A5100408675- Video Surveillance and Tracking Methods
- Combustion and flame dynamics
- Computational Fluid Dynamics and Aerodynamics
- Face and Expression Recognition
- Advanced Vision and Imaging
- Image Enhancement Techniques
- Radiative Heat Transfer Studies
- Advanced Image Processing Techniques
- Image and Signal Denoising Methods
- Advanced Measurement and Detection Methods
- Color Science and Applications
- Infrared Target Detection Methodologies
- Robotic Path Planning Algorithms
- Face recognition and analysis
- Ultrasound Imaging and Elastography
- Wireless Communication Security Techniques
- Remote Sensing and Land Use
- Indoor and Outdoor Localization Technologies
- Industrial Vision Systems and Defect Detection
- Functional Brain Connectivity Studies
- Sparse and Compressive Sensing Techniques
- Robotics and Sensor-Based Localization
- Acoustic Wave Phenomena Research
- Numerical methods in inverse problems
- EEG and Brain-Computer Interfaces
First Affiliated Hospital of Xi'an Jiaotong University
2025
Hangzhou Dianzi University
2020-2024
Southwest University of Science and Technology
2008-2022
Shijiazhuang University
2014-2017
East China University of Technology
2017
Weifang University
2011
Beijing University of Technology
2008
This paper reports experiments on a hydrogen-fueled scramjet performed at different equivalence ratios in ground pulse combustion wind tunnel with Mach-2.5 incoming flow. In the non-chemical reaction flow before fuel was ignited, field oscillatory, and from pressure monitor, oscillation period 0.07 s amplitude 0.035 MPa. Schlieren flame self-luminescence images of combustor were obtained synchronously, flow-field structure stable concentrated shear layer. Deep learning used to subject...
Considering the mobile robot faces complex environments, this paper proposes a path planning method in three-dimensional grid environment. Combined with plane method, according to design target can surpass or climb up some obstacles and higher which cannot be leaped over directly could move downwards turn. The height position of are known advance is optimized through genetic algorithm (GA), meanwhile an improved GA proposed. Path environment breakthrough from Experimental results show...
To deal with the problems of illumination changes or pose variations and serious partial occlusion, patch based multiple instance learning (P-MIL) algorithm is proposed. The divides an object into many blocks. Then, online MIL applied on each block for obtaining strong classifier. takes account both average classification score scores all blocks detecting object. In particular, compared whole algorithm, P-MIL detects according to unoccluded patches when occlusion occurs. After object, rates...
We propose a compressive sensing based method to recognize face with pose variations. The recognition framework includes two issues: feature extraction and classification. For extraction, we present random measurement matrix compress an image from high dimensional space low space. has powerful discrimination because it can preserve most salient information of the image. Meanwhile, requires only uniform generator. Consequently, computational complexity is very low. classification, adopt...
To detect the edge of object, an efficient scheme based on Multi-Oriented Local Energy (MOLE) is presented. The MOLE constructed by Multi-oriented Gaussian second differential and its Hilbert transformation. obtained using a linear combination basis filters with arbitrary orientations. overcome affection illumination, Phase Congruency employed normalizing directional energies. At last, our method compared other common methods. experimental results reveal that this can extract more continuous...
Light field images can record multiple information about the light rays in a scene and provide views from single image, offering new data source for 3D reconstruction. However, ensuring quality of themselves is challenging, distorted image inputs may lead to poor reconstruction results. Accurate assessment pre-judge used as input reconstruction, providing reference results before work, significantly improving efficiency based on images. In this paper, we propose an Adaptive Vision...
This paper puts forward a simple algorithm to make color images lighter based on Wavelet transformation and histogram smoothing.This uses HIS space which is human visual features.Without changing hue, it can conduct smoothing in low brightness frequency domain through wavelet decomposition the lighter, at same time avoids vagueness of details increase noises.According experiment results, this effectively with complexity.
Dimensionality reduction is a crucial step for pattern recognition and data mining tasks to overcome the curse of dimensionality. Principal component analysis (PCA) traditional technique unsupervised dimensionality reduction, which often employed seek projection best represent in least-squares sense, but if original nonlinear structure, performance PCA will quickly drop. An supervised algorithm called Linear discriminant (LDA) seeks an embedding transformation, can work well with Gaussian...
To overcome illumination changes and pose variations, a pose-invariant face detection method is presented. First, an compensation based on reference white presented to the lighting variations. The obtained according component Y from YCbCr color space. Then, mixture model constructed by Cb Cr space H HSV extract faces colorful image. At last, eyes designed locate in images, which can distinguish neck arms ultimately. conducted CASIA database. experimental results have shown that our robust it...
This paper focuses on the theory of side information generation, and puts forward an improved scheme based unsupervised learning motion vector which improves system's performance.The generates optimizes through introduction rate control at encoder EM model decoder.Lastly, a DVC system is implemented scheme, its principle work presented.A MATLAB simulation has been developed for system.By comparing with typical generation methods, outstanding performance proposed justified, validity...
Background: Capacitive micromachined ultrasonic transducers (CMUTs) is a promising component of mechanical-electrical-acoustical conversion, which shows valuable applications in non-distructive testing and obstacle detection fields. In order to study the influence dimension parameters on directivity CMUTs, three-dimensional multi-physics coupling model proposed for CMUTs this paper. Methods: The combines multiple physical fields through finite element method (FEM), more efficient accurate....
Background: Visual tracking is a crucial component of computer vision systems. Objective: To deal with the problems occlusion, pose variation, and illumination in long-time tracking, we propose new kernel-based multiple instances learning tracker. Method: The tracker captures five positive bags, including occlusion bag, scale object to appearance changes an complex environment. A Gaussian kernel function used compute inner product for selecting powerful weak classifiers, which further...
The seismic signal denoising is to extract useful information from data, remove interference noise and improve the signal-noise ratio lay foundation of subsequent treatment data.The paper gives basic principles curvelet transformation puts forward method reducing in by using threshold.Finally, a practical example provide some references for relative teachers.
To deal with the problems of occlusion, pose variations and illumination changes in object tracking system, a regression model weighted multi-templates mean-shift (MS) algorithm is proposed this paper. Target templates occlusion are extracted to compose set. Then, MS applied set for obtaining candidate areas. Moreover, trained estimate Bhattacharyya coefficients between Finally, geometric center tracked areas considered as object’s position. The evaluated on several classical videos....