A laser module was packaged by a pulse width regulated circuit, a driving circuit,a laser protection circuit and laser chips to use in a pulse semiconductor laser sending system. When a laser pulse width was around several nanoseconds, the pins of packaged lasers would introduce resistances to distort the waveform shape and to reduce the energy coupling into the laser chips. To get a high peak power laser pulse with short pulse width and fast risetime, the structure of the laser chips was improved, and an optoelectronic hybrid integration was used to package the driver and the laser chips together into a module. In this way, the short current pulses could be coupled into laser chips effectively.Analysis and experiments prove that the improved laser module can produce optical pulses with better output parameters. On the same condition, the peak optical power produced by the proposed laser module is 6 times larger than that produced by traditional packaged lasers with the same driving circuit under a pulse width around 4.5 ns.The U-P curve of the laser module was tested at a pulse width around 7 ns,and the result shows that it can obtain a peak optical power of 176 W.

A refractive/diffractive spherocylindrical Phakic Intraocular Lens (PIOL) is designed to improve the white light visual quality of the human eye. On the basis of measured individual corneal data, eyeball depths and wavefront aberrations, an individual eye model is constructed with the optical design software Zemax,then the spherocylindrical PIOL and the refractive/diffractive spherocylindrical PIOL are designed by the established eye model. A comparison with spherocylindrical PIOL shows that the refractive/diffractive spherocylindrical PIOL can optimize overall white light Modulation Transfer Functions (MTF) and the MTFs of non-center wavelengths to improve the visual quality. The experiment also shows that a zone of 4 mm is enough for the binary structure. Finally, the fabrication issues are discussed. Obtained results indicate that the designed refractive/diffractive spherocylindrical PIOL has great potential ities in advanced vision correction.

In consideration of corneal aspheric surfaces and from the point of view of an optical imaging,a way combining of a corneal bi-conic model with human eye models was proposed with the optical design software Zemax (Zemax Development Corp. San Diego,Calif.). On analysis of measured results of Chinese emmetropic cornea and the collected data of Chinese myopic eyes,the initial model eye was optimized based on choosing the wavefront aberration of a human eye as the merit function. The corneal models of the emmetropic and myopic eyes and corrected eye models were presented and the relevant corneal ablation model was also given. The results show that the maximum ablation depth located at the center is about 26.5 μm.It is concluded that the corneal bi-conic model combing with the human eye models is accord with corneal surface characters,which can be applied in clinical analysis and can also provide a useful digital model of calculating wavefront ablation profiles for wavefront aberration supported customized corneal ablations.

In order to meet the design requirements of a Long Slit Spectrograph(LSS) worked at the World Space Observatory,the LSS primary structure was designed and a structure optimization method for the LSS primary structure was studied. An optimization model of the LSS primary structure was established by taking the basic frequency as the constrained conditions and the total mass as a objective function,and an Ant Colony Optimization(ACO) algorithm with an improved pheromone updating rule was put forward. Based on the improved ACO algorithm and the Finite Element Analysis(FEA) software ANSYS,a combined optimization design method was presented to solve above structural optimization problems. The optimized structure obtained from the combined method has reduced about 25% in mass as compared with that of the original design and by 7% that of the zero-order optimization method provided by the Design Opt module of ANSYS.Experimental results show that the optimal primary structure satisfies the design requirements. The combined method takes advantages of the characteristics of the ACO algorithm and software ANSYS and has potentials to be applied to other similar engineering structural optimizations.

A new algorithm model for target location was proposed to increase the detecting capability of infrared panoramic searching systems and a location error formula was also deduced.In the new algorithm model,the effects of the distance a target,measurement errors of azimuth and elevation and the measurement errors of longitude and latitude for receivers on the detection precision of target location were investigated.The formulae for the 3D coordinate and distance of the target were obtained by a matrix analytical method.The location errors of target in four different methods including three-station method,square-station method,rhombic-station method and five-station method were analyzed and compared using the Geometrical Dilution of Precision (GDOP) graphs.Analyzed results show that the GDOP distributing order of five-station is similar to that of the square-station,but the error values are higher than that of the square-station. The error distribution of rhombic-station is different from that of the square-station,and the location behavior is worse than that of the square station.The results show that the location error of square station is the smallest among the four different models mentioned above.

A compact hyper-spectral imaging system characterized by all the optical elements integrated on the glass surface combined two pieces of normal glass was reported.The working principle and configuration layout of the system were introduced and a designed example and its image quality evaluation were given.The optimized hyper-spectral imaging system consisting of three concentric spherical optical elements is less than 30 mm×30 mm×30 mm in sizes and its common center,entrance slit and the CCD detector are in the same plane.it works at the visible wavelength band (0.4~0.8 μm) with the optical system higher than F/2.5,unit magnification,negligible distortion,and both objective and imagery in telecentrics.Moreover,the imager has the advantages of high spectral resolutions, good stability,and imaging quality closed to diffraction limit,and can be widely used in the fields of aerospace,biomedical sciences,etc.

To reduce the contradiction between the system space and the load-carrying capacity of a launching system and with the demand of reflectors with large apertures and high resolutions,a light-weight and flexible membrane reflector was developed and the precise control of the electrostatic stretched membrane reflector was realized.For a Ф300 mm electrostatic stretched polyimide membrane reflector coated with aluminium controlled by concentric annular electrodes,the control matrix for the reflector shape was determined by deducing the influence function of each electrode and the needed voltages on the distributed electrodes for the shape precise control were acquired by the least-square fitting on the basis of the approximate solution of the Poisson equation.The result by proposed method was compared with that obtained by ANSYS.It is shown that when the membrane center deflection is higher than 2.5 mm,the result is very different from that obtained by ANSYS,and the computed deflection deviates from an ideal deflection greatly.It cames to the conclusions that predicting and controlling the membrane shape of the small deflection exactly must use synthetically numerical computations and the least-square fitting,and also must find the membrane control matrix of the distributed electrode to solve the distributed voltages.

For calibration of a recent developded spectral radiation remote sensor, a new integrating sphere radiation source with an illuminated outside laser was developed.The radiation source not only could meet the demand of the calibration for radiance responsivity,but also could scan the spectral responsivity of hyperspectral detectors accurately.A single-wavelength laser was illuminated to research the basic radiometric characteristics of the integrating sphere source.The results show that the radiance instability of the radiation source is 0.09% in half an hour,the radiance plane nonuniformity is spatially 0.16% in the central region of the sphere exit port,and the angular nonuniformity in horizontal and vertical planes within a range of ±22° are 1.8% and 0.7% respectively.These results indicate that the proposed system can calibrate directly the spectral radiance responsivity for optical instrumerts under the radiance mode,for it has advantages of higher spectral flux,lower wavelength uncertainty,and wider tunable source wavelengths.

The general expressions of the ground resolution and the telescope focal length for an imaging spectrometer were deduced to investigate the increase of energy integral time with motion compensation.The ground resolution is inverse proportion to the pointing angle for the imaging spectrometer with the motion compensation.It will decrease when the pointing angle increases,and vice versa.By giving a pixel size of 20 μm for a detector,an orbit altitude of 600 km and a beam deflexion angle low than 30° for the imaging spectrometer, the effective focal length of the telescope was calculated(400,470 or 435 mm),and the relation curves between the relative ratios of ground resolution and the beam deflexions were presented.On the situation of certain integral time, the expression of anti-scanning angular velocity of the pointing mirror was deduced.The result indicates that the anti-scanning angular velocity of the pointing mirror is the functions of the velocity of secondary planet,orbit altitude,the beam deflexion angle ahead of schedule on gazing an object,aviation time and the radius of the earth.The integral time will increase by N times when anti-scanning angular velocity of the pointing mirror is changed.

The effect of the collimation degree of light beam of a Light Emittering Diode(LED) on the performance of optical encoders was researched.By analyzing on the impact of radiation angles of a light source on the grating flux,and by applying of a rotation matrix method and frequency-domain analysis, the light distribution function formula of Moiré patterns was deduced and its simulation waveforms with various slit widths were presented.Furthermore,the influences of collimated light source on the contrast ratio of Moiré patterns and the grating gaps were analyzed.The Experimental results indicate that the contrast ratio of Moiré patterns is reduced by 17.66% when the value of slit width is reduced by 5 μm,and obtained results is well coincident with the simulation waveforms.Therefore,the proposed Moiré pattern formula relative to the collimation degree is very helpful to the design and production of high quality encoders.

The polarization aberrations of an optical system were analyzed to control the polarization effects and to improve the optical performance of the transmitting rotationally symmetric optical systems.The expressions of the polarization aberrations were deduced by a method of polarization ray tracing based on a polarization aberration theory.Then,the factors affecting the polarization aberrations were analyzed and the affecting factor of a transmitting rotationally symmetric optical interface was defined.Finally,the polarization aberrations in a telescope objective were analyzed.Obtained results show that the polarization aberrations at an incident angle of 18.77° are 5.5 times as large as those at an incident angle of 7.86°,the polarization aberrations of the interfaces coated with Layers 2 are decreased by 30% as compared with those coated with MgF2 and the polarization aberrations at λ=1.1 μm are 10 times as large as those at λ=0.55 μm.The results indicate that polarization aberrations must be controlled to improve the optical performance of the transmitting rotationally symmetric optical systems with large fields of view,large incident angles or worked in wide wavelengths.This paper presents two kinds of methods for control the polarization aberrations of this type of optical systems effectively,and they are to keep the angles of incidence as small as possible and to minimize the polarization separation by optimizing the coating design.

Signal processing method for quadrant photo-detectors are investigated and how to obtain the spot parameters from quadrant photo-detectors is analyzed.Based on the research of the sources of their indices and errors for the detectors,a new calculation method on obtaining the spot parameters of quadrant photo-detector by a jiggling method is proposed.In comparison to the traditional methods,the proposed method by jiggling method can obtain not only the spot center,but also the spot radius,which can provide more relevant information different from any traditional method.In addition,this method has the advantage over the traditional method that the errors are unrelated to the spot center and the spot radius.Theoretical analysis and simulation experiments have proved that obtaining the spot parameters of quadrant photo-detector by jiggling method has considerable advantages.Field tests in the atmospheric laser photoelectric precision positioning system of inter-vehicle communications is shown that this method is advantageous in speed,accuracy and convergence.Obtained parameters are from 2 to 3 with the method.When the 30 mm GaAs-PIN quadrant detector is used,the maximum error decreases from 18% to 1.2%.

A new technique called fixed abrasive surfacing technique is presented for manufacturing SiC reflecting mirrors,and a mirror with good quality is obtained by grinding with bigger diamond abrasives at a high speed and the fixed motion between abrasives and workpieces.The material removal characteristics on the SiC are measured at certain rotation speed and pressure by W7,W5,W3.5,W1.5 pellets,and the removal characteristic curves obtained by above experiments show that this method can offer a higher removal rate and stability for the mirror manufacturing.In addition,a surface roughness experiment is undertaken,and the results indicate that a surface roughness of 42.758 nm rms can be obtained by W7 pellets.The experiment also indicates that the surface roughness is decrease with changing pellets in a smaller diamond diameter.At the end of the experiment,a smooth surface in a roughness of 1.591 nm rms is gotten after using W1.5 pellets.These results conmonstrate that the proposed technique for manufacturing SiC reflecting mirrors with fixed abrasives is able to replace the traditional manufacturing technology with slurry abrasives completely in certain finishing phases.

A new method for precise machining non-cylinder pin holes of pistons was presented by using Giant Magnetostrictive Material(GMM) smart components.To eliminate the impact of GMM smart component hysteresis and nonlinearity,a real-time hysteretic compensation control strategy combining a CMAC neural network feedforward controller and a sliding mode controller was proposed to implement the precision position tracking control of the smart component.The output and the output rate were used as the input data of CMAC neural network of the current smart component,the input current as the output of the neural network,and an inverse hysteresis model based on the GMM smart component was established by the CMAC network on-line learning.The model approximate error of CMAC neural network and the external disturbance were eliminated by using the discrete sliding controller.Simulation results show that the control strategy could on-line obtain the inverse hysteresis model of the smart component with a controll error less than 1.5%,and could eliminate the hysteretic nonlinear impact on the smart component.

In order to research the Total Dose Effects (TDEs) of the storages for Very Large Scale Integrated circuits(VLSIs),two kinds of devices,SRAM and ROM, are used as samples in a radiation experiment,and a60Co γ source is selected as the radiation source.Experiments indicates that both devices are sensitive to the TDEs,and their damage thresholds are 10~15 krad(Si).Experiments also suggest that research on the TDEs of SRAM and ROM should take the static consumption current and dynamic consumption current as the effective criterions for the damage threshold of TDEs other than testing the data access function only,because the static consumption current and dynamic consumption current are all sensitive to TDEs.On the foundation of analysing the device structure and experiment results,the damage reasons of TDEs are analyzed,and results show that the damage of CMOS SRAM is from the interface effect caused by TDEs and the damage of floating gate ROM is from the oxide charges and interface trapped charges caused by TDEs.

A resolver excitation and reolver-to-digital conversion system based on a single DSP is designed and some experiments are carried out to examine its validity.A linear angle calculation method with little operation or a few memories is proposed,for the inverse tangent operation of the resolver-to-digital conversion in an original system needs complex computations and a large memoried space,even an additive memory chip. Then,the voltage amplitude of the resolver excitation is reduced based on enough angle precision to obtain an excellent integrated system combinng the resolver excitation and reolver-to-digital conversion system and to provide the low power consumption of the system.The experimental results can offer the angle precision of 0.013°,and the CMG gimbal motor velocity precision of 0.000 4°/s.The bulks,weights and power consumption of this system have reduced by more than 80% compared with those of the resolver excitation module JD20-D15C36MK and resolver-to-digital conversion module 19XSZ2413,which meets the space application demands of small size and low power consumption.

To shorten the design period of control system of a satellite attitude control flywheel,and to verify the feasibility and effectiveness of design schemes,a control model based on a mathematics model for flywheels was established combined with acceleration techniques,reversal connection braking control and commutation control for the control system of flywheels. By taking the speed instruction in a speed control mode for an example,the speed responses,currents,terminal voltages and torques were analyzed by the variable voltage speed control method and the variable voltage speed control combined with reversal connection braking method.The results indicate that the latter has better performance than that of the variable voltage speed control method at a speed instruction of 0.025 Hz.

To fabricate micro/nano structures on a large-format substrate,a novel miro-area micro/nanoimprinting lithographic system was presented,and the design of this imprinting system was introduced.Focused on improving stability and realizing the rotation of an imprinting mould,the gas circuit and configuration of the imprinting mould were discussed in detail.By use of polycarbonates,the gratings with a period of 400 nm were fabricated and the precision of the platform was given.Then,by taking the fabrication of a diffusing film in a backlight unit of flat panel displays as an example,the potential application of a distributed micro-area micro/nano-imprinting lithographic method was proposed.The experimental results indicate that the positioning precision of the platfom in an 1 m long stroke is 1 nm and the rotating angle varies from -90° to 90° when the imprinting depth is controlled by heating temperature and gas pressure.A diffusing film with a thickness of 125 μm was fabricated,whose diffusing radius is 5 mm.These results suggests that the proposed micro-area nanoimprinting lithographic system should be especially suitable for fabrication of optical variable diffraction images and devices used in flat-panel displays.

In order to research the levitation capability of a ultrasonic bearing,a bidirectional ultrasonic bearing which levitates and supports a shaft in axial and radial directions respectively was introduced,and the max rotation speed of the shaft was measured.The generation mechanism of the ultrasonic radiation pressure from this kind of bearing and the working principle of the ultrasonic bearing were analyzed,then,a bidirectional support ultrasonic bearing was designed,and the max rotation speeds of the shaft with different levitation clearances were measured.The results indicate that when the shaft weigh is 85.5 g and the levitation clearance is 8.53 μm,the max rotation speed could be 8 946 r/min.The results also show that the bidirectional support ultrasonic bearing could support the shaft in a levitation state,and the smaller the levitation clearance and the greater the bracing stiffness of the bearing support are,the better the stability and the higher the rotation speed of the shaft are.

The control tactics and the relevant software architectures for a microassembly were studied and the requirement, main process and the human-robot interaction in the microassembly were analyzed in detail.A microassembly workstation based on the machine vision was constructed and a look-then-move control pattern for the micromanipulation was proposed.Then,A three-layered control architecture consisting of a task layer,a strategy layer and a behavior layer was established.In order to improve reusability,the relation between core classes and associations was analyzed in software architectures.Finally,the classes in the task layer and strategy layer were mainly constructed with an aggregation mode to finish the software for microassembly control.The assembly experiments of the microsystem composed of six microparts without special surface markers were carried out with the microassembly workstation.Several critical technical indexes of the assembled microsystem were measured and analyzed with a tool microscope.The experimental results show that the average values of coaxiality errors with the automatic microassembly are close to the results with the full manual microassembly,but the average values of symmetric errors are lower than that with the full manual microassembly and the uncertainty with the automatic assembly is better than that of the traditional one.The efficiency,repeatability,reliability and the reusability of microassembly have been improved with the control tactics based on the machine vision and the software based on three-layered control architectures.

A registration method for unorganized point clouds in a structure light 3D scanning measurement is researched and a inheriting and optimizing algorithm is proposed to implement the precise registration of point clouds.The principle of the algorithm is explained,and a mathematical model is established to obtain the rotation and translation parameters.The realizing procedure of the algorithm is described and analyzed.6 groups of point clouds of a car flashing are acquired by a fringe projection 3D scanning instrument,then,the point clouds are registered by the proposed algorithm.Registration errors are analyzed by a multi resolution hiberarchy method,which indicates that the proposed algorithm can implement the precise registration of unorganized point clouds.Experimental results show that the registration standard error is less than 0.10 mm,and the registration time of 2 point clouds is less than 2 s.The required number of iterations are reduced by more than 5 times as compared with that of an ICP algorithm.

A new gravity matching aided navigation algorithm based on maximum correlation theory is put forward after analysing on the algorithm of INS/gravity matching integrated navigation.The new algorithm can subdivide the gravity map and can eliminate large numbers of interferential data in data gravity map searching process.By constraint conditions to improve the traditional gravity matching algorithm based on maximum correlation effectively.The simulation result shows that the new algorithm can lower the magnitudes of inertial navigation errors from 10-2 to 10-5 and can reduce calculation capacities from 1015 to 107,which meets the requirement of a matching error less than 0.01°.These data reported here prove that proposed algorithm is more effective and more adaptive to restrain the errors of the gravity data,and can improve the orientation precision,decrease the influence of Etvs effect on the gravity matching,and can speed up the gravity matching for navigation in real-time.

In order to improve the measuring precision and controling precision of the Line Edge Roughness (LER) in microelectronics fabrication technologies,the influence factors of LER measure-ments for semiconductors using Atomic Force Microscope(AFM) is studied.Firstly,the LER features are extracted from an AFM image of single crystal silicon based on image processing techniques,and the parameters are determined to quantitatively characterize the LER features.Then,according to the characteristics of LER measurement and characterization method,the influence factors of LER measurements using AFM,including the nonideal properties of the sizes and shapes of probe tips,the signal noises of AFM images,scan sampling intervals,the driving precision of piezoelectricity,cantilever oscillation and the free parameters of edge detection algorithm and so on,are analyzed theoretically and experimentally,and the methods for restraining and amending the measuring errors are respectively proposed.The study indicates that a more precise measurement result for LERs using AFM can be obtained by eliminating or reducing the influences of various factors on measuring errors by proposed methods,which can provide a theoretical and methodologic support for improving the accuracy of nanometer-scale line structure topographic measurement.

A method to calibrate the transform matrix between the sun sensor coordinate system and the cubic prism coordinate system was proposed.Firstly,the sun sensor coordinate system and a 2D rotator coordinate system were calibrated with a sun simulator and a high-accuracy 2D rotator.Then,the cubic prism coordinate system and the 2D rotator coordinate system were calibrated with a diaphragm screen and a CCD image-collimate system.Finally,the transform matrix between the sun sensor coordinate system and the cubic prism coordinate system was derived by taking the 2D rotator as a medial coordinate system.Thereby,the sun vector measured by the sun sensor was indirectly transferred to the cubic prism coordinate system.Experiment results show that the calibration precision is better than 4″(1σ) ,which can meet the requirement of sun sensor measurements.Proposed method can relax the demands for precise machining and installation,and the sun sensor and 2D rotator coordinate systems shouldn't be in coincidence when calibration.It shows that the method is easy to carry out and worth to use.

A multi-point real-time nonuniformity correction method based on a System on Programmable Chip (SOPC) from Altera Company is proposed,for a large amounts of image data need to be processed in the nonuniformity correction of infrared focal plane arrays (IRFPA) and the data are difficult to correct in real-time.By creating custom instructions for the NiosⅡ embedded processor,the software algorithms are implemented by a hardware to improve the nonuniformity correction speed of the IRFPA in real time.The experimental results show that the correction speed of utilizing the SOPC hardware has increased by 150% as compared with that of utilizing the TMS320C6201 DSP hardware for the same multi-point correction.The SOPC correction implementation is characterized by its simple and flexible method and clear images,it can satisfy the requirement of real-time nonuniformity correction of IRFPA.

According to the module principle of visual information processing,the color constancy under variable illumination is studied and a novel color constancy image enhancement algorithm in two steps under variable illumination is presented.After estimating the color offset of an original image using the pixel near edges, the color offset is corrected with white balance.Then,based on the correlation characteristics of three components in RGB color spaces,the RGB three components of the color image are enhanced at the same proportion by brightness gain curved surface to ensure the hue to be a constance.The experiment on the SFU database indicates that the contrast of the processed image is four times as high as that of the original image,and the brightness change is suitable.The algorithm overcomes the ill-posed problem of traditional color constancy algorithm and can better maintain the color constancy for the images under variable illumination with color offsets or lower illuminating,and can enhance effectively the contrast and brightness.

In the Interacting Multiple Model (IMM) algorithm based on a current statistical model,the current model probability can not easily be decided,and the system parameters amax and a-max can not be adaptively adjusted.In order to overcome the above disadvantages,an IMM algorithm based on current statistical model is presented with adaptive fuzzy logic.The working principle of the algorithm is introduced,and basic steps are given.Then,a modified IMM algorithm based on current statistical model is presented with adaptive fuzzy logic.The selection probability of the modified model is given by fuzzy inference to improve the tracking speed,and the design parameters of the modified algorithm are optimized with an ant algorithm to improve the tracking precision.Finally,the modified IMM model is used to simulate maneuvering targets.The experimental results indicate that the modified algorithm can improve the tracking precision by 20%,and the simulation time of one maneuvering tracking process is 0.047 s,which can satisfy the system requirements of fast tracking and high precision.

For improving classification accuracy,a novel fingerprint classification algorithm was proposed by combining the special capability of a singularity method and the Hidden Markov Model(HMM).The belief functions of the singularity classification and the HMM classification was assigned,respectively,then the combined belief function from the proposed method was determined by the Dempster-shafter(D-S).Finally, fingerprint classification was accomplished according to the classification criteria.The results show that the proposed method explores the effectiveness of singularity extraction and the capability of HMM in dealing with low-quality images in fingerprint classification.An experiment based on standard fingerprint datasets has verified that the classification accuracy reaches 94.5%,which indicates that the performance of the proposed algorithm is better than that of the singularity classification and HMM classification,respectively.

In order to realize the automatic customizing for shoe lasts,a segment-based adaptive algorithm is proposed.After fitting the regular data clouds to NURBS curves,the standard shoe lasts are contracted or enlarged in each feature segment according to foot shapes to perform local modification.Then,an energy optimization-based method to modify the control points and the weight factors is proposed to fit the standard shoe lasts adaptively and to realize digital customizing the shoe lasts.The automatic custormizing shoe lasts are performed by the presented method,and the procedure from scanning the foot shape to establish a last 3D model takes time less than 120 s.The results show that the algorithm can meet the requirements of comfort and aesthetics,and can shorten the customizing time.

A location method combining of the fusion for color and edge features with the object authenticity confirmation is presented to solve the problem of Vehicle License Plate (VLP) location under complex environments.The algorithm fused by color features and edge features can extract all possible objects because the vehicle license plate is charaterized by its fixed color assortments and rich charater edges.Then,the object authenticity confirmation by virtue of the features of concomitance and complement is used to realize a feedback location to obtain the multi-object or multi-type object locations under complex environments.An experiment is undertaken with 981 actual RGB color images from various complex environments,and the experimental results show that the object location accuracy is over 99 percent,which proves the proposed object location method is effective and practical.

Aerial cameras on the aircraft will work with a side-oblique angle,when aircraft flies in a steep state,and many image motions with different rates occur on the focal plane array simultaneously.To compensate the different rates of image motion,an image algorithm is presented to improve the quality of blurred motion images in the aerial camera system.On the basis of the mechanism of the image motion with different rates generated in side-oblique situation,the entire blurred image is segmented into many slices by their different motion rates.Then,the slices are divided into pixel lines continuously to improve the computing speed,and a specific parallel computing scheme with 1D Wiener filters is utilized to restore all the pixel lines.Finally,a restored image is obtained by processed pixel lines combined together.The experimental results show that the proposed parallel computing algorithm can restore a blurred image of 2 048 bit×2 048 bit in 17.11 ms,with the help of the parallel computing capability of a GPU,and can offer good image details with a Peak Signal Noise Ratio (PSNR) of 30.469.

In order to realize defect inspection for train wheel treads,a defect inspecting system of the train wheel treads based on computer vision is established and defect region extraction techniques for the images of the train wheel treads are investigated.The tread region is extracted by an adaptive thresholds algorithm,then,a method combining the approximate location based on the block segmentation with the accurate location is adopted to extract the burning defect regions,and the peeling defect extraction method based on searching tread edge lines is used to extract the peeling defect regions.Two experimental examples are carried out to verify the effectiveness of the proposed method.The experimental results indicate that the missing recognition rates of the burnings and peelings are 8.3% and 5.3%,respectively,and false acceptance rate is 5.1%,which shows that proposed method can lay the basis for the feature extraction and defect recognition further.

An improved star identification algorithm based on the traditional triangle method is proposed in this paper to overcome the low character dimensions and high percentage of redundant matches in star identification.When a triangle is recognized by a labeling method,the fourth star is introduced as a vertex to make up a tetrahedron with the triangle,and the accessorial angular distances among them can be used not only as the supplementary character to reduce redundancies,but also to verify whether the result is correct.The proposed algorithm takes the angular distance as the main identification character and relative magnitude information as the assistant character,so that the capacity of guide star database decreases to 0.4 Mb.During star-pair matching,the hashing search is used to construct the function for identification characters to reduce travelling times and time consumption for the guide star database and to speed up the identification process,and the average identification time is 0.43 ms.The result of simulation demonstrates that the improved identification algorithm has advantages of high identification efficiency and short identification time in the all-sky autonomous star identification.

By considering the advantages of the 2D Dual Tree Complex Wavelet Transfer(DTCWT) in shift invariance and directionality,a threshold denoising formula based on parental and neighboring coefficients is constituted and a novel Parental and Neighboring DTCWT(PNDTCWT) image denoising method is presented.By proposed method,the shrinkage threshold of each coefficient is calculated to use in denoising for the current coefficient.After 2D DTCWT transfer to an original image,the final image is obtained by the inverse DTCWT of these denoised coefficients.Experimental results show that the denoising performance of the PNDTCWT is better than those of other denoising methods based on DWT or other DTCWT methods,and its Peak Signal Noise Ratios(PSNRs) have improves by 0.5 dB averagely as compared with that of parental coefficients based DTCWT denoising method.In terms of visual quality,PNDTCWT can get the images with more details, smooth profiles and without confusion effect.

Two kinds of algorithms are proposed to dynamically estimate the precise of the Virtual Channel Model(VCM) to improve the performance of Distribution Video Coding(DVC) system.According to the temporal dynamic property of the VCM,a new frame level VCM estimation algorithm at the decoders is proposed.For each coding frame,the bidirectional vectors acquired in side information interpolation are used to get motion compensated versions of the key frames,then,the residuals of the key frames and their compensated motion versions are used to estimate the VCM of current coding frame.According to the spatial dynamic property of the VCM,a novel block level VCM estimation algorithm is proposed also.The video frames are divided into blocks,then the VCM estimation algorithm proposed is adopted to get its own VCM for each block.Experimental results show that for different video sequences,the proposed algorithms improve rate distortion performance by 0.5~1 dB for the DVC system,and reduce transmission code rates for video encoding systems.

An extraction method based on voxel connectivity and region growth is proposed for the extrusion features of range images.A voxel model of range image data is established with linear octree codes,and a fast transforming algorithm from octree codes to grid codes is introduced to improve the efficiency of searching voxel neighbors,thereby the computing time has been reduced by 50%.Then,the region growth is carried out based on the connectivity of the seed voxel in a specific direction,and a threshold is used to judge the growth result to extract the extrusion feature data.Finally,the method is applied to the liquid cabin range image data of an actual ship and the results of the extraction and threshold option show that the extracting accuracy reaches about 90% when setting the comparison precision as 10－4 mm.The experimental results support a conclution that the method can be used in a point-cloud cutting method to reconstruct the extrusion features.

By considering the color randomicity of foreigns and the color difference between the regular materials and the foreigns in fast foreign real-time recognition,the specialities of an One-class Support Vertor Machine(OC-SVM) for foreign recognition are studied,and an approach to real-time recognition based on hypersphere OC-SVM is presented.In this method,the OC-SVM is used to determine the color distribution of the regular materials to find out the foreigns out of the range of the regular materials in color.The centrifugal coefficient ω is presented to reduce the amount of computation,and the Zoutendijk fastest decline method is adopted to seek the working sets to simplify the number of iterations.The experimental result shows that the proposed method for color recognition can offer a higher computation speed and recognition ability,especialy for some foreignes whose color are similar to those of regular materials.The data from experiments indicate that the compution time has reduced by 20% as compared with that of LibSVM,and the general accuracy of recognition by using this method is 8%~10% higher than that by one dimension normal distribution and three dimension normal distribution fitting algorithms.