To accurately and quickly segment the line laser stripes under interferences of strong noise and strong reflection, a lightweight weld line laser stripe segmentation network based on the DeepLabv3+ network, named WLS-Net, is proposed. To improve the segmentation speed of the network, the shallow residual network ResNet-18 is selected as the backbone network. Combining the multi-Dconv head transposed attention (MDTA) and the convolutional block attention module (CBAM), the multi-dimensional CBAM (MD-CBAM) is proposed, and the dynamic upsampling method (DySample) is chosen to replace the traditional bilinear interpolation to improve the segmentation accuracy. To address the foreground-background class imbalance in the weld line laser stripe image, the sum function of the Dice loss function (Dice Loss) and the pixel-wise cross-entropy loss function is chosen as the loss function of the model. The experimental results show that compared with the original DeepLabv3+ network, the WLS-Net achieves absolute improvements of 6.00% in IoU, 5.67% in precision, and 3.76% in F1 score on the weld line laser dataset, and the inference time of a single image is reduced by 18 ms. Compared with other semantic segmentation networks, the WLS-Net also achieves better segmentation effect and higher segmentation rate.
| [1] |
GUO J C , ZHU Z M , SUN B W . A multifunctional monocular visual sensor based on combined laser structured lights[J]. Transactions of the China Welding Institution, 2019, 40(10): 1—7, 161. (in Chinese)
|
| [2] |
GILLIES D. Close range photogrammetry[J]. The Photogrammetric Record, 2015, 30(151): 318—322.
|
| [3] |
BÁSACA—PRECIADO L C , SERGIYENKO O Y , RODRÍGUEZ—QUINONEZ J C , et al. Optical 3D laser measurement system for navigation of autonomous mobile robot[J]. Optics and Lasers in Engineering, 2014, 54: 159—169.
|
| [4] |
HUANG Y G , WANG D Q , JIANG M , et al. Laser fringe segmentation and feature points location method of weld image based on multi—task learning[J]. Chinese Journal of Lasers, 2023, 50(16): 79—89. (in Chinese)
|
| [5] |
ZHANG S K , WU Q X , LIN Z Y . Detection and segmentation of structured light stripe in weld image[J]. Acta Optica Sinica, 2021, 41(5): 88—96. (in Chinese)
|
| [6] |
CHEN B , HE S , LIU J , et al. Weld structured light image segmentation based on lightweight DeepLab v3+ network[J]. Chinese Journal of Lasers, 2023, 50(8): 49—58. (in Chinese)
|
| [7] |
YANG L , FAN J F , HUO B Y , et al. Image denoising of seam images with deep learning for laser vision seam tracking[J]. IEEE Sensors Journal, 2022, 22(6): 6098—6107.
|
| [8] |
LU J , ZHANG J , LUO J , et al. Plate additive, seam—tracking technology based on feature segmentation[J]. Optics & Laser Technology, 2024, 168:109848.
|
| [9] |
MILLETARI F , NAVAB N , AHMADI S A . V—net: fully convolutional neural networks for volumetric medical image segmentation[C]// 2016 Fourth International Conference on 3D Vision (3DV). New York: IEEE, 2016: 565—571.
|
| [10] |
CHEN L C , PAPANDREOU G , KOKKINOS I , et al. DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834—848.
|
| [11] |
ZAMIR S W , ARORA A , KHAN S , et al. Restormer: efficient transformer for high—resolution image restoration[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 5728—5739.
|
| [12] |
WOO S , PARK J , LEE J Y , et al. CBAM: convolutional block attention module[C]// Computer Vision—ECCV 2018. Cham: Springer, 2018: 3—19.
|
| [13] |
LIU W Z , LU H , FU H T , et al. Learning to upsample by learning to sample[C]// 2023 IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2023: 6004—6014.
|
Funding
National Natural Science Foundation of China(U1831123)
PDF
(6739KB)
