Skin defect inspection is one of the most significant tasks in the conventional process of aircraft inspection.This paper proposes a vision-based method of pixel-level defect detection,which is based on the Mask Scori...Skin defect inspection is one of the most significant tasks in the conventional process of aircraft inspection.This paper proposes a vision-based method of pixel-level defect detection,which is based on the Mask Scoring R-CNN.First,an attention mechanism and a feature fusion module are introduced,to improve feature representation.Second,a new classifier head—consisting of four convolutional layers and a fully connected layer—is proposed,to reduce the influence of information around the area of the defect.Third,to evaluate the proposed method,a dataset of aircraft skin defects was constructed,containing 276 images with a resolution of 960×720 pixels.Experimental results show that the proposed classifier head improves the detection and segmentation accuracy,for aircraft skin defect inspection,more effectively than the attention mechanism and feature fusion module.Compared with the Mask R-CNN and Mask Scoring R-CNN,the proposed method increased the segmentation precision by approximately 21%and 19.59%,respectively.These results demonstrate that the proposed method performs favorably against the other two methods of pixellevel aircraft skin defect detection.展开更多
As the development of machine vision technology, the color line-scan system is widely applied in the on-line inspection. Due to the non-uniform gray scale and color distortion of the image acquired by the system, the ...As the development of machine vision technology, the color line-scan system is widely applied in the on-line inspection. Due to the non-uniform gray scale and color distortion of the image acquired by the system, the image correction is needed to reduce the problem of image processing and the stability system. Based on reasons mentioned above, a method that using polynomial fitting to correct the image is presented to solve the problem in this paper. The method has been used in the automatic optical inspection of PCB, and has been proved to be effective. So this method will have a potential application to the development of the color line-scan machine vision system.展开更多
Automatic visual inspection of fabric is not only one of the potential application of machinevision but a considerable challenge in textile engineering as well.This paper mainly discusses howto inspect fabric defects ...Automatic visual inspection of fabric is not only one of the potential application of machinevision but a considerable challenge in textile engineering as well.This paper mainly discusses howto inspect fabric defects using machine vision.The introduced inspection system has a feature of:(?)Categorizing the fabric defects into 4 groups,for each group diffcrent image processing and recog-nizing methods are designed for fast and efficient inspection:2.The inspection and recognitionparameters are determined by training and self learning,these parameters vary with different kindsof fabric;3.Human inspetor’s experiences are summed up as rules to ensure the system has a s(?)lar evaluation performance of human inspector.This system can detect most of the fab(?) defects.the total recognition error is less than 5% except for the detection error of yarn irregularity,whichcould be as high as 20%.展开更多
基金National Natural Science Foundation of China(Nos.U2033201 and U1633105)。
文摘Skin defect inspection is one of the most significant tasks in the conventional process of aircraft inspection.This paper proposes a vision-based method of pixel-level defect detection,which is based on the Mask Scoring R-CNN.First,an attention mechanism and a feature fusion module are introduced,to improve feature representation.Second,a new classifier head—consisting of four convolutional layers and a fully connected layer—is proposed,to reduce the influence of information around the area of the defect.Third,to evaluate the proposed method,a dataset of aircraft skin defects was constructed,containing 276 images with a resolution of 960×720 pixels.Experimental results show that the proposed classifier head improves the detection and segmentation accuracy,for aircraft skin defect inspection,more effectively than the attention mechanism and feature fusion module.Compared with the Mask R-CNN and Mask Scoring R-CNN,the proposed method increased the segmentation precision by approximately 21%and 19.59%,respectively.These results demonstrate that the proposed method performs favorably against the other two methods of pixellevel aircraft skin defect detection.
文摘As the development of machine vision technology, the color line-scan system is widely applied in the on-line inspection. Due to the non-uniform gray scale and color distortion of the image acquired by the system, the image correction is needed to reduce the problem of image processing and the stability system. Based on reasons mentioned above, a method that using polynomial fitting to correct the image is presented to solve the problem in this paper. The method has been used in the automatic optical inspection of PCB, and has been proved to be effective. So this method will have a potential application to the development of the color line-scan machine vision system.
文摘Automatic visual inspection of fabric is not only one of the potential application of machinevision but a considerable challenge in textile engineering as well.This paper mainly discusses howto inspect fabric defects using machine vision.The introduced inspection system has a feature of:(?)Categorizing the fabric defects into 4 groups,for each group diffcrent image processing and recog-nizing methods are designed for fast and efficient inspection:2.The inspection and recognitionparameters are determined by training and self learning,these parameters vary with different kindsof fabric;3.Human inspetor’s experiences are summed up as rules to ensure the system has a s(?)lar evaluation performance of human inspector.This system can detect most of the fab(?) defects.the total recognition error is less than 5% except for the detection error of yarn irregularity,whichcould be as high as 20%.
