Wei Zhu

Beijing Institute Of Technology, Peping, Beijing, China

Are you Wei Zhu?

Claim your profile

Publications (20)1 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: t is a effective means by using high-speed vision to locate mobile targets. Under the circumstance of high frame rate and high sensitivity (300Hz), in addition to the Gaussian noise and impulse noise, the image quality is also influenced by atmospheric instability, and it is mainly expressed as Gaussian noise. An improved adaptive threshold weighted mean (IATWM) de-noising algorithm is proposed in this paper. According to the characteristics of impulse noise, the algorithm is able to obtain the threshold adaptively and separate the impulse noise. Then, the weighted median filtering algorithm is used to remove the impulse noise. And the improved weighted average filter algorithm is adopted to remove the Gaussian noise for graphics with Gaussian noise. The algorithm could deal with the Gaussian noise and impulse noise separately, avoiding the weaken handling for the parts not subject to pixels pollution of the impulse noise. The experimental results show that the processing result of the algorithm is able to retain the image details, superior to the traditional filtering algorithms and MTM algorithm. In addition, the algorithm provides an effective way to eliminate the mixed noise, along with a good effect on the high-speed sequence image processing.
    09/2013; 415:318-324. DOI:10.4028/www.scientific.net/AMM.415.318
  • [Show abstract] [Hide abstract]
    ABSTRACT: The panoramic annular optical lens can acquire the information of 360 degrees scene of horizontal plane and condense the image onto a CCD. In this paper a single-lens catadioptric panoramic imaging system is designed by the software ZEMAX to form a panoramic annular image on a CCD which is needed to be unwrapped for easily observation. An improved cubic spline interpolation algorithm is proposed in this paper based on the characters of the annular image for the unwrapping of the image. The new method can preserve the edge and detail of the image much better compared with the nearest interpolation algorithm and bilinear interpolation algorithm. The image already unwrapped still has distortion due to the intrinsic imaging characters of the catadioptric panoramic imaging system. Then the paper shows the evaluation for the distortion of the unwrapped image which derived from the result of image simulation using the optical system mentioned above. A new distortion method is devised when black and white fringe image is used as the input picture for the image simulation to get a distortion correction model which can be applied for the distortion correction of any other unwrapped images. Simulation results show that the panoramic annular optical system can obtain the omnidirectional information factually based on the new unwrapped algorithm and distortion correction algorithm.
    Proceedings of SPIE - The International Society for Optical Engineering 10/2012; DOI:10.1117/12.976309 · 0.20 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: A representation method for line edge roughness (LER) parameters based on wavelet-analysis is proposed. Multi-resolution analysis is processed on both spatial features and frequency features of line edge. Different feature details are achieved with different analysis resolution. The features of line edge are able to be represented well by wavelet analysis. And an extraction method for contour line of LER is designed based on wavelet baseline. The experimental results show that more details of LER features are obtained by multi-resolution wavelet analysis. The method is feasible to estimate the nano-scale LER. Therefore, it is possible to reduce the processing errors of nano-scale manufacturing line.
    02/2012; 472-475:2436-2443. DOI:10.4028/www.scientific.net/AMR.472-475.2436
  • [Show abstract] [Hide abstract]
    ABSTRACT: Using high-speed visual equipment is an effective method to locate mobile targets. Under the circumstance of high sensitivity(500Hz), except for the Gaussian noise, atmospheric instability has also an important impact on the image quality. To solve the problem, a method is proposed in this paper based on image power spectrum to analyze and evaluate the Gaussian noise, atmospheric noise meanwhile combined with wavelet denoising to remove the noise aiming at the images acquired by DALSA's 8192*32 high-sensitivity camera. Firstly, image databases are established based on the outdoor working conditions, including normal images, Gaussian noise images loaded with different simulated characteristics and atmospheric noise images in different simulated frequencies. Power spectrum ratio of all the images in the databases is calculated, and the image power spectrum critical value is determined. Then the evaluation and classification of the image noise is got according to the databases and the threshold. wavelet denoising is introduced to remove the noise subsequently. Finally, the comparison of power spectrum between the image untreated and treated is made to evaluate the effect of the method above. Experimental results show that the way can evaluate and remove the noise of image effectively for high-speed visual images.
    Proceedings of SPIE - The International Society for Optical Engineering 11/2011; DOI:10.1117/12.905013 · 0.20 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Laser image often contains mixed noise mainly including salt and pepper noise and Gaussian noise. Not only the quality of laser image is made deteriorated but also the important details of laser image are flooded by it. A new mixed noise filtering method for laser image is proposed in this paper to filter out mixed noise effectively meanwhile ensure laser image details clear and completed. In the first space, salt and pepper noise in laser image is filtered out in spatial domain. A Novel Adaptive Switching Median (NASM) filter based on local salt and pepper noise density which determines filter window of every salt and pepper noise point adaptively is introduced here. Secondly, image is transferred to wavelet domain to filter out Gaussian noise. Local Adaptive BayesShrink Threshold (LABT) wavelet denoising on basis of Gaussian Mixture Model (GMM) is used in this step where wavelet coefficients are modeled by GMM and LABT is used for adjusting the threshold adaptively which utilizes local relevant characters of sub band wavelet coefficients. Experimental results show that new method can remove mixed noise effectively meanwhile protect details of laser image well thus getting better filtering performance compared with other filters for laser image.
    Proceedings of SPIE - The International Society for Optical Engineering 11/2011; DOI:10.1117/12.904678 · 0.20 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper a new approach for obtaining the intrinsic camera parameters based on circular points is proposed. This method just requires the camera to observe a planar pattern with square at a few different unknown orientations (at least three). With the image of four corner points of square we can obtain the image of circular points (ICPs) linearly. Without conic fitting, pattern matching and metric information of square, camera intrinsic parameters can be obtained automatically. This method is particularly useful for those people who want to get camera intrinsic parameters fast with relative high precision. Simulation and real planar pattern experiment show that the new approach is convenient, accurate and robust.
    4th International Symposium on Computational Intelligence and Design, ISCID 2011, Hangzhou, China, October 28-30, 2011, 2 Volumes; 10/2011
  • [Show abstract] [Hide abstract]
    ABSTRACT: Coordinate measuring system which based on the theory of binocular stereo vision is widely used in many areas, whereas their effective measuring ranges are usually not larger than ten meters. In modern times the surveillance of large scale is used more and more in the civil and military area with the development of camera and computer technology. So based on this requirement this paper developed a new measuring model for this binocular stereo vision measuring system which is proper used in outdoor surveillance to get the 3D coordinate of the moving object. When the distance between two cameras is hundreds meters, the installation and camera calibration are quiet simple and convenient without expensive calibration apparatus and an elaborate setup or a planar pattern shown at a few different orientations or complicated camera imaging model and the parameters of math model are easy to get. After building the model of measuring system error analysis is performed to show influence of every parameter on the measuring system error. Both computer simulation and real data have been used to test the validity of our new simple measuring system model.
    International Symposium on Photoelectronic Detection and Imaging 2011; 06/2011
  • [Show abstract] [Hide abstract]
    ABSTRACT: In order to measure accurately position of object in remote large area with vision measurement method, a special layout of vision measurement system was designed. Some calibration objects which have the same height with cameras were placed at specific position in the measurement area. By adjusting carefully the posture of cameras, those calibration objects were imaged at the same pixel row which contained the camera principal point. As a result, the master planes of cameras which were determined by the pixel row which contained the principal point and the optical axis of cameras were horizontal and coplanar. So, the u pixel coordinate of any space point was same with the one of its projection point in the xoy coordinate plane and a simplified calculation model was derived. Compared with the traditional vision measurement method, this method is simpler and more suitable for large remote area measurement. In order to verify the validity of this method, a verification experiment had been made. The experimental results show that the resolving accuracy of this model is very high, and this measurement method is very good to meet the requirement of large remote area measurement.
    International Symposium on Photoelectronic Detection and Imaging 2011; 06/2011
  • Xiao Kang, KeJie Li, Wei Zhu
    [Show abstract] [Hide abstract]
    ABSTRACT: A new localization method Genetic Simulated Annealing Monte Carlo Localization (GSAMCL) is presented for mobile robots in this paper. By using the observation matching as the fitness function to make the particles adjust to the high probability area meanwhile utilizing the high optimization performance of Genetic Simulated Annealing Algorithm, GSAMCL alleviates particle recession and improves the convergence efficiency compared with Monte Carlo Localization (MCL). Implementation of a system for multiple mobile robots localization using GSAMCL is gained based on the establishment of motion model and RSSI-based awareness model of mobile robots. Through analyzing of simulation results of the mobile robots system above, it shows that, using GSAMCL, mobile robots need fewer particles and less time to achieve higher localization efficiency and obtain higher localization accuracy under the same condition in global localization compared with MCL.
    01/2011; DOI:10.1109/ICMA.2011.5986249
  • [Show abstract] [Hide abstract]
    ABSTRACT: In order to measure precisely the position of remote falling point in large-range using stereo vision measurement technology, a fast accurate installation method with calibration objects for the field vision measurement system was proposed. The first thing of this method was to calculate accurately parameters of system based on measurement conditions. Then, some special calibration objects were emplaced with the same height along the theoretical main axis of camera. According to the images of calibration objects, the pose of camera was adjusted carefully to complete the fast accurate installation of camera. Simulation and experimental results show that this method is very suitable for locating the remote falling point in large-scale field. And the measurement accuracy of this method is about 0.3%.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Laser image often mixes with salt and pepper noise when obtained and transmitted by image sensor. The salt and pepper noise not only makes the quality of laser image deteriorated but also causes the detail feature of laser image flooded which contains very important structure information. To remove salt and pepper noise effectively meanwhile ensure image details clear and completed, a Novel Adaptive Switching Median (NASM) filter based on local salt and pepper noise density is presented in this paper. Pixel points are divided into salt and pepper noise points and signal points according to two level detection mechanisms firstly, then, local salt and pepper noise density is introduced here to determined filter window of every noise point, only noise points are filtered by different size window adaptively whereas signal points are kept unprocessed finally. Through analyzing of experimental results, it shows that NASM filter can remove salt and pepper noise effectively meanwhile protect image details well thus getting better filtering performance compared with Traditional Median (TM) filter and other improved Switching Median (SM) filters for laser image.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Coal mine detect robot is a kind of mobile robot. It goes into explosion environment and detects gas content, temperature, etc. Environment dates are send to control people in safe field. Coal mine is a dangerous field, varies fatal factors menace people. Especially when coal mine disaster is occur. Rescuer doesn't know mine tunnel actual situation. It is very dangerous to go into mine tunnel without environment detect because second explosion may occur at any time. This paper introduces a coal mine robot electric and control system. It has many characters suitable to mine tunnel. It is compose of electric, computer, control, communicate, sensor, etc. The main problem of it is weight than ordinary robot, communicate in turn tunnel. It can run in explosion environment, climb ruins, check gas, and carry some food and medicine to people pin down in disaster. This robot can rescue people in coal mine disaster.
    Electronic Measurement & Instruments, 2009. ICEMI '09. 9th International Conference on; 09/2009
  • [Show abstract] [Hide abstract]
    ABSTRACT: After earthquake, many people are buried under ruins. Rescue is very important in the first 48 hours. Snake robot is a very useful tool in climbing into ruins to detect people. This paper designs a new snake robot with all body drive system based on rope system. All snake robot body has drive ropes to drive it move. It is simple in drive system and control system. Any point touched ground, wall or roof can have drive force. A force feed back system is in snake robot body. Snake robot knows all force it touching and know environment. Force feedback control is used to cross obstacle. Snake body can auto controlled to make a detour to avoid obstacle. This snake can climb into ruins to detect people and control is simple than other snake robots. It is a practical scheme in disaster rescue robot in earthquake ruins.
    Mechatronics and Automation, 2008. ICMA 2008. IEEE International Conference on; 09/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: The purpose for this research is to develop a realtime video stabilization algorithm for mine tunnel inspectional robot. The inspectional robot moves violently and randomly depending on the tunnel terrain, the video flow captured from the camera bounded on the robot is affected by jitters. These jitters mean there are errors between the captured video flow and the desired video flow. These errors affect the robot operator's visual experience. In this paper, a real-time scheme of video stabilization is proposed to overcome the rotational and translational movement of video flow. In proposed scheme, the LSE analytical solution is adopted to estimate the global motion parameters through local motion vectors, and for the purpose of getting smooth video flow, the weighting least error fitting algorithm is used to smooth the video frames. Finally, the experimental results show the effectiveness of this scheme.
    Robotics and Biomimetics, 2007. ROBIO 2007. IEEE International Conference on; 01/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: Light mobile robotpsilas weight design is a important problem which decides function and ability of mobile robot. In this paper, a total design method is advanced. There are many factors to consider, include mechanical part, electrical part, and task part. Each parts of robot must be carefully calculated and designed. There are many experiences in it. A light mobile robot is carefully analyzed as an example. The method can help mobile robot designers on how to design a robotpsilas weight quickly and not waste time and money.
    Proceedings of the International Joint Conference on Neural Networks, IJCNN 2008, part of the IEEE World Congress on Computational Intelligence, WCCI 2008, Hong Kong, China, June 1-6, 2008; 01/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: Coal mine is a dangerous field, varies fatal factors menace people. Especially after coal mine disaster, rescuer doesn't know mine tunnel situation. It is very dangerous to go into mine tunnel without environment detect because second explosion may occur. Coal mine detect and rescue robot is a kind of mobile robot. It can go into explosion environment and detect gas content, temperature, etc. The date can send to control man in safe field. This paper designs a coal mine detect and rescue robot. It has many characters suitable to mine tunnel. It is compose of mechanic, electric, computer, control, communicate, sensor, etc. It can run in explosion environment, climb ruins, check gas, and carry some food and medicine to people pin down in a disaster. This robot can rescue many people in coal mine disaster.
    Proceedings of the IEEE International Conference on Networking, Sensing and Control, ICNSC 2008, Hainan, China, 6-8 April 2008; 01/2008
  • Information Computing and Automation - The International Conference; 01/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: Space robot reliability is a fatal important problem. Any fault and error may damage spacecraft. This paper suggests a space robot system with two alternate drive system, two alternate control system, redundant freedom, two alternate communicate system. The space robot is a fault-tolerant system. Fault tree is used to analysis space robot. Any one element on the space robot fault doesnpsilat influence robot work. Space robot reliability is raised to a high level through this design. This design scheme is a practice scheme for space robot.
    Proceedings of the International Joint Conference on Neural Networks, IJCNN 2008, part of the IEEE World Congress on Computational Intelligence, WCCI 2008, Hong Kong, China, June 1-6, 2008; 01/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: A new wall climbing robot suction method called ldquoNPT methodrdquo has been primarily proposed by our lab, it is based on compositive effect of negative pressure force and thrust. This paper presents the suction principle in detail, then the key parameters are given and a model of the robot is designed. Using computational fluid dynamics simulation, key parameters are analyzed. Finally the attraction-gap-motor speed curve is drawn through experimental tests. It is proved that this method is suitable for improving the robot's adhering ability and obstacle crossing ability.
  • Jize Li, Kejie Li, Wei Zhu
    [Show abstract] [Hide abstract]
    ABSTRACT: To provide proper coverage of their random deployment regions, wireless sensor networks (WSN) should employ abundant nodes. In this paper we correct such situations by employing some mobile robots as mobile nodes in WSN which can actively move to desired locations for repairing the broken networks. According to the pre-research work we know that the number of nodes employed by WSN is closely relevant to quality of service (QoS) in sensing coverage when the nodes are randomly deployed in the target region. A modified particle swarm optimization (PSO) named particle swarm genetic optimization (PSGO), which imports selection and mutation operators in PSO to overcome the premature fault of classical PSO, is proposed to redeploy the mobile robots according to the node density for repairing the sensing coverage hole after their initial random deployment. It is suggested by the simulated experiment results that the WSN employing the mobile robots can improve the QoS in sensing coverage than the stationary WSN by redeploying mobile robots according to the node density.