Search

• An accurate method for detecting complex component surface defects is proposed. • The defect detection accuracy is high to 99.3% for the complex components. • Normal feature misdetection can be guaranteed by machine learning method. • Spreading algorithm is proposed to classify and recognize the number of defects. • Sizes of defects can be detected by the covariance matrix 3D measurement method. Detecting the surface defects of complex components exhibiting different visual properties in positions, shapes, number and sizes is a challenging problem. In this paper a novel computational framework is developed to accurately detect the component surface defects through three steps. In the framework, the positions and shapes of the components surface defects are extracted based on the support vector machine and the point cloud model. Then a novel unsupervised classification method termed as spreading algorithm is proposed to classify the defects for the recognition of the number of defects, and finally the sizes of defects are calculated using the covariance matrix 3D measurement method. Experimental cases on two typical complex components, the blade with depression and protrusion defects and the transmission case with defect and normal features, are investigated. The results are particularly compared with the ones with the state-of-the-art methods, showing the practicality and effectiveness of the proposed technique. [ABSTRACT FROM AUTHOR]

• A novel robot hand-eye calibration method is proposed without complicated error identification. • Robot hand-eye can be effectively calibrated without calculating the complex matrix equations. • Robot relocalization is implemented to reduce the introduction of robot kinematic error. • The IRLS spherical fitting algorithm is used to suppress the influence of measurement noise. • The error of the proposed method in the criterion sphere splicing verification is only 0.0323 mm. The accuracy of hand-eye calibration is easily affected by robot kinematic error, measurement error, random error, etc., and can be to a certain extent enhanced by the complicated error identification. To address this problem, we propose a novel errors-unidentified hand-eye calibration method. A relocalization-based hand-eye calibration is implemented to overcome the challenge from robot kinematics at first owing to the high robot relocalization accuracy. The tool center point (TCP) coordinates are then obtained based on an iterative reweighted least squares (IRLS) spherical fitting algorithm. An iterative combinatorial refinement algorithm is finally presented to search the solutions in the form of permutation and combination when solving the rigid transformation matrix. Experiments on criterion sphere and blade demonstrate that by converting the hand-eye matrix solving problem into a point set matching problem, the accuracy of the proposed method is enhanced by 60% on average compared with three state-of-the-art hand-eye calibration methods. [ABSTRACT FROM AUTHOR]

• An interference-resistant gear defect detection method is proposed based on improved YOLOv5. • Ability of the network to extract defect features is enhanced by new module CBAMC3. • Combined module BiFPN_concat makes the network capable of fusing the features. • Cosine annealing function is used to improve the learning ability of the network. • Improved YOLOv5 network enhances the recall rate by 13.1% and the mAP @0.5 by 12%. Automatically detecting the metallic gear defects may encounter the misdetected and undetected errors when the defect features are very close to the interferences of oil stains and image shadow. To address these challenges, in this paper an industrial interference-resistant gear surface defect detection method is presented based on the improved YOLOv5 network. The method combines the approach using the attention module CBAMC3 constituted by CBAM and C3 modules together with the analysis of the module BiFPN_concat, which makes the network capable of extracting and fusing defect features, respectively. The cosine annealing function is then employed to improve the learning ability of the network by modifying the learning rate. The experimental results indicate that the improved YOLOv5 network enhances both the recall rate by 13.1% and the mAP @0.5 by 12% compared with the original YOLOv5 network, and also outweighs the classical algorithms in detection speed with average 25 frames per second. [ABSTRACT FROM AUTHOR]

• A convergent binocular vision algorithm under extended imaging dynamic range is proposed. • The proposed method is used to guide the machining robot for precise positioning. • Re-projection error with the extended dynamic range is enhanced by nearly 1 pixel level. • The average positioning error of the robot can be controlled within 3%. • This work provides reference for subsequent vision-guided robotic machining. The machining quality of large and complex components highly depends on the positioning accuracy of the vision guided robot. In order to solve the problems of limited imaging field-of-view and dynamic range of existing parallel binocular cameras, a convergent binocular vision algorithm under extended imaging dynamic range is proposed to guide the machining robot for precise positioning. Specifically, synthesize high dynamic range images combined with sigmoid function model, and chromaticity space color correction algorithm introduced for adaptive correction of image chromaticity information. The pose between the convergent binocular camera and the robot is calibrated to correct the three-dimensional coordinates of the target workpiece CAD model. The experimental results indicate that the high dynamic images synthesized by the proposed algorithm have clear surface details and balanced colors. This work further enhances the positioning accuracy of the machining robot to the target workpiece which is placed in arbitrary pose within the field-of-view. [ABSTRACT FROM AUTHOR]