An adaptive optical system based on an piezoelectric deformable mirror was established to compensate the wavefront aberration of laser beams by using a 37-unit piezoelectric deformable mirror controlled by the hybrid Simulated Annealing-Hill Climbing (SAHC) algorithm. With the hybrid SAHC algorithm, Simulated Annealing (SA) algorithm was used first to provide a good start point, and then the hill climbing algorithm was used to achieve wavefront aberration compensation. It needed only a few iterations of SA process to provide a good start point. As the hill climbing algorithm has a very fast process speed, the number of iterations to obtain the same correction effect has been reduced by 50% as compared with that of the normal SA algorithm. Meanwhile, the probability of falling into the local optimum is very small, for it has the good start point. The SAHC algorithm is proved to have better correction efficiency for laser beam wavefront by experiments.

A time tester was developed to evaluate the quality of a traffic safety lighting in mesopic vision conditions. Firstly, test parameters including background brightness, visual target contrast and visual target eccentric angle were set by collecting the reaction time of human eyes to observing objects in three kinds of light sources(Light Emitting Diode/High-pressure Sodium Lamp/Metal Halide Lamp,LED/HPS/MH),and the vision efficiencies of different sources were researched.Then, utilizing the regression analysis of correlation coefficient method, the relationship between reaction time and three test parameters was obtained, respectively. Finally, based on visual performance method, the reaction time tendencies and road traffic lighting performance of the three light sources were analyzed in the mesopic vision. The experimental results show that under the same visual performance conditions, when the background brightness is less than or equal to 1.5 cd/m2, the background brightness of HPS is the maximum, MH lamps take the second place and the LEDs are the minimum. The conclusion is that the visual performance of the LEDs is better than those of other two light sources in the mesopic vision, therefore they are more suitable for road traffic safety needs.

Since the shear directly determines the displacement gradients of both out-of-plane and in-plane in the shearing speckle interferometry, the measurment precision of the shear should be improved to obtain the accurate strain distribution during the measurements of out-of-plane and in-plane strains. In this paper, four methods for the shear measurement: formula method,imaging method,Moiré fringe method and correlation method are compared in detail. An experimental setup is established to measure the shear using the imaging method, Moiré fringe and correlation method. The measurement results based on the three methods are compared with that of the formula method as a criterion. The result shows that when the shear is more than 3 mm, the correlation method has higher accuracy; however, once the shear is less than 3 mm, the Moiré fringe has higher accuracy. It suggests that the appropriate measurement method can be chosen after estimating the amount of the shear to obtain the higher out-of-plane and in-plane strains.

A modified Czerny-Turner spectral imaging system was developed based on aberration theory to minimize the large astigmatism in classical Czerny-Turner spectrometers. The astigmatism from a plane grating placed in the divergent light beam was used to compensate the astigmatism from an objective lens. The broadband astigmatism corrected simultaneously conditions were deduced, and the astigmatism was corrected in a wide spectral region. The principle and method of astigmatism correction were analyzed in detail, and the initial parameter computing was programed. As an example,a Czerny-Turner imaging spectral system operating in 540-780 nm was designed. The ray tracing and optimization for the spectral imaging system were performed with ZEMAX-EE software. The analyzed results demonstrate that the total field-of-view modulation transfer function is higher than 0.52 in the whole working spectra.The system shows good imaging quality due to the astigmatism to be corrected in the wide spectral region synchronously.Obtained results prove the feasibility of the modified method.

A dual-channel active laser detection system based on an Avalanche Photo Diode(APD) and a Charge-couple Device(CCD) is designed to effectively detect weak echo signals of “cat-eye” targets and to acquire the intuitive image information of a target area. Firstly, the concept of hardware design is introduced, including the functions and constitution of some modulies and the types of primary assemblies. Then,based on the key problems of designing receiving-magnifying circuit in the APD detection mode, the circuit constitution, working principle and the selection of main parts are discussed,and the match of response spectrum for the CCD and laser is analyzed. Finally, the detection experiments are performed by using typical “cat-eye” targets at different distances. The experimental results show that the echo signal power of the typical “cat-eye” target is 2.87 times that of mirror-like surface at a distance of 550 m in the APD mode; and the echo signal power of a telescope and that of optical window for a laser finder are 2.72 times and 2.31 times those of background at distances of 550 m and 2 500 m in the CCD mode,respectively. The dual-channel active laser detection system can capture the “cat-eye” targets from background and jamming targets easily,which proves the rationality and validity of the proposed system.

An optical system to perform the semiconductor laser myringotomy with the help of a video display and a pilot laser was designed and manufactured. First,the 650 nm laser output from a single emitter and the 810 nm laser emitted from a laser diode array were coupled into a fiber with a core diameter of 200 μm and a numerical aperture of 0.22 by using beam shaping and wavelength beam combination technologies. Then, an achromatic lens was used to collimate the dual wavelength laser output from the optical fiber and the other achromatic lens which could move along principal axis and consequently change the spot size was used to focus the dual wavelength laser to the eardrum. Meanwhile, laser and imaging lights were coupled as coaxial by using a hot mirror and the imaging components from markets. Moreover,the location of myringotomy could be piloted by red light on a computer monitor.According to the results of measurement,the laser power output from the system can be continuously adjusted from 0 to 13.3 W, and the spot size can also be adjusted between 1-3 mm. The system can decrease the operation time and the possibility of surgery side effect; moreover, the myringotomy for children needs not any anasthesia. In addition, the system has the virtues of low volume, light weighted and high wall-plug efficiencies compared with other candidates.

A calibration approach based on active vision for a light plane is proposed to achieve the self-calibration of a line structured light sensor. The proposed method is different from the previous plane fitting methods using the target points on the light plane, and the entire calibration process can be accomplished in two steps by several translational motions of a sensor. Firstly, the projections of parallel lines in the light plane are captured through some translational motions, and the normal vector of light plane is obtained via computing vanishing points. Then,in order to satisfy the model of perspective-three-points, the sensor is controlled to move along the set direction twice which is perpendicular to the normal vector, and the distance from the origin of camera coordinate system to the light plane is derived eventually. Experimental result shows that this calibration method has high accuracy, its average measuring accuracy is 0.015 4 mm, and relative error is about 0.183 0%. This calibration approach belongs to the self-calibration method without calibration targets, therefore, the calibration cost is reduced. It concludes that the whole process is easy, flexible and suitable for calibration on sites.

A new method to measure the spectrum of a pulse light was developed. As the pulse width could be broaden due to the dispersion when a pulse light passed through the optical fiber, the calibration and measurement were proposed to correct and recover the broaden waveform to obtain the real spectrum of a pulse light source. In calibration, the time response function of measuring system was gotten with a ps laser firstly; then the group refractive indexes at different wavelengths were obtained by measuring the time delay of given wavelengths; finally, the relative attenuation index of each wavelength in optical fiber was given by measuring the spectral attenuation. On the basis of the calibration results mentioned above, broaden waveform was recovered by the digital deconvolution, time delay and attenuation .The pulse spectrum of a red scintillator was measured with this method. Results demonstrate that the difference of the center wavelength of the scintillator given by this method and other methods is less than 2 nm, which shows the spectra of pulse light can be measured based on fiber dispersion when the pulse width is much less than its dispersion.

Many small aperture satellite optical communication terminals utilize a periscope as the Coarse Pointing Appliance(CPA), however, the model of coarse tracking arithmetic becomes much complex since the optic structure of the periscope-type CPA varies with the angle of attitude in the CPA . To improve the tracking accuracy of the CPA, this paper researches coarse tracking arithmetic of periscope-type satellite optical communication terminals. Based on the optical matrix method, a matrix model with geometric optics for tracking in the periscope-type inter-satellite optical communication terminals was established. After obtaining the CCD-based angle measuring model, an automatic tracking model of the CPA was derived to implement the fast tracking for the target beams. A test was performed to testify the arithmetic. The test results prove that the tracking arithmetic can express the configuration of optical system exactly, and can process the information of locality from CCD accurately. The average tracking accuracy is less than 10 μrad. The test results also show that the tracking model can meet the requirements of the satellite optical communication. The method proposed in this paper would supply some helps for similar optical study.

The CO laser frequency mixing in high-quality ZnGeP2 and GaSe crystals is studied to obtain the 2.15-1 500 μm coherent sources. Secondary Harmonic Generation(SHG), sum- and difference frequency generations are considered as a method for the CO laser frequency mixing and the mode-locking as an efficient way to improve the mixing efficiency.Results show that the internal SHG efficiency of Q-switched multiline CO laser radiation has exceeded by 0.3% for GaSe crystal and has reached by 1.1 % for ZnGeP2 crystal. When the SHG is in ZnGeP2, the internal efficiency of electron beam sustained discharge frequency-tunable mode-locking CO laser is up to 12.5 %. Simultaneous SHG and sum frequency generation show the same output spectrum. It is shown by modeling that the sum and difference frequency generations of neighboring lines of both fundamental and first overtone bands can allow one to get the oscillation,respectively, at 4.0-5.0 μm and 100-≥1 200 μm (THz). In conclusions, the frequency mixing of mode-locked CO laser emission lines in ZnGeP2 crystals allows some one to design 2.15-≥1 500 μm coherent sources with the power frequency conversion efficiency up to or over 12.5%.

According to the tool path characters of non-cylindrical hole precision machining by giant magnetostrictive components, a dynamic hysteresis model of giant magnetostrictive components was established by a pure delay transfer function and the linearity model between high-frequency driving currents and micro-displacement responses. The pure delay compensation parameters of the system were obtained by the relevant identification of driven currents and output displacements with a certain frequency. Then, a mapping model of the driven currents and output displacements without delay was established. The output displacement met the ideal tool paths of non-cylindrical hole boring by direct inverse model and delay compensation in real-time control. The results in verification experiments indicate that the maximum control error is 2.7 μm, and the maximum relative error is about 10%. By integration of micro-displacement feedback control,the accuracy of the component is improved further, the maximum control error is 1.2 μm, and the maximum relative error is about 7%.

To implement the integral optimization of satellite engineering, a upper-stage spacecraft is proposed based on the combination of a small satellite and a small launcher, and a reconfigurable integrated electronic system for the upper-stage spacecraft is designed to achieve the high function density development of aerospace electronic systems. By taking the Microblaze as a soft-core processor, the electronic system improves its expandability based on a distributed network structure with CAN bus. It realizes the time sharing of a central computer in the core of electronic system and satisfies different requirements of the launching phase and on-orbit phase. Furthermore, the control algorithm has implemented by hardware, which reduces the calculation of processor and improves the computing and processing capabilities of the system. A semi-physical system is built and the full-mode simulation is carried out to verify effectiveness of the designed system. The experiment results show that the control period in the launching phase is 10 ms, the running time of control algorithm is about 7.5 ms, and the attitude orientation accuracy and the stability are 0.035°and 0.000 68(°)/s, respectively. The reconfigurable integrated electronic system can achieve the full-mode flight mission of upper-stage spacecrafts, satisfy the requirements of strong real time in the launching phase, high stability in the on-orbit phase and can offer higher control accuracy.

The control precision of gimbal system for a Control Moment Gyro (CMG) is a main factor affecting the output torque accuracy of the CMG. Generally,a harmonic driver a harmonic driver is adopted to improve the dynamic response ability of the gimbal system, however,the induced mechanical resonance seriously reduces the control precision of the gimbal system. In order to eliminate the resonance and satisfy the control precision of the gimbal system, a kinetic model was set up. A proper damping ratio was chosen to design the dominating poles according to the demands of gimbal system for dynamic performance. The zeroes induced by a controller were calculated to be equal to the poles induced by mechanical resonance, so that the mechanical resonance was eliminated. Simulation and experiment results show that the proposed method eliminates the gimbal resonance of the CMG, and the control precision of 0.175 rad/s is 0.002.When the gimbal system tracks a 0.175 sin(2πt) rad/s sine given velocity, the relative amplitude error is 3.28% and the phase error is 0.13 rad. The control performance satisfies the demands of CMG for high output torque precision very well.

The motion accuracy of a platform based on two compensated flexible four-bar linkages was researched and three error sources,namely, batten size error, the sphere center distance error of ratio lever and the assembling angle between the inner and external battens were analyzed. A semi-beam model was established to analyze the deformation of the batten and the relationship between working displacement and coupled displacement. The expressions of the platform straightness error and the limiting coupled displacement error which was concerned with the platform displacement and the batten sizes were derived. Finally,an autocollimator and a high precise measuring head were used in the measurements of the platform straightness and the coupled displacement. The experimental results show that the straightness error is smaller than 1.5″ and the coupled displacement error is less than 13.7 μm when the stroke of the displacement platform is no longer than 5 mm. It is proved that the platform works with high motion accuracy and the experimental result is consistent with the analysis results.

To improve the bonding efficiency of microfluidic chips, an in-mold bonding method is researched by taking a Polymethyl Methacrylate(PMMA) microfluidic chip as an research object and a micro injection machine as a working platform. Clamping forces and bonding temperatures are provided by the micro injection machine and a mold temperature machine. Isopropanol is selected as assistant solvent to reduce the bonding temperature and bonding pressure with its solubility. At 30 ℃ to 70 ℃，a microscope and a step profiler are used to analyze the influence of assistant solvent on surface morphologies and micro channel structures of chips at different temperatures. An in-mold bonding experiment is operated with the assistant solvent, then the electric universal testing machine is utilized to measure the bonding strength and to optimize the process parameters. Experimental results illustrate that the influence of isopropanol on bonding quality is related with the bonding temperature and bonding period. The surfaces of chips chap and the micro grooves become deformed or block at higher temperatures. The surface quality and morphologies of micro grooves are correspondingly improved when the bonding temperature is 35 ℃, the bonding time is 5 min, and the bonding strength is no less than 2.64 MPa.

To measure the location parameters for projectile burst on a proximity fuze in high altitude,an intersection imaging method by using two planar array cameras and taking a simulant object as reference was put forward to calculate three coordinate parameters of projectile burst location. According to the request of proximity fuze test, the principle and method of camera intersection imaging were analyzed,and the measuring model of projectile burst location was set up. The relations of space geometric locations of the camera layouts, the technology of image processing and the sizes of objects were analyzed,the spatial three coordination of projectile burst location were calculated, and the parameter modification method was studied. A differential method was employed in analyzing the location errors, and results show that the differential coordinate errors are less than 40 mm.Experiments and image processing were performed,which shows that the planar array camera intersection imaging can obtain the spatial three coordination of projectile burst.Furthermore, the data by proposed method are compared with that of the four screen across system,and their average error is 0.158 m in the high and low directions and 0.114 m in right and left directions.

To quickly and accurately predict the thermal error of a ball screw, this paper focuses on the modeling of temperature field of the screw. A theoretical temperature model is achieved based on the heat conduction equation of the screw and the modified boundary value conditions and a parameter α′ which is changed with the temperature are used to improve the model. Then, a model parameter identification process is proposed. The thermal error of the screw is predicted by the model and mechanical thermal deformation theory. Finally, experiments are performed to identify parameters and check up the effect of the model. Result shows that the maximum error between the predicted and the actual temperatures is 0.8 ℃ and the maximum error between the predicted and the actual thermal errors is 3.8 μm. It concludes that the temperature model accurately predicts the temperature distribution and the thermal error of the screw under a single heat.

For the complex control and unpractical calculation of the surface shape control method for an electrostatic stretching membrane mirror in engineering, a semi-analytical method of surface shape control based on an inverse approach was proposed.Firstly,the relation between required pressure distribution and voltages and the desired parabolic shape were derived based on the Karman equation and electrostatic field theory.Then,the finite element analysis was used to establish a model to validate the pressure distribution and a verified test was performed on a Ф300 mm three concentric annular electrode membrane mirror. The experimental results show that the measured values of max displacement are consistent with the theoretical ones basically.The surface shape error has been reduced as compared with that of traditional method and the best RMS and PV values of shape surfaces have reduced by 9.76% and 15.38% respectively under a uniform pressure. In conclusion, the method described can control surface shape and can improve the surface accuracy. Compared with other methods, the method is simpler and more practical.

In order to characterize the anisotropic wet etching rule of quartz crystals, a method to forecast the sidewall profiles of quartz microstructures was studied based on wet etching rates of quartz crystal planes. First, etching rates of quartz crystal planes were summarized and the etching rate vector graphs of crystal planes were plotted for x and y groups. Then, the plane lines denoting their crystal planes were obtained by plotting etching rate vector graphs at each edge of a mask,and the etching shape was the minimum profile enclosed by plane lines. Finally,sidewall profiles of quartz beams at x and y directions were forecasted according to this method. An x direction quartz micro-beam and a y direction quartz micro-beam with a thickness of 500 μm were fabricated after 5 h wet etching process in a mixture of hydrofluoric acid and ammonium fluoride at 70 ℃. Experimental results indicate that -x sidewall of the y direction micro-beam presents a crystal facet with a height of 210 μm and +x sidewall is smooth; and crystal facets with the height of 450 μm and 240 μm appear on ±y sidewalls of x direction micro-beam respectively. These sidewall profiles of the beams are in agreement with the analysis results, which validates the forecasting method of wet etching profile. Results demonstrate that the quartz micromachined structure can be optimized by introducing fabrication processes based on the proposed method in designed steps.

To solve the problem of salivate, an unsteady phenomenon caused by the mismatch of feed velocity and high-voltage in the electrospinning to fabricate nanoscale fibers, this paper proposes an on-line monitor system based on the multi-featured pattern recognition for realizing both the stability of nanofiber manufacturing and the controllability of fiber diameter. Firstly,an industry CCD is used to collect the image of the Taylor cone continously, and then these images are preprocessed by sharpening, filtering and threshold segmentation. The perimeter, area and height of the Taylor cone are recognized by the multi-featured pattern recognition algorithm to judge the shape and size of the Taylor cone. Finally, the result of pattern recognition is viewed as the feedback signal to adjust the feed module and high-voltage module. Experimental results indicate that when the voltage is 8 kV, the system can be stabilized in 0.43 s,and it maintains a stable shape of Taylor cone without the salivate phenomenon.Moreover,obtained nanoscale fiber shows a uniform diameter. In conclusion, this method has reliably and effectively solved the instability and uncontrollability of nanoscale fiber manufarturing.

An accurate global calibration method for a muti-camera videogrammetric system with large field-of-view is presented. Firstly, a cross target with feature points pasted on its two sides is designed as the calibration reference pattern, and the world coordinates of these feature points are reconstructed using the XJTUDP photogrammetric system. Then, the calibration target is placed in the public field-of-view and its images shown at different orientations are captured by the four cameras simultaneously. Finally, based on a 10-parameter nonlinear camera model,the capatured images are calibrated successively and the camera interior parameters and global orientation parameters with respect to a predefined global coordinate system are obtained. Experimental result indicates that the proposed calibration method can offer a considerable accuracy in a re-projection error less than 0.05 pixel.By the proposed method, the four-camera videogrammetric system can achieve a relative accuracy of 1/4 000 on length measurement. It can satisfy the demands of the multi-camera videogrammetric measurement with large field-of-view for precision and efficiency.

High-resolution photoelectric encoder usually achieves a high resolution by the subdivision of two-way orthogonal sine and cosine signals from a fine disk code. In order to improve the subdivision technology, this paper researches the subdivision errors specially. The signal DC component errors, magnitude errors, phase errors, harmonic errors, noise errors and quantization errors are analyzed, respectively. Based on properties of subdivision errors, the laws of errors and their computing formulae are obtained, then, the fairly complete mathematical results of subdivision errors and precision analysis for photoelectric encoders are formed. The results show that the subdivision precision is about 1.5% generally. It points out that it should pay more attention to the effects of subdivision errors on different technological steps such as code selection, axis design, signal extraction, circuit design and adjusting method, when the subdivision is used to improve the encoder accuracy. The research results are useful for analyzing and distributing errors, forecasting product accuracy and reducing design errors in the product design.

In consideration of the effect of radiometric specification of a mapping camera on the matching accuracy of imaging, the quantification of main radiometric specifications are analyzed, and the mathematical model between them is developed to provide theoretical foundations for determining the specifications of the mapping camera. Firstly, Least Squares Image Matching Algorithm (LSIM) is deduced mathematically, and then the main factors impacting matching accuracy are presented, including the scene characteristics, Modulation Transfer Function (MTF) for an imaging system, Signal to Noise Ratio (SNR) and Radiometric Distortion (RD) metrics. Then, the image degradation owing to the above factors is modeled and simulated. Based on LSIM, the mathematical model between matching accuracy and radiometric specifications is developed using Genetic Algorithm and Back-propagation Neural Network (GABPNN). Finally, the model is verified using simulated images and on-orbit images. Experimental results indicate that the model precision is less than 0.01 pixels. It suggests that when the MTFN is set to be greater than 0.08, SNR greater than 45 and RD less than 4% based on the model, the error of matching accuracy can be less than 0.1 pixel.

A network training method for star pattern recognition was designed by combining a classific Radial Basic Function(RBF) neural network and star pattern samples. Firstly,the star pattern abstraction method was discussed and a triangulation based on star magnitudes was induced to connect the stars which probably appear in the same field of view.By taking extrated angular distances as the characteristic of star pattern, a star pattern sample set with completion, translation and rotation invariance was established. Then, RBF neural network was studied to recognize the star patterns. RBF network training method was classified as sequence learning and batch learning. Some typical algorithms that could represent the two methods were studied on their advantages and disadvantages,and a new training method was designed based on the specialty of above star pattern sample sets.Experiments indicate that the designed method is more appropriate than those typical algorithms. Several star images were simulated through software, which was regarded as the observatory data and entered into the trained RBF neural network to test. The experiment results show that the network can recognize all the star patterns successfully.

Without detection of the location of sea-sky line in advance,a feasible method combining with the characters of marine visible images and a complex sea-sky background was proposed to detect distant weak targets in the sequential images from a surface vehicle. Firstly, the images damaged by the noise such as exposures and reflections was mended by the border color of ruined range. Then, the complexity of sub-images and their neighborhood average gray difference were measured to predict the sea-sky region.If the sea-sky region was obtained,the location of the sea-sky should be predicted. Otherwise,the consequent process of images could be eliminated. The strategy of filtering and clustering in Mean-shift segmentation was adopted,in which the surround suppression filter was applied to smoothing the sea-sky region and the smoothed image was clustered by taking the pixels and points as units. Finally, the largest region was assumed to be the background and the rest was taken as target information. The experiment results prove that this method can locate the objects efficiently and the time cost is 35 ms/frame, which is robust and real-time.

An algorithm combining an improved Auxiliary Particle Filter(APF) with a Markov random field is proposed to achieve multiple target tracking in an infrared scene. First, targets are described according to the gray histogram of target regions.Then, sampling particles of all targets are optimized roughly by using standard APF. Meanwhile, Mean-shift is introduced to the process of auxiliary particle sampling to improve the exponential growth of particle numbers and to increase the percentage of efficient particles and the real-time ability. As for the failure tracking from that targets often are covered each other, a graphic model theory is introduced, in which multi-tracking model by the Markov random field is used to describe the multi-tracking model and convert the problem of multi-target tracking into an inferential problem of the graph model. Results indicate that the new algorithm proposed can track targets only by a few particles, and the accurate rate for multi-target tracking is up to 84%, the failure tracking caused by targets covered mutually can be solved effectively.

To realize fast and accurate evaluation for flatness errors, a Modified Artificial Bee Colony (MABC) algorithm was proposed to implement the minimum zone evaluation of flatness errors. The minimum zone method and the criteria for flatness errors were introduced.According to the minimum zone condition, the mathematic model of flatness error evaluation was presented. By introducing two traction bees and a Tabu Strategy(TS), this modified method could enhance the rate of convergence and the quality of optimum solution.The implementation steps of the method were expounded. Then, two test functions were selected in the simulation experiments through analysis, and the results verified the feasibility of MABC algorithm. Finally, proposed approach was used to evaluate flatness errors.The results calculated meet the criterion of minimal condition. On the basis of a group of metrical data, this approach can find the optimal plane by 0.436 second, which saves 0.411 second as compared with that of ABC algorithm.In addition,the flatness value from the MABC algorithm is 18.03 μm lower than that of the Least-Square Method(LSM), and 6.13 μm than the Genetic Algorithm(GA). According to other five measurement data sets available from the Coordinate Measuring Machines(CMMs), the results obtained by the MABC algorithm are more accurate than those by the GA and Particle Swarm Optimization (PSO), and the maximum gap of flatness values is 0.9 μm. Experimental results show that the MABC-based approach outperforms ABC-base method in optimization efficiency, solution quality and stability, and its calculating precision is superior to that given by LSM, GA or PSO. It is suited for the evaluation of position measuring instruments and CMMs.

The Phase Diversity (PD) method in adaptive optical waveforn detection shows great computations when it is used to estimate the wave-front phase aberration and to restore the degraded images, and it is difficult to increase the speed of PD to achieve its real time application on a PC platform. The computational-hardware such as Digital Signal Processor(DSP) and Field Programming Gate Array( FPGA) is a proper way to improve its performance, however, the complex structure of the PD object function and plenty of Fourier transformations in each computation loop influence on its hardware implementation. According to the theory of Zernike polynomial, a method which utilizes polynomial operation instead of Fourier transformations is proposed to modify the PD object function. The modified computing formula and gradient formula are given, by which the computation of the PD object function only depends on the polynomials and the hardware implementation of DSP, FPGA and the parallelism of hardware processing are more easily. A test and an experimental platform are designed, and simulative images and grabbed images for a resolution plate and an optical fiber bundle are restored respectively. Experiment results indicate that the modified PD is still a good means to restore the degraded images, the optical fiber bundle has a higher resolution and its particle profile can be distinguished.

A new on-line boosting algorithm based on sub-regional classifiers was presented to solve the problem that traditional on-line boosting based tracking algorithm often leads to drifting or failure due to the accumulated error during on-line updating under serious occlusions. Firstly, the feature pool was divided into a number of sub-regional feature pools to correspond to their sub-regional classifiers. Then, the sub-regional classifiers were selected adaptively into a strong classifier to eliminate the influence of occluded sub-regions on the target location when occlusions took place. Finally, the sub-regional feature pools were updated partly to solve the problem of accumulated error during on-line learning. The proposed algorithm was tested with variant video sequences and results show that proposed algorithm achieves exact tracking for the object occluded, and the average computing frame rate is 15 frame/s when the object scale is 36 pixel×40 pixel. In conclusion,the algorithm can satisfy the requirements of stability under occlusion as compared with the original on-line boosting algorithm.

A fast local feature description algorithm based on the region histogram of Greyvalue Differential Invariants(GDIs) is presented to improve the traditional GDI feature descriptor with complex computation, worse stability and less information. The GDIs with the meaning of differential geometry derived from low order derivatives are used to describe a feature point to reduce the computation complexity of the feature descriptor and enhance the stability. The intensity and relative location of the feature point neighborhood are made full use of to increase the content of information and improve the robustness of the feature descriptor. Finally, the feature description is used to match images.The experiment results demonstrate that the proposed algorithm can get better matching results in the cases of image zoom, rotation, blurring, illumination varying, smaller viewpoint changes as well as JPEG compression. Furthermore, the processing speed is about twice that of the Scale Invariable Feature Transform(SIFT).