Guihua Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Guihua Liu has a total of 42 co-authors. Here are the top co-authors by number of shared publications:
Lei Deng
College of Information Engineering , Southwest University of Science and Technology , Mianyang China
5 shared publicationsTao Song
College of Information Engineering , Southwest University of Science and Technology , Mianyang China
5 shared publications- Huiming Huang
Southwest University of Science and Technology
4 shared publications Fuping Qin
College of Information Engineering , Southwest University of Science and Technology , Mianyang China
3 shared publications- Hao Deng
Mianyang Teachers’ College
3 shared publications Huiming Huang
2 shared publications
Latest Publications
- 15 September 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (37) , 375003
A binocular calibration method based on a coded 1D target for large-field measurement
To address the practical difficulty of deploying large planar targets in large-field binocular calibration and the insufficient accuracy and robustness of existing calibration methods based on one-dimensional (1D) targets, a binocular calibration method based on a coded 1D target is proposed. First, an effective matched observation set is constructed through coded point centre extraction, cross-view matching, and epipolar distance screening, and is then divided into an accuracy evaluation observation set and a candidate calibration observation set. Then, during each single calibration process, a calibration subset is randomly sampled from the candidate calibration observation set, and the initial binocular parameters are obtained through fundamental matrix estimation and refinement, initialisation based on the physical parameters of the cameras, essential matrix decomposition, and rod length constraints. The optimisation variables are further released stage by stage during optimisation, enabling the optimisation of intrinsic, extrinsic, and distortion parameters while suppressing interference from parameter coupling. Finally, multiple single calibration results are evaluated in a unified manner on the fixed accuracy evaluation observation set, and the final calibration parameters are selected according to the minimum error criterion. Experimental results show that the mean reprojection error and mean absolute error of the proposed method are 0.0277 px and 0.0422 mm, respectively, both of which are lower than those of the comparative methods. The proposed method also maintains stable calibration performance under different object depths and different binocular baseline lengths. When a standard sphere device is reconstructed, the maximum error of the sphere centre distance is 0.117 mm. The proposed method avoids dependence on a large planar target while ensuring calibration accuracy and robustness, thereby providing a technical solution for the calibration of visual measurement and three-dimensional reconstruction systems.
- 15 September 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (37) , 375003
A binocular calibration method based on a coded 1D target for large-field measurement
To address the practical difficulty of deploying large planar targets in large-field binocular calibration and the insufficient accuracy and robustness of existing calibration methods based on one-dimensional (1D) targets, a binocular calibration method based on a coded 1D target is proposed. First, an effective matched observation set is constructed through coded point centre extraction, cross-view matching, and epipolar distance screening, and is then divided into an accuracy evaluation observation set and a candidate calibration observation set. Then, during each single calibration process, a calibration subset is randomly sampled from the candidate calibration observation set, and the initial binocular parameters are obtained through fundamental matrix estimation and refinement, initialisation based on the physical parameters of the cameras, essential matrix decomposition, and rod length constraints. The optimisation variables are further released stage by stage during optimisation, enabling the optimisation of intrinsic, extrinsic, and distortion parameters while suppressing interference from parameter coupling. Finally, multiple single calibration results are evaluated in a unified manner on the fixed accuracy evaluation observation set, and the final calibration parameters are selected according to the minimum error criterion. Experimental results show that the mean reprojection error and mean absolute error of the proposed method are 0.0277 px and 0.0422 mm, respectively, both of which are lower than those of the comparative methods. The proposed method also maintains stable calibration performance under different object depths and different binocular baseline lengths. When a standard sphere device is reconstructed, the maximum error of the sphere centre distance is 0.117 mm. The proposed method avoids dependence on a large planar target while ensuring calibration accuracy and robustness, thereby providing a technical solution for the calibration of visual measurement and three-dimensional reconstruction systems.
- 21 November 2025
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (11) , 115207
Multi-line laser 3D reconstruction without light-plane calibration via trinocular spatial consistency and cross reprojection constraints
To address the problem that conventional multi-line laser three-dimensional (3D) reconstruction methods depend on a calibrated laser plane to match and reconstruct laser stripe centre point (LSCP), are susceptible to calibration errors and laser plane deformation, and experience declining reconstruction accuracy as the number of laser stripes increases, a multi-line laser reconstruction method that does not require laser plane calibration is proposed, based on trinocular spatial consistency (SC) and cross reprojection constraints. The method first obtains a coarse set of LSCP triplets through trinocular SC and then the correct matching triplet is selected using cross reprojection constraints. Once accurate trinocular correspondences are established, initial 3D coordinates of each LSCP are computed by stereo triangulation between the left and right cameras, after which a nonlinear optimization model is constructed from the trinocular reprojection error and iteratively solved to improve 3D point accuracy. Rigid transformation matrices determined from preset circular markers are finally used to register the local point clouds to the global coordinate system, completing the reconstruction of the measured object. Experiments show that the proposed method attains the LSCP matching accuracy (LMA) of 99.01% and the LSCP matching recall LMR of 96.10%. When reconstructing a standard sphere apparatus, the maximum absolute deviations in sphere diameter and centre-to-centre distance are 0.111 mm and 0.166 mm, respectively. On an 11-line system, the algorithm averages 57.327 ms per frame. In deployment, the trade-off between accuracy and speed can be tuned to meet application requirements. Robustness tests further show that the method is compatible with hardware configurations featuring different baseline lengths, camera convergence angles and lens focal lengths. The method reduces the dependence of multi-line laser reconstruction systems on a laser plane and provides a technical solution for multi-line laser measurement.
- 21 November 2025
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (11) , 115207
Multi-line laser 3D reconstruction without light-plane calibration via trinocular spatial consistency and cross reprojection constraints
To address the problem that conventional multi-line laser three-dimensional (3D) reconstruction methods depend on a calibrated laser plane to match and reconstruct laser stripe centre point (LSCP), are susceptible to calibration errors and laser plane deformation, and experience declining reconstruction accuracy as the number of laser stripes increases, a multi-line laser reconstruction method that does not require laser plane calibration is proposed, based on trinocular spatial consistency (SC) and cross reprojection constraints. The method first obtains a coarse set of LSCP triplets through trinocular SC and then the correct matching triplet is selected using cross reprojection constraints. Once accurate trinocular correspondences are established, initial 3D coordinates of each LSCP are computed by stereo triangulation between the left and right cameras, after which a nonlinear optimization model is constructed from the trinocular reprojection error and iteratively solved to improve 3D point accuracy. Rigid transformation matrices determined from preset circular markers are finally used to register the local point clouds to the global coordinate system, completing the reconstruction of the measured object. Experiments show that the proposed method attains the LSCP matching accuracy (LMA) of 99.01% and the LSCP matching recall LMR of 96.10%. When reconstructing a standard sphere apparatus, the maximum absolute deviations in sphere diameter and centre-to-centre distance are 0.111 mm and 0.166 mm, respectively. On an 11-line system, the algorithm averages 57.327 ms per frame. In deployment, the trade-off between accuracy and speed can be tuned to meet application requirements. Robustness tests further show that the method is compatible with hardware configurations featuring different baseline lengths, camera convergence angles and lens focal lengths. The method reduces the dependence of multi-line laser reconstruction systems on a laser plane and provides a technical solution for multi-line laser measurement.
- Open Access
- 17 September 2025
- Published by Walter de Gruyter GmbH in Measurement Science Review
- Vol. 25 (5) , 229-247
The Joint Calibration Method of Multi-line Laser and Tracking System based on Conjugate Gradient Iteration
The multi-line laser 3D reconstruction system mainly relies on marking points to acquire 3D data. To simplify the acquisition of 3D data for objects, we use a binocular tracking method to achieve unmarked point stitching of the multi-line laser reconstructed 3D data. The key challenge with this system is the joint calibration between the multi-line laser system and the tracking ball cage. Traditionally, planar calibration plates are used for calibration. However, due to the extensive calibration field, the production of large calibration plates incurs high costs and compromises machining accuracy. As a result, significant joint calibration errors occur between the tracking ball and the multi-line laser system, making high-precision calibration impossible. To solve these problems, an iterative method based on multi-position attitude and conjugate gradient is proposed to achieve high-precision joint calibration. A simple and convenient cross pole with multiple coding points is used as a calibrator. The 3D data of these coding points are determined beforehand using a coordinate measuring machine (CMM). First, the internal and external parameters of the binocular tracking system are calibrated using this cross pole. During the joint calibration process, in which both the multi-line laser system and the tracking ball cage are involved, the cross pole is imaged at different positions simultaneously with the binocular tracking system and the multi-line laser system. This allows us to determine the positions and orientations of both systems relative to each other and relative to the cross pole. The transformation relationship between the multi-line laser system and the tracking ball cage is calibrated using an iterative conjugate gradient optimization algorithm based on these positions and orientations, which completes the entire system calibration and eventually achieves three-dimensional reconstruction of the unmarked points. Compared to conventional planar calibration plate-based methods, our proposed approach requires only one cross pole to perform two crucial calibration steps, improving the joint calibration accuracy. While the final reconstruction accuracy of conventional methods is about 0.1 mm, our proposed method can achieve an accuracy of about 0.02 mm.
- 17 September 2025
- Published by Walter de Gruyter GmbH in Measurement Science Review
- Vol. 25 (5) , 229-247
Open AccessThe Joint Calibration Method of Multi-line Laser and Tracking System based on Conjugate Gradient Iteration
The multi-line laser 3D reconstruction system mainly relies on marking points to acquire 3D data. To simplify the acquisition of 3D data for objects, we use a binocular tracking method to achieve unmarked point stitching of the multi-line laser reconstructed 3D data. The key challenge with this system is the joint calibration between the multi-line laser system and the tracking ball cage. Traditionally, planar calibration plates are used for calibration. However, due to the extensive calibration field, the production of large calibration plates incurs high costs and compromises machining accuracy. As a result, significant joint calibration errors occur between the tracking ball and the multi-line laser system, making high-precision calibration impossible. To solve these problems, an iterative method based on multi-position attitude and conjugate gradient is proposed to achieve high-precision joint calibration. A simple and convenient cross pole with multiple coding points is used as a calibrator. The 3D data of these coding points are determined beforehand using a coordinate measuring machine (CMM). First, the internal and external parameters of the binocular tracking system are calibrated using this cross pole. During the joint calibration process, in which both the multi-line laser system and the tracking ball cage are involved, the cross pole is imaged at different positions simultaneously with the binocular tracking system and the multi-line laser system. This allows us to determine the positions and orientations of both systems relative to each other and relative to the cross pole. The transformation relationship between the multi-line laser system and the tracking ball cage is calibrated using an iterative conjugate gradient optimization algorithm based on these positions and orientations, which completes the entire system calibration and eventually achieves three-dimensional reconstruction of the unmarked points. Compared to conventional planar calibration plate-based methods, our proposed approach requires only one cross pole to perform two crucial calibration steps, improving the joint calibration accuracy. While the final reconstruction accuracy of conventional methods is about 0.1 mm, our proposed method can achieve an accuracy of about 0.02 mm.
- 25 October 2024
- Published by Elsevier in Optics and Lasers in Engineering
- Vol. 184
Circular marker-aided multi-view laser point cloud registration based on adaptive-weighted bundle adjustment
- 25 October 2024
- Published by Elsevier in Optics and Lasers in Engineering
- Vol. 184
Circular marker-aided multi-view laser point cloud registration based on adaptive-weighted bundle adjustment
- Open Access
- 9 October 2024
- Published by Springer Nature in Scientific Reports
- Vol. 14 (1) , 1-23
Multi-line laser 3D reconstruction method based on spatial quadric surface and geometric estimation
The traditional multi-line laser 3D reconstruction is based on binocular epipolar constraints and the equation of the laser optical plane. Firstly, matching points are identified based on the epipolar constraints. Then, the correct matching points are selected based on the optical plane equation of the multi-line laser. Finally, 3D reconstruction is performed using the selected matching points. In actual production processes involving multi-line lasers, noticeable distortion occurs due to the projection of Diffractive Optical Elements (DOE) at a large angle. This results in curved laser light planes and curved reflections on the measured objects. Under such circumstances, efficiently screening matching points using the traditional method based on the laser plane equation becomes challenging. Additionally, inherent noise in multi-line laser systems introduces errors in extracted laser center coordinates, making it impossible to directly obtain high-precision 3D data. To address these issues, this paper proposes a method that utilizes spatial quadric surfaces and geometric estimation for completing multi-line laser 3D reconstructions. By analyzing the distortion principle of DOE and the positional relationship after optical diffraction, the multi-line laser manifests as a quadric surface on the optical output plane. Consequently, by calibrating the equations of the quadric surface and applying binocular polar constraints, suitable matching points for the multi-line laser can be chosen. Once an accurate matching point is identified, a minimum geometric distance estimation can be established based on the distance constraint between the point and its corresponding epipolar line. This distance represents the separation between the left and right camera’s laser center points and their respective epipolar lines. Utilizing this estimate of the minimum geometric distance, a refined calculation can be performed to better satisfy epipolar constraints and obtain new matching points. Ultimately, utilizing these new matching points enables the completion of 3D reconstruction for multi-line lasers. In comparison with methods relying on spatial plane and epipolar constraints, our algorithm exhibits a superior degree of matchability and accuracy.According to multiple groups of test data, the average error before optimization is 0.3126 mm, and can be increased to 0.0336 mm after optimization.
- 9 October 2024
- Published by Springer Nature in Scientific Reports
- Vol. 14 (1) , 1-23
Open AccessMulti-line laser 3D reconstruction method based on spatial quadric surface and geometric estimation
The traditional multi-line laser 3D reconstruction is based on binocular epipolar constraints and the equation of the laser optical plane. Firstly, matching points are identified based on the epipolar constraints. Then, the correct matching points are selected based on the optical plane equation of the multi-line laser. Finally, 3D reconstruction is performed using the selected matching points. In actual production processes involving multi-line lasers, noticeable distortion occurs due to the projection of Diffractive Optical Elements (DOE) at a large angle. This results in curved laser light planes and curved reflections on the measured objects. Under such circumstances, efficiently screening matching points using the traditional method based on the laser plane equation becomes challenging. Additionally, inherent noise in multi-line laser systems introduces errors in extracted laser center coordinates, making it impossible to directly obtain high-precision 3D data. To address these issues, this paper proposes a method that utilizes spatial quadric surfaces and geometric estimation for completing multi-line laser 3D reconstructions. By analyzing the distortion principle of DOE and the positional relationship after optical diffraction, the multi-line laser manifests as a quadric surface on the optical output plane. Consequently, by calibrating the equations of the quadric surface and applying binocular polar constraints, suitable matching points for the multi-line laser can be chosen. Once an accurate matching point is identified, a minimum geometric distance estimation can be established based on the distance constraint between the point and its corresponding epipolar line. This distance represents the separation between the left and right camera’s laser center points and their respective epipolar lines. Utilizing this estimate of the minimum geometric distance, a refined calculation can be performed to better satisfy epipolar constraints and obtain new matching points. Ultimately, utilizing these new matching points enables the completion of 3D reconstruction for multi-line lasers. In comparison with methods relying on spatial plane and epipolar constraints, our algorithm exhibits a superior degree of matchability and accuracy.According to multiple groups of test data, the average error before optimization is 0.3126 mm, and can be increased to 0.0336 mm after optimization.
- 8 August 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (11) , 116004
FR-YOLOv7: feature enhanced YOLOv7 for rotated small object detection in aerial images
For the problems of weak object feature expression and variable object orientation in aerial image object detection, this paper proposed a feature enhanced YOLOv7 for rotated small object detection in aerial images. Firstly, for the problem of feature loss in the feature extraction stage, the feature enhanced spatial pyramid pooling and cross stage partial connections module was proposed, which effectively boost the feature expression of small object. Secondly, an attention guided max-pooling module was constructed to address the problem of feature loss. Then, the rotated object detection head was introduced to solve the problem of false negatives caused by variable object angles and dense object distribution. Finally, a multi-scale loss function was proposed for improving the detection effects of rotated small objects detection in aerial images. Extensive experiments were conduct on the public datasets of DOTA and University of the Chinese Academy of Sciences-AOD, with the help of the proposed method, we can achieve the detection accuracy that the mean average precision are 79.7% and 98.9%, respectively. Experimental results demonstrate that the proposed method has a significant improvement on the detection of small targets in aerial images.
- 8 August 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (11) , 116004
FR-YOLOv7: feature enhanced YOLOv7 for rotated small object detection in aerial images
For the problems of weak object feature expression and variable object orientation in aerial image object detection, this paper proposed a feature enhanced YOLOv7 for rotated small object detection in aerial images. Firstly, for the problem of feature loss in the feature extraction stage, the feature enhanced spatial pyramid pooling and cross stage partial connections module was proposed, which effectively boost the feature expression of small object. Secondly, an attention guided max-pooling module was constructed to address the problem of feature loss. Then, the rotated object detection head was introduced to solve the problem of false negatives caused by variable object angles and dense object distribution. Finally, a multi-scale loss function was proposed for improving the detection effects of rotated small objects detection in aerial images. Extensive experiments were conduct on the public datasets of DOTA and University of the Chinese Academy of Sciences-AOD, with the help of the proposed method, we can achieve the detection accuracy that the mean average precision are 79.7% and 98.9%, respectively. Experimental results demonstrate that the proposed method has a significant improvement on the detection of small targets in aerial images.
- Open Access
- 23 April 2024
- Published by Springer Nature in Research Square
The joint calibration method of binocular tracking multi-line laser system based on Multi-position attitude optimization algorithm
The traditional multi-line laser 3D reconstruction system relies on marker points to obtain complete 3D information of objects. However, in practical applications, many scenarios do not allow marker points to be pasted on objects. To solve this problem, we adopt a binocular tracking-based method to achieve marker-free stitching of multi-line laser reconstructed 3D data. The calibration of the binocular tracking system and the joint calibration between the multi-line laser system and the tracking ball cage constitute the core challenges of the system. To quickly and accurately complete the calibration of the entire system, this paper proposes a bundle adjustment optimization method based on a cross pole. This method utilizes multiple encoded marker points on the cross pole and their real 3D data. By capturing the cross pole images at different positions through the binocular tracking system, the internal and external parameters of the binocular tracking system are calibrated using the bundle adjustment optimization method in combination with the 3D data of the cross pole and the recognized image points. During the joint calibration of the multi-line laser system and the tracking ball cage, we simultaneously use the binocular tracking system and the multi-line laser system to capture cross pole images. By capturing cross pole at different positions, we can obtain the positional relationship between the cross pole, the binocular tracking system, and the multi-line laser system. Based on these positional relationships, we adopt a multi-position attitude optimization algorithm to calibrate the transformation relationship between the multi-line laser system and the tracking ball cage, thus completing the calibration of the entire system. Finally, based on all the calibration results, we can achieve 3D reconstruction of objects. Compared with the traditional method based on a planar calibration board, the calibration method proposed in this paper only requires one cross pole to complete the calibration of the two key steps, which not only improves the calibration accuracy but also simplifies the calibration process.
- 23 April 2024
- Published by Springer Nature in Research Square
Open AccessThe joint calibration method of binocular tracking multi-line laser system based on Multi-position attitude optimization algorithm
The traditional multi-line laser 3D reconstruction system relies on marker points to obtain complete 3D information of objects. However, in practical applications, many scenarios do not allow marker points to be pasted on objects. To solve this problem, we adopt a binocular tracking-based method to achieve marker-free stitching of multi-line laser reconstructed 3D data. The calibration of the binocular tracking system and the joint calibration between the multi-line laser system and the tracking ball cage constitute the core challenges of the system. To quickly and accurately complete the calibration of the entire system, this paper proposes a bundle adjustment optimization method based on a cross pole. This method utilizes multiple encoded marker points on the cross pole and their real 3D data. By capturing the cross pole images at different positions through the binocular tracking system, the internal and external parameters of the binocular tracking system are calibrated using the bundle adjustment optimization method in combination with the 3D data of the cross pole and the recognized image points. During the joint calibration of the multi-line laser system and the tracking ball cage, we simultaneously use the binocular tracking system and the multi-line laser system to capture cross pole images. By capturing cross pole at different positions, we can obtain the positional relationship between the cross pole, the binocular tracking system, and the multi-line laser system. Based on these positional relationships, we adopt a multi-position attitude optimization algorithm to calibrate the transformation relationship between the multi-line laser system and the tracking ball cage, thus completing the calibration of the entire system. Finally, based on all the calibration results, we can achieve 3D reconstruction of objects. Compared with the traditional method based on a planar calibration board, the calibration method proposed in this paper only requires one cross pole to complete the calibration of the two key steps, which not only improves the calibration accuracy but also simplifies the calibration process.
- Open Access
- 1 March 2024
- Published by AIP Publishing in AIP Advances
- Vol. 14 (3)
Spatial quadric calibration method for multi-line laser based on diffractive optical element
The traditional multi-line laser calibration method relies primarily on calibrating the spatial plane equations of multiple laser lines and reconstructing the multi-line laser in three dimensions through the plane equation and camera parameters. In traditional methods, the multi-line laser is projected as a straight line by default, but in reality, the laser line projected by the multi-line laser is not a plane due to the Diffractive Optical Element (DOE). Instead, it is a surface with a certain curvature in space. During the production process, the multi-line laser experiences significant distortion because of the DOE projecting at a large angle. Consequently, if traditional methods are used, accurate calibration of the multi-line laser cannot be achieved. To address this issue, this paper introduces a new calibration algorithm for multi-line laser based on a spatial quadric equation. By calibrating the spatial quadric equation of the multi-line laser, the spatial characteristics of the laser can be accurately captured. The algorithm involves projecting a crossed 7-line laser onto a calibration plate and changing the position of the calibration plate while capturing images. This allows for obtaining the spatial three-dimensional coordinates of each laser line in the camera coordinate system. Next, spatial quadric equations are fitted for each laser line based on the three-dimensional coordinates. Finally, 3D reconstruction is performed using the quadratic surface parameters of the multi-line laser and camera parameters. The results from 3D reconstruction demonstrate that, compared to traditional laser calibration methods based on planes, the proposed algorithm achieves higher calibration accuracy. It also improves the number of successful binocular multi-line laser matches and overall accuracy.
- 1 March 2024
- Published by AIP Publishing in AIP Advances
- Vol. 14 (3)
Open AccessSpatial quadric calibration method for multi-line laser based on diffractive optical element
The traditional multi-line laser calibration method relies primarily on calibrating the spatial plane equations of multiple laser lines and reconstructing the multi-line laser in three dimensions through the plane equation and camera parameters. In traditional methods, the multi-line laser is projected as a straight line by default, but in reality, the laser line projected by the multi-line laser is not a plane due to the Diffractive Optical Element (DOE). Instead, it is a surface with a certain curvature in space. During the production process, the multi-line laser experiences significant distortion because of the DOE projecting at a large angle. Consequently, if traditional methods are used, accurate calibration of the multi-line laser cannot be achieved. To address this issue, this paper introduces a new calibration algorithm for multi-line laser based on a spatial quadric equation. By calibrating the spatial quadric equation of the multi-line laser, the spatial characteristics of the laser can be accurately captured. The algorithm involves projecting a crossed 7-line laser onto a calibration plate and changing the position of the calibration plate while capturing images. This allows for obtaining the spatial three-dimensional coordinates of each laser line in the camera coordinate system. Next, spatial quadric equations are fitted for each laser line based on the three-dimensional coordinates. Finally, 3D reconstruction is performed using the quadratic surface parameters of the multi-line laser and camera parameters. The results from 3D reconstruction demonstrate that, compared to traditional laser calibration methods based on planes, the proposed algorithm achieves higher calibration accuracy. It also improves the number of successful binocular multi-line laser matches and overall accuracy.
- Open Access
- 20 December 2023
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (3) , 035019
Slender defect segmentation network of workpiece surface based on deep learning
As a key part of the workpiece production process, workpiece surface defect segmentation detection plays an important role in the quality assurance of the workpiece. Aiming at the problem of low accuracy of workpiece surface slender defect segmentation, we propose a slender defect segmentation network based on feature enhancement module (FEM) and feature fusion module (FFM). Firstly, the ConvNeXt is improved and used as the backbone network. Then, the FEM is designed to obtain rich defect features. Finally, the FFM is designed to expand the receptive field and fuse shallow features with deep features to enhance the features of slender defects. The proposed method improves the segmentation accuracy of slender defects. The mean pixel accuracy) of slender defect segmentation on the self-built diode glass shell dataset and public magnetic sheet dataset is 97.87% and 94.86%, respectively. The mean Intersection over Union of slender defect segmentation is 94.32% and 83.89%, respectively. The segmentation speed reaches 43FPS, and the parameter is 7.12 M, which balances the segmentation accuracy, speed and model size.
- 20 December 2023
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (3) , 035019
Open AccessSlender defect segmentation network of workpiece surface based on deep learning
As a key part of the workpiece production process, workpiece surface defect segmentation detection plays an important role in the quality assurance of the workpiece. Aiming at the problem of low accuracy of workpiece surface slender defect segmentation, we propose a slender defect segmentation network based on feature enhancement module (FEM) and feature fusion module (FFM). Firstly, the ConvNeXt is improved and used as the backbone network. Then, the FEM is designed to obtain rich defect features. Finally, the FFM is designed to expand the receptive field and fuse shallow features with deep features to enhance the features of slender defects. The proposed method improves the segmentation accuracy of slender defects. The mean pixel accuracy) of slender defect segmentation on the self-built diode glass shell dataset and public magnetic sheet dataset is 97.87% and 94.86%, respectively. The mean Intersection over Union of slender defect segmentation is 94.32% and 83.89%, respectively. The segmentation speed reaches 43FPS, and the parameter is 7.12 M, which balances the segmentation accuracy, speed and model size.
- 31 July 2023
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 34 (11) , 115009
Dimensional measurement of anti-stress cone of HV cable joints based on 3D point cloud
To solve the problem of automatic measurement of anti-stress cone of high voltage cable joints, a measurement algorithm based on 3D point cloud is proposed. Firstly, denoising and coordinate adjustment of point cloud of a cable joint are realized. Then the characteristic curve model of fitting circle radius indicating structure characteristic of cable joints is constructed by fitting circles with random sample consensus algorithm. After that, the discrete points on the characteristic curve are extracted according to the characteristics such as the fluctuation range of the characteristic curve for straight line fitting and the intersection of the fitted lines is calculated to complete the segmentation of the anti-stress cone area of the cable joint. Finally, the parameter measurement of the anti-stress cone was completed according to the structure of the anti-stress cone, and the error correction post-processing was performed. Two types of cable joint point clouds with standard dimensions and surface defects were used for measurement experiments, the absolute errors of the measurement results are less than 0.2 mm and 1.0 mm respectively, and the relative errors are less than 0.25% and 2% respectively, indicating that the proposed method has high robustness and measurement accuracy, which meet the measurement requirements in engineering practice.
- 31 July 2023
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 34 (11) , 115009
Dimensional measurement of anti-stress cone of HV cable joints based on 3D point cloud
To solve the problem of automatic measurement of anti-stress cone of high voltage cable joints, a measurement algorithm based on 3D point cloud is proposed. Firstly, denoising and coordinate adjustment of point cloud of a cable joint are realized. Then the characteristic curve model of fitting circle radius indicating structure characteristic of cable joints is constructed by fitting circles with random sample consensus algorithm. After that, the discrete points on the characteristic curve are extracted according to the characteristics such as the fluctuation range of the characteristic curve for straight line fitting and the intersection of the fitted lines is calculated to complete the segmentation of the anti-stress cone area of the cable joint. Finally, the parameter measurement of the anti-stress cone was completed according to the structure of the anti-stress cone, and the error correction post-processing was performed. Two types of cable joint point clouds with standard dimensions and surface defects were used for measurement experiments, the absolute errors of the measurement results are less than 0.2 mm and 1.0 mm respectively, and the relative errors are less than 0.25% and 2% respectively, indicating that the proposed method has high robustness and measurement accuracy, which meet the measurement requirements in engineering practice.