The printed circuit board(PCB)is an indispensable component of electronic products,which deter-mines the quality of these products.With the development and advancement of manufacturing technology,the layout and struct...The printed circuit board(PCB)is an indispensable component of electronic products,which deter-mines the quality of these products.With the development and advancement of manufacturing technology,the layout and structure of PCB are getting complicated.However,there are few effective and accurate PCB defect detection methods.There are high requirements for the accuracy of PCB defect detection in the actual pro-duction environment,so we propose two PCB defect detection frameworks with multiple model fusion including the defect detection by multi-model voting method(DDMV)and the defect detection by multi-model learning method(DDML).With the purpose of reducing wrong and missing detection,the DDMV and DDML integrate multiple defect detection networks with different fusion strategies.The effectiveness and accuracy of the proposed framework are verified with extensive experiments on two open-source PCB datasets.The experimental results demonstrate that the proposed DDMV and DDML are better than any other individual state-of-the-art PCB defect detection model in F1-score,and the area under curve value of DDML is also higher than that of any other individual detection model.Furthermore,compared with DDMV,the DDML with an automatic machine learning method achieves the best performance in PCB defect detection,and the Fl-score on the two datasets can reach 99.7%and 95.6%respectively.展开更多
A dynamic learning rate Gaussian mixture model（GMM）algorithm is proposed to deal with the problem of slow adaption of GMM in the case of moving object detection in the outdoor surveillance,especially in the presence...A dynamic learning rate Gaussian mixture model（GMM）algorithm is proposed to deal with the problem of slow adaption of GMM in the case of moving object detection in the outdoor surveillance,especially in the presence of sudden illumination changes.The GMM is mostly used for detecting objects in complex scenes for intelligent monitoring systems.To solve this problem,a mixture Gaussian model has been built for each pixel in the video frame,and according to the scene change from the frame difference,the learning rate of GMM can be dynamically adjusted.The experiments show that the proposed method gives good results with an adaptive GMM learning rate when we compare it with GMM method with a fixed learning rate.The method was tested on a certain dataset,and tests in the case of sudden natural light changes show that our method has a better accuracy and lower false alarm rate.展开更多
Accurately identifying the location and type of internal defects in gas-insulated switchgear(GIS)is a challenge.To address this challenge,this study proposes a novel method for the nondestructive detection of GIS inte...Accurately identifying the location and type of internal defects in gas-insulated switchgear(GIS)is a challenge.To address this challenge,this study proposes a novel method for the nondestructive detection of GIS internal defects.This method is based on x-ray digital radiography(X-DR)technology and an improved real-time models for object detection(RTMdet)algorithm,namely GIS-specific localised internal defect-RTMdet.Firstly,the X-DR images of GIS are preprocessed by dynamic limit adaptive histogram equalisation algorithm to improve the images contrast.Then,a convolution shuffle upsample module for upsampling is proposed,which enlarges the defect feature map by multi-convolution and pixel shuffling,reduces the information loss,and enhances the interaction between the feature information.Finally,both the multi-channel attention net and the global attention mechanism are integrated into the neck network for enhancing local feature extraction and global information association.Experiments demonstrate that the pro-posed method achieves a mean average precision@0.5:0.95 of 94.9%,showcasing excellent overall performance and generalisation ability,and is more suitable for accurate nondestructive detection of internal defects of GIS in complex scenarios.展开更多
The nutritional value of perishable food items,such as fruits and vegetables,depends on their freshness levels.The existing approaches solve a binary class problem by classifying a known fruit\vegetable class into fre...The nutritional value of perishable food items,such as fruits and vegetables,depends on their freshness levels.The existing approaches solve a binary class problem by classifying a known fruit\vegetable class into fresh or rotten only.We propose an automated fruits and vegetables categorization approach that first recognizes the class of object in an image and then categorizes that fruit or vegetable into one of the three categories:purefresh,medium-fresh,and rotten.We gathered a dataset comprising of 60K images of 11 fruits and vegetables,each is further divided into three categories of freshness,using hand-held cameras.The recognition and categorization of fruits and vegetables are performed through two deep learning models:Visual Geometry Group(VGG-16)and You Only Look Once(YOLO),and their results are compared.VGG-16 classifies fruits and vegetables and categorizes their freshness,while YOLO also localizes them within the image.Furthermore,we have developed an android based application that takes the image of the fruit or vegetable as input and returns its class label and its freshness degree.A comprehensive experimental evaluation of proposed approach demonstrates that the proposed approach can achieve a high accuracy and F1score on gathered FruitVeg Freshness dataset.The dataset is publicly available for further evaluation by the research community.展开更多
基金the Natural Science Foundation of Shanghai(No.20ZR1420400)the State Key Program of National Natural Science Foundation of China(No.61936001)。
文摘The printed circuit board(PCB)is an indispensable component of electronic products,which deter-mines the quality of these products.With the development and advancement of manufacturing technology,the layout and structure of PCB are getting complicated.However,there are few effective and accurate PCB defect detection methods.There are high requirements for the accuracy of PCB defect detection in the actual pro-duction environment,so we propose two PCB defect detection frameworks with multiple model fusion including the defect detection by multi-model voting method(DDMV)and the defect detection by multi-model learning method(DDML).With the purpose of reducing wrong and missing detection,the DDMV and DDML integrate multiple defect detection networks with different fusion strategies.The effectiveness and accuracy of the proposed framework are verified with extensive experiments on two open-source PCB datasets.The experimental results demonstrate that the proposed DDMV and DDML are better than any other individual state-of-the-art PCB defect detection model in F1-score,and the area under curve value of DDML is also higher than that of any other individual detection model.Furthermore,compared with DDMV,the DDML with an automatic machine learning method achieves the best performance in PCB defect detection,and the Fl-score on the two datasets can reach 99.7%and 95.6%respectively.
文摘A dynamic learning rate Gaussian mixture model（GMM）algorithm is proposed to deal with the problem of slow adaption of GMM in the case of moving object detection in the outdoor surveillance,especially in the presence of sudden illumination changes.The GMM is mostly used for detecting objects in complex scenes for intelligent monitoring systems.To solve this problem,a mixture Gaussian model has been built for each pixel in the video frame,and according to the scene change from the frame difference,the learning rate of GMM can be dynamically adjusted.The experiments show that the proposed method gives good results with an adaptive GMM learning rate when we compare it with GMM method with a fixed learning rate.The method was tested on a certain dataset,and tests in the case of sudden natural light changes show that our method has a better accuracy and lower false alarm rate.
基金National Engineering Research Center of UHV Technology and New Electrical Equipment Basis of China Southern Power Grid Research Institute Co.,Ltd,Grant/Award Number:NERCUTNEEB-2022-KF-08。
文摘Accurately identifying the location and type of internal defects in gas-insulated switchgear(GIS)is a challenge.To address this challenge,this study proposes a novel method for the nondestructive detection of GIS internal defects.This method is based on x-ray digital radiography(X-DR)technology and an improved real-time models for object detection(RTMdet)algorithm,namely GIS-specific localised internal defect-RTMdet.Firstly,the X-DR images of GIS are preprocessed by dynamic limit adaptive histogram equalisation algorithm to improve the images contrast.Then,a convolution shuffle upsample module for upsampling is proposed,which enlarges the defect feature map by multi-convolution and pixel shuffling,reduces the information loss,and enhances the interaction between the feature information.Finally,both the multi-channel attention net and the global attention mechanism are integrated into the neck network for enhancing local feature extraction and global information association.Experiments demonstrate that the pro-posed method achieves a mean average precision@0.5:0.95 of 94.9%,showcasing excellent overall performance and generalisation ability,and is more suitable for accurate nondestructive detection of internal defects of GIS in complex scenarios.
文摘The nutritional value of perishable food items,such as fruits and vegetables,depends on their freshness levels.The existing approaches solve a binary class problem by classifying a known fruit\vegetable class into fresh or rotten only.We propose an automated fruits and vegetables categorization approach that first recognizes the class of object in an image and then categorizes that fruit or vegetable into one of the three categories:purefresh,medium-fresh,and rotten.We gathered a dataset comprising of 60K images of 11 fruits and vegetables,each is further divided into three categories of freshness,using hand-held cameras.The recognition and categorization of fruits and vegetables are performed through two deep learning models:Visual Geometry Group(VGG-16)and You Only Look Once(YOLO),and their results are compared.VGG-16 classifies fruits and vegetables and categorizes their freshness,while YOLO also localizes them within the image.Furthermore,we have developed an android based application that takes the image of the fruit or vegetable as input and returns its class label and its freshness degree.A comprehensive experimental evaluation of proposed approach demonstrates that the proposed approach can achieve a high accuracy and F1score on gathered FruitVeg Freshness dataset.The dataset is publicly available for further evaluation by the research community.
