A5021196481- Industrial Vision Systems and Defect Detection
- Image and Object Detection Techniques
- Advanced X-ray and CT Imaging
- Medical Imaging Techniques and Applications
- Integrated Circuits and Semiconductor Failure Analysis
- Advanced Vision and Imaging
- Cardiac Imaging and Diagnostics
- Advanced Surface Polishing Techniques
- Image Processing Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Advanced machining processes and optimization
- Robotics and Sensor-Based Localization
- Non-Destructive Testing Techniques
- Cardiovascular Disease and Adiposity
- Video Surveillance and Tracking Methods
- Advanced Measurement and Metrology Techniques
- Educational Technology and Pedagogy
- Advanced Neural Network Applications
- Advanced Measurement and Detection Methods
- Higher Education and Teaching Methods
- Radiation Dose and Imaging
- Remote Sensing and Land Use
- Surface Roughness and Optical Measurements
- Advanced Image Processing Techniques
- Medical Image Segmentation Techniques
Yibin University
2025
Sichuan University of Science and Engineering
2025
Case Western Reserve University
2016-2025
Anhui University of Technology
2018-2025
Harvard University
2024-2025
San Francisco Conservatory of Music
2024
Xi'an University of Technology
2024
North China Electric Power University
2024
Shanghai Jiao Tong University
2007-2024
Xiamen University
2023
Recently, deep learning has developed rapidly and contributed in many fields like the classification radar sonar applications. In some special underwater acoustic signals, dataset for training may be scarce due to reason of security or other restrictions, which affects performance methods as those need a big ensure high accuracy. Furthermore, original is formats audio, makes difficult capture features, especially insufficient sample case because interference. this paper, we present novel...
This article proposes a novel solder joint recognition method based on the state-of-the-art Mask Region-convolutional neural network (R-CNN) deep learning method. Traditional classification methods, such as networks and statistical can only classify defect type, template-matching match position of object. Based R-CNN, our proposed approach classify, position, segment at same time. To train R-CNN-based detection method, transfer uses ResNet-101, which is initialized trained Microsoft COCO...
Purpose This paper aims to inspect the defects of solder joints printed circuit board in real-time production line, simple computing and high accuracy are primary consideration factors for feature extraction classification algorithm. Design/methodology/approach In this study, author presents an ensemble method joint defects. The new is based on extracting color geometry features after image acquisition using decision trees guarantee algorithm’s running executive efficiency. To improve...
We developed machine learning methods to identify fibrolipidic and fibrocalcific A-lines in intravascular optical coherence tomography (IVOCT) images using a comprehensive set of handcrafted features. incorporated features previous studies (e.g., attenuation A-line peaks). In addition, we included vascular lumen morphology three-dimensional (3-D) digital edge texture Classification were expansive datasets (∼7000 images), consisting both clinical <italic>in-vivo</italic> an...
As the adoption of new energy sources like photovoltaic and wind power increases alongside influx advanced electronic devices, there has been a significant rise in quality disturbance events (PQDs) within systems. These disturbances, including harmonics voltage dips, severely impact stability microgrids efficiency equipment. To enhance accuracy identifying disturbances microgrids, this paper introduces Multi-level Global Convolutional Neural Network combined with Simplified double-layer...
Soldering of printed circuit board (PCB)-based surface-mounted assemblies is a critical process, and to enhance the accuracy detecting their multi-targeted soldering defects, we propose an automated sample generation method that combines ControlNet Stable Diffusion Model. This can expand dataset by quickly obtaining images with high quality containing both defects normal detection targets. Meanwhile, Cascade Mask R-CNN model ConvNeXt as backbone, which performs well in dealing multi-target...
In China, Hangzhou is the first city to set up Public Bicycle System. Now, system has been largest bike sharing program in world. The software of System was developed by our team. Accurate and precise prediction public bicycle traffic flow important planning, design, operation s, etc. According highly complexity, nonlinearity uncertainty flow, a single model difficult ensure th e accuracy efficiency. To overcome lack method, this paper uses hybrid that combining clustering with support...
In order to adapt the high standard of defect detection in industrial production for electronic component soldering, and address issues unbalanced integrated circuit (IC) samples difficulty segmentation, this paper proposes a model based on Swin Transformer architecture automatic segmentation solder defects components. The approach involves feeding real collected from into Cycle-Generative Adversarial Network, which generates balanced sufficient sample dataset through adversarial generation....
In the production process of steel strips, detection surface defects is very important. However, traditional methods defect bring problems low accuracy and dependence on subjective judgment. this study, strips are detected by a classic convolutional neural network method that improved use transfer learning model. This model has advantages shorter training time, faster convergence, more accurate weight parameters. The obtained through experiments secures better results in than method, as its...
Feature selection is a key step for image registration. The success of feature has fundamental effect on matching image. Corners determine the contours characteristics target image, and number corners far smaller than pixels, thus can be good By considering algorithm speed registration accuracy registration, paper proposes an improved Harris corner detection method effective This effectively avoids clustering phenomenon occurs during process, points detected distribute more reasonably,...
Abstract Checkerboard corners are routinely used during tasks such as camera calibration, and accurate detection of them is essential. Traditional methods those based on the Harris corner detector usually affected by image artefacts noise blur. The more recently proposed deep network approach also suffers from lack training data. In this article, we make following contributions: First, propose a synthetic data generation method that simulates real imaging process, with most notable being...
Non-contrast, cardiac CT Calcium Score (CTCS) images provide a low-cost cardiovascular disease screening exam to guide therapeutics. We are extending standard Agatston score include risk assessments from features of epicardial adipose tissue, pericoronary heart size, and more, which currently extracted Coronary Angiography (CCTA) images. To aid such determinations, we developed deep-learning method synthesize Virtual (VCTA) CTCS retrospectively collected 256 patients who underwent CCTA our...