, Lingmei Nie, Jun Zhang, Yuxi Tang, Youquan Zhao. Improved Faster R-CNN-Based Contact Lens Surface Defect Detection[J]. Laser & Optoelectronics Progress, 2023, 60(20): 2015004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Nov. 22, 2022

Accepted: Dec. 19, 2022

Published Online: Sep. 28, 2023

The Author Email: Youquan Zhao (zhaoyouquan@tju.edu.cn)

DOI:10.3788/LOP223132

Topics

laser devices and laser physics

Lasers and Laser Optics

laser manufacturing

Instrumentation, Measurement and Metrology