1.School of Computer and Communication Engineering,Northeastern University at Qinhuangdao,Qinhuangdao 066004,China
2.Hebei Key Laboratory of Marine Perception Network and Data Processing,Northeastern University at Qinhuangdao,Qinhuangdao 066004,China. Corresponding author: WANG Huai,E-mail: 2372415@stu. neu. edu. cn
Light stripe center extraction algorithm based on an improved U-shaped network (UNet) was proposed to resolve the contradiction among uneven strong reflections on workpiece surfaces, background noise interference, and the demand for high-precision measurement in complex environments. Firstly, grouped depthwise separable convolution modules and dense residual convolution modules were introduced into the encoder and decoder of the UNet model, respectively. A U-shaped segmentation model based on multiple convolutions, feature fusion, and attention mechanisms was established to reduce interference caused by abnormal exposure in stripe images and improve the overall accuracy of stripe segmentation. Experiments on the test set show that for the stripe segmentation task, the Dice coefficient, mean intersection over union, and precision reach 0.901 8, 0.835 4, and 0.913 5, respectively, which are improved by 6.43%, 9.13%, and 15.51% compared with the original UNet model. Finally, based on the segmentation results, Steger algorithm based on the Hessian matrix is applied to accurately extract the centerline of the light stripe. Experimental results verify the effectiveness of combining the improved UNet model with Steger algorithm, which demonstrates excellent accuracy even under abnormal exposure conditions in light stripe images.
ZhangYao, XiaYu-feng, WangZi-qi, et al. Extraction method of light stripe center based on the matching of adjacent pixels of skeleton[J]. Journal of Northeastern University (Natural Science), 2023, 44(11): 1529-1536.
YangJian-hua, YangXue-rong, ChengSi-yuan, et al. Review of extracting the centers of linear structured light stripes for 3D visual measurements[J]. Journal of Guangdong University of Technology, 2014, 31(1): 74-78.
[9]
ZhangT Y, SuenC Y. A fast parallel algorithm for thinning digital patterns[J]. Communications of the ACM, 1984, 27(3): 236-239.
[10]
ZhouP, XuK, WangD D. Rail profile measurement based on line-structured light vision[J]. IEEE Access, 2018, 6:16423-16431.
MeiJun-hua, LaiLei-jie. Research on extraction algorithm of light strip center in structured light measurement system[J]. Electronic Measurement Technology, 2018, 41(21): 8-13.
[13]
ParkJ, KimC, NaJ, et al. Using structured light for efficient depth edge detection[J]. Image and Vision Computing, 2008, 26(11): 1449-1465.
HuBin, LiDe-hua, JinGang, et al. New method of obtaining the center of structured light stripe by direction template[J]. Computer Engineering and Applications, 2002(11): 59-60, 109.
[16]
StegerC. An unbiased detector of curvilinear structures[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998, 20(2): 113-125.
ZhangYu, HuangDan-ping, TianYing, et al. Research on extraction method of structured light center line of complex object[J]. Electronic Measurement Technology, 2023, 46 (1): 167-172.
[19]
YeC A, LiY H, WangC, et al. A method of binocular laser 3D scanning imaging for reflective workpieces[J]. IEEE Sensors Journal, 2023, 23(13): 15188-15198.
[20]
JiangH, FuW J, ZhangX M, et al. Extraction of laser stripe centerlines from translucent optical components using a multi-scale attention deep neural network[J]. Measurement Science and Technology, 2024, 35(8): 085404.
ZhangYong-ting, LiFeng. Center extraction of line structured light strip based on improved UNet network[J]. Computer & Digital Engineering, 2023, 51(9):2092-2097.
[23]
GuoX Z, ShiZ Y, YuB, et al. 3D measurement of gears based on a line structured light sensor[J]. Precision Engineering, 2020, 61: 160-169.
[24]
ZhangX Y, ZhouX Y, LinM X, et al. ShuffleNet: an extremely efficient convolutional neural network for mobile devices[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, 2018: 6848-6856.
[25]
LinZ C, LiS Y, JiangY Z, et al. Feedback multi-scale residual dense network for image super-resolution[J]. Signal Processing: Image Communication, 2022, 107: 116760.