A5090805274- Advanced Neural Network Applications
- Industrial Vision Systems and Defect Detection
- Advanced Vision and Imaging
- Image and Object Detection Techniques
- Integrated Circuits and Semiconductor Failure Analysis
- Advanced Image and Video Retrieval Techniques
- Video Surveillance and Tracking Methods
- Optical measurement and interference techniques
- Image Retrieval and Classification Techniques
- Spinal Fractures and Fixation Techniques
- Autonomous Vehicle Technology and Safety
- Image Processing and 3D Reconstruction
- Wireless Signal Modulation Classification
- Robotics and Sensor-Based Localization
- Advanced Wireless Communication Technologies
- Handwritten Text Recognition Techniques
- Oral microbiology and periodontitis research
- Superconducting and THz Device Technology
- Intensive Care Unit Cognitive Disorders
- Terahertz technology and applications
- Advanced Steganography and Watermarking Techniques
- Advanced Technology in Applications
- Chemical Synthesis and Characterization
- Digital Media Forensic Detection
- Advanced Text Analysis Techniques
China University of Petroleum, Beijing
2025
Guangzhou Vocational College of Science and Technology
2019-2024
China National Salt Industry Corporation (China)
2024
First Affiliated Hospital of Anhui Medical University
2024
Anhui Medical University
2024
Beijing University of Chemical Technology
2023
Shandong Provincial Hospital
2023
Shandong First Medical University
2023
Affiliated Hospital of Jining Medical University
2021
Shihezi University
2021
IMPORTANCE Postoperative delirium is a common and impactful neuropsychiatric complication in patients undergoing coronary artery bypass grafting surgery. Cognitive training may enhance cognitive reserve, thereby reducing postoperative delirium. OBJECTIVE To determine whether preoperative reduces the incidence of grafting. DESIGN, SETTING, PARTICIPANTS This prospective, single-blind, randomized clinical trial was conducted at 3 university teaching hospitals southeastern China with enrollment...
The composite of ZnWO4/12WO3-0.5 wt % rGO was successfully synthesized for the first time by one-step precursor calcination equimolar W and Zn precursors, followed loading prepared hydrazine reduction. structure, composition, morphology as-synthesized were characterized various spectral analyses. Compared with reported metal-oxide-based NH3 sensors, our sensor can detect low concentrations NH3, method used herein is simple, low-cost, effective. composite-based exhibits highest response...
This study is aimed at exploring the effects of lentinan on small intestinal mucosa as well lung and liver injury in mice with gut-origin sepsis. Cecal ligation perforation (CLP) were used to construct a mouse model The randomly divided into six groups: sham operation group (sham), sepsis (CLP), ulinastatin-positive drug control (UTI), low concentration (LTN-L, 5 mg/kg), medium (LTN-M, 10 high (LTN-H, 20 mg/kg). H&E staining was detect pathological damage intestine, liver, lung. serum each...
The field of medical technology has begun to use artificial intelligence for auxiliary diagnosis. This article uses multi-target tracking automatically track and locate thyroid nodules. In the application scenario ultrasound imaging, this a tracking-by-detection method. multi-object model ByteTrack performs by associating each detection box instead only high-scoring boxes. For low-scoring boxes, their similarity trajectories is exploited recover real objects filter out background detections....
Abstract At present, fatigue driving is the third most dangerous factor affecting traffic accidents. Therefore, how to accurately and quickly identify driver’s state a common concern in world. With rapid development of machine vision its application field detection, new solution problem detection proposed. Based on this, this paper proposes technology based face recognition. By means computer image processing technology, drivers detected. The specific contents are as follows: dlib...
Abstract In recent years, the breakthrough of deep learning technology has brought a qualitative leap in field industrial detection. past, defects that are difficult to detect by traditional image methods can be detected automatic feature extraction find expression samples. silicon wafer defect detection, with help target detection algorithm, it effectively adapt different size, illumination, depth, length defects. this paper, yolov4 algorithm is applied crack problem. Yolov4 uses many...
Background:: Fetal small bowel obstruction (SBO) is a serious condition with high morbidity and mortality rates. Prenatal ultrasound an important tool for detecting SBO, but the optimal cutoff value intestinal diameter remains undefined. Objective:: This study aimed to investigate ultrasonic characteristics of fetal SBO determine enhance prenatal diagnosis. Methods:: We retrospectively analyzed postpartum data 76 cases diagnosed SBO. Receiver operating characteristic (ROC) curve analysis was...
Abstract Lane detection technology is widely used in the field of automatic driving. In this paper, Hough transform to detect vertical straight lanes interest. Firstly, edge features image are extracted by preprocessing. Then, region interest delineated. contour transformed find line segment image. Finally, end filtered and simulated least square method. Combine linear equation mark lane line.
In this paper, a literature frontier evaluation method based on latent semantic analysis was proposed in terms of collection front keywords and the surface information literature. Firstly, to build keywords, metric keyword studied for evaluating keyword. Secondly, literatures achieved according built space literatures' such as title, abstract. Finally, new evaluated each literature, relevance with impact factor journal which belonged to. The experimental results show that by outperforms...
Steganalysis is used for preventing the illegal use of steganography to ensure security network communication through detecting whether or not secret information hidden in carrier. This paper presents an approach detect quantization index modulation (QIM) G.723.1 based on analysis probability occurrence values before and after studying influence adjacent voice over internet protocol (VoIP). According change value distribution characteristics, this extracts matrix transition as feature...
Abstract Analog dial are widely used in modern production and life. In the fields of meter detection scale value recording, demand for automatic reading analog is increasing. This paper proposes an improved method fast images, which refines circle uses branch point extraction algorithm to extract intersection line express indirectly. Finally, pointer calculated based on ratio deflection angle total range.
Sparse code book multiple access (SCMA), due to its good link performance, adapt large-scale scenarios, helps cope with the surge of 5G and next generation mobile communication equipment challenge capacity improvement. The implementation SCMA system faces two major challenges: optimal codebook design efficient decoding algorithm. At present, method based on deep learning has been studied in system. In this paper, a scheme generative network introduces Transformer attention mechanism encoder...
Abstract Dimensionality reduction is an important idea in machine learning. In the learning, some high-dimensional data sets are often encountered. case of data, samples sparse and distance calculations difficult. Such problems serious faced by all learning methods. “Dimensional disaster.” addition, linear correlation between features easy to occur features, which means that redundant.
Abstract With the development of deep learning technology, pedestrian re-identity technology has been widely used in multi-target tracking and cross mirror tasks. In this paper, classical ResNet18 network is for The advantage that it can easily realize lightweight deployment. addition, labeled smooth entropy loss function migration are process training network, which accuracy map 67.8 on Market1501 data set while lightening promote engineering landing network.
Abstract Each camera has its own model and parameters. Accurate calibration of the camera’s parameters can effectively prevent distortion effects ensure image imaging. In this paper, three-dimensional space points are linked with two-dimensional points, in-plane obtained by using checkerboard method, thereby distorting restoring imaging effect.
This case report is to demonstrate that a female patient had suddenly become unconscious 14 hours after percutaneous vertebroplasty. Bedside echocardiogram showed the strong echo in right heart with small amount of pericardial effusion. CT high density distal branches both pulmonary arteries and heart. With help that, doctor made diagnosis intracardiac cement embolism very short time. The bone was removed under emergency cardiopulmonary bypass, then discharged smoothly.
Abstract Lane detection algorithm is widely used in the field of driverless. In this paper, a lane method proposed to estimate vehicle’s driving position and attitude. Firstly, frame image transmitted from monocular camera. each pre-processed. After Sobel edge detection, x-direction features single are obtained. Then, region interest demarcated, line segment found through Hough transform, ends filtered, two selected respectively for vehicle calculation According angle pose, current camera...
Abstract As deep learning shines in the field of machine vision, this article uses YOLACT instance segmentation algorithm to solve detection and pedestrians, vehicles other objects traffic scenes. In paper, backbone is changed from original resnet50 lightweight resnet18 ensure that can be deployed on Jetson-AGX-Xavier, processing speed reach 9.4 frames per second, map 81%. NVIDIA 2070 super, inference graphics card 40 second.
Deep learning target detection algorithm is maturing with the development of machine vision technology. This article applies YOLOV4 to industrial traffic scene detect and recognize cars, trucks, buses, pedestrians, two- wheelers, three-wheelers in scenes. Six types common goals. The this paper cuts network model structure, uses 12 anchors different sizes adapt size targets image, accuracy trained can reach 94.2%, tensorrt acceleration technology used deploy NVIDIA Jetson AGX Xavier embedded...
As the development of machine vision technology matures, this paper builds a set license plate recognition algorithms based on deep learning and applies them in traffic monitoring scenarios, which can accurately identify information passing vehicles at intersections, detection accuracy mAP reach more than 99%, The 94%, algorithm uses YOLOV4, detect position image where is located, CRNN, detected ROI area cut out for text recognition, deployed used Deployed with tensorrt acceleration library...
Abstract In order to solve the problem of large barrel distortion in fisheye lens, an image correction algorithm based on double longitude is proposed improve visual effect. Firstly, case unknown lens angle and non-circular image, spherical center its radius are obtained by using features, optimized reduce pole distortion; this basis, mapped surface orthogonal projection strategy model method, which transformed into transverse longitudinal coordinates sphere The a square plane horizontal...