A kind of novel fabrication system of array optical fiber gratings is proposed and multi-wavelength array optical fiber gratings are obtained on 8-core ribbon optical fiber through this structure with only one phase mask. By means of image processing, the dynamic images are collected during the array gratings fabrication and the relative pixel shifts are monitored, and then fiber slip regulation are analyzed and obtained. In order to resist the influence of fiber slip, the specially-designed fixture is designed, and the material and shape of the clamping surface are improved many times. The experiment results show that the macrobending surface of the clamping surface can not only fix the ribbon optical fiber, but also keep the coating in good condition. We successfully obtained array optical fiber Bragg gratings with 0.2 nm bandwidth, wavelength interval less than 0.5 nm, wavelength deviation less than 0.05 nm and reflectivity more than 80% on ribbon optical fiber by the above system.

For the incompletion of the eigenspace and the poor generalization ability of the pattern classifier in the past cognitive system, an ensemble cognitive method for intrusion behavior based on blending features is explored. Initially, use the Ensemble Empirical Mode Decomposition (EEMD), Power Spectral Analysis (PSA) and Discrete Wavelet Transform (DWT) to extract the information of the distribution tendency of the fiber optic signal on the time domain, frequency domain and the wavelet domain to build a relatively completed eigenspace of the fiber optic signal. And then use the Discriminative Function Pruning Analysis (DFPA) feature subset selection method to evaluate the ability of the feature element to discriminate different kinds of intrusion behavior, and then find the best feature subset. The simplification procedure for the feature group is thus accomplished. Lastly, use the ensemble modeling based on Random Vector Functional-Link net (RVFL) to improve the generalization ability of this cognitive model. Simulation experiment on fiber optic signal collected from the fiber optic perimeter security system based on M-Z(Mach-Zehnder, M-Z)interferometer has shown the effectiveness of this cognitive method.

The physical model for the inscription of mid-infrared Fiber Bragg Grating (FBG) is established based on the Finite Element Method (FEM). The diffracted light field by the phase mask is analyzed. Two light sources as the plane wave and spherical wave are adopted in the simulation. When using the plane wave as the incident beam, the variation of the Talbot diffraction pattern is deeply analyzed by changing the incident angle of the plane wave and the groove depth of the phase mask. When using the spherical wave as the incident light, the variation of the diffraction pattern of the phase mask is analyzed by changing the distance between the light source and the phase mask. As a result, when the plane wave is incident in an arbitrary angle, the Talbot diffraction pattern is formed along the same direction as the input. The profile and maximum energy of the Talbot diffraction is changed with the groove depth of the phase mask. With the groove depth increasing, the maximum diffraction energy increases gradually. However, it begins to decrease after the depth reaches certain value. At the same time, the Talbot image tends to evolve to the uniform fringes. When the spherical wave is incident on the phase mask, the distribution of Talbot diffraction pattern is along the propagation direction of the spherical wave. As the distance between the spherical wave and phase mask is decreased, the maximum diffraction energy is higher. Then it begins to decrease after the distance is smaller than some certain value.

7 153.7cm-1 and 7 154.3cm-1 absorption light spectrums of H2O vapor were chosen to measure temperature distributions and H2O vapor concentration distributions of the combustion region of CH4/Air flat flame furnace based on Tunable Diode Absorption Spectroscopy (TDLAS) and modified Algebraic Reconstruction Technique (ART). In the numerical simulation, combustion region was described by 1 000 pixels×1 000 pixels. There were 40 laser rays cross the combustion region in orthogonal two directions. By introducing the initial distribution, we found the temperature field reconstruction relative error was less than 4.5% and H2O vapor concentration field reconstruction relative error was less than 4% based on modified ART algorithm and cubic polynomial interpolation processing. In the experiment, 24 laser rays were used to scan the target region of CH4/Air flat flame furnace in orthogonal two directions with rotary table and slide way. Temperature field and H2O vapor concentration field were obtained based on modified ART algorithm and cubic polynomial interpolation. A comparison of temperature reconstruction result and thermo couple result indicates the method is effective to achieve temperature field reconstruction of the combustion region.

In order to improve the operating and locating accuracy of intelligent equipment under special working environment (high radioactivity, high temperature, high pressure, boric acid and so on) in nuclear power plant. The nonlinear mathematical model of underwater binocular vision is established, considering both refraction effect and lens distortion. To achieve this, it is analyzed that the imaging principle of the boric acid underwater vision as well as the influence factors and presented that a kind of underwater binocular vision positioning method to unstructured environment. In order to verify the method, the comparison experiments are done in the air, common water and different concentrations of boric acid water. The experimental results show that different concentration of boric acid has little influence on the camera imaging refraction, and target attribute dimension calculated is close to actual value.

To improve the sinusoidal feature of binarized sinusoidal fringe pattern, a new error-diffusion method based on Floyd-Steinberg error-diffusion dithering algorithm in traditional halftoning is proposed. Firstly, the error-diffusion kernel is expanded to reduce quantization error of the binarized sinusoidal fringe pattern and restrain the unsymmetrical texture. And then the error-diffusion coefficients are distributed linearly to smooth the fringe pattern and enhance the dispersion. The simulations and experiments indicate that, the proposed method smoothes the fringe pattern and reduces the phase error of 3D measurement compared with traditional Floyd-Steinberg dithering algorithm.

The performance of the inertial stabilized platform will be affected heavily by inner or outer factors, which will change the transfer characteristic of the inertial stabilized platform and result in the degradation of the precision of the system. Therefore, the high-gain acceleration feedback will be inserted in the velocity feedback loop to compensate the change of these factors. And the acceleration feedback loop can stabilize the platform because that the accelerometers are inertial sensors, the active disturbance suppression of the platform is the product of the error attenuation of the acceleration loop and the velocity loop. Extensive experimental results show that the acceleration feedback can enhance the active disturbance suppression of the platform.

The support structure based on kinematic principle had been wildly used in the axial support on large-diameter telescope. But it is not maturely applied in the lateral support system. A lateral support structure based on kinematic principle is explained. Based on a primary mirror with a diameter of 2.04 m, a support structure with more details was designed and analyzed by the ANSYS Parametric Design Language. Selected the Root Mean Square (RMS) of the primary mirror when the optical axis was vertical as the objective function, the support reaction was optimized by the simulated annealing algorithm. Final surface error RMS was 16.67 nm. Surface error with different obliquities were calculated, which meets the requirement that the RMS should be less than λ/3.

The large relative error between the measurements and the fitted values with the least squares method easily occurred when the small outputs of CCD. According to this problem, the evaluation method with minimizing the value of the maximum relative error was proposed to solve the optimum curve fitting with the neural dynamic optimization algorithm. Firstly, the CCD opto-electronic responses were achieved by using the high power LED as a linear light source, and the polynomial fitting error model was advanced. Then the neural dynamic optimization algorithm was used to solve it. The experimental results show that the optimum algorithm is stable and effective, and the maximum relative error is not more than 2.5% in the third order polynomial fitting with the neural dynamics. The relative error is obviously reduced in comparison with the least square method. By analyzing the responses, the CCD ICX694AL is with linear characteristics in the unsaturated region. However, it has obvious nonlinear effects in the saturated region.

Laser Induced Fluorescence Detection (LIFD) technology is one of the most sensitive means for the microfluidic chip detection in which application field it is important to achieve accurate focus due to the small size of microfluidic chip. A novel LIFD based on liquid lens is proposed to obtain the optimal excitation light and the induced fluorescence collection, which is capable of adjusting focus position in three-dimension automatically according to the information provided by the images. The autofocus algorithm for LIFD is accomplished with the combination of SUSAN operator and the mountain climbing control algorithm. A tunable liquid lens based on LIFD experimental prototype is set up to testify the proposed system and algorithms. Experimental results show that the system can achieve the autofocus of microfluidic chip. The repeatability experimental results indicate that the evaluation of the last picture in each single experiment has a steady feature with a fluctuation of 0.5% and coefficient of variation of 1.5‰. It provides a solution to improving the detection accuracy and a design flexibility of microfluidic chip.

Firstly, the relationship among the interface, the density difference and the interfacial tension is analysed with the Laplace equation. Then, as for the combination and selection of materials, we use a new ionic conductive liquid and four kinds of common insulating liquid so as to obtain four combinations of the conductive liquid and the insulating liquids according to density difference. Finally, by simulation, we calculate the corresponding equations and analyse the imaging quality of the liquid lens with aspheric interface. It is concluded that the imaging quality is improved when the liquid lens with aspheric interface is compared to a liquid lens with spherical interface, but it has no positive correlation with the degree of the asphericity and the imaging quality.

The Labeling Gaussian Mixture Hypothesis Probability Density filter (LGM-PHD) cannot get the spawn targets. Addressing this problem, an improved algorithm is presented. Firstly, the labels are applied to the Gaussian items in the GM-PHD filter to distinguish different targets, and their tracks are determined. After that, in the period of filtering, the track labels between the current step and former step are matched, associated and maintained. Finally, the spawn threshold is used to determine if there are spawn targets or not and determine the number of possible spawn targets, then the labels for Gaussian items of new targets and possible spawn targets are reallocated. The simulation results show, in the situation of existing spawn targets, the improved algorithm has better tracking performance than the LGM-PHD.

To measure the spectral sensitivity of the photosensitive material, an exposure system for the spectral sensitivity detection of the photosensitive material is designed by use of commercial software of Code V and SolidWorks. The structural differences of the optical path are analyzed using line density of grating under diverse incident angle. The analysis result is that the incident angle of 25° and line density of 350 L/mm are optimum parameters of the exposure system. The experimental result in the system shows that the rectangular spectra of the size of 201.5 mm×80 mm can be achieved, the wavelengths cover the range of 360 nm~900 nm, the intensity of spectra can be adjusted in the range of 0.52 mW?10.65 mW when an 18 gray scale level of slices are inserted in the exposure system, and the resolution of spectra is 0.021 nm, which satisfies the requirement of the exposure system.

Traditional sparse representation tracker use simple grayscale characteristics in calculating sparse coefficient, which is easily affected by the heavy occlusions and deformation. To this end, a local adaptive weighting algorithm is put forward to increase degree of differentiation between the candidate targets affected by shade, deformation, etc and not affected by the shade, deformation, etc. In addition, the general sparse representation algorithm use a small number of target templates to build a complete dictionary, which unable to get a better sparse coefficient. Inverse structure sparse representation algorithm, using the candidate target which contains rich target and background features to build a complete dictionary to reconstruct the target template under the condition of the same dimension target template better sparse coefficient can be obtained, is proposed. Experiments show that the proposed algorithm in the small differences between target and background or serious barrier, deformation, can better track the target.

To deal with the defects of Compressive Tracking (CT) in the tracking error and the sample collection, firstly, the predictive vector was introduced to search samples that can direct motion of the target. Then the fan-shaped sampling areas reduced the amount of computation greatly. Furthermore, we could determine complex background or occlusion through comparison of the neighboring target images, and then update the classifier parameters automatically by applying the Bhattacharyya coefficient. Experiment shows that these improvements can avoid the failure of compressive tracking and the adaptive predictive compressive tracking (VACT) is better than the original algorithm (CT) in robustness and speed.

In order to eliminate the imaging gap of focal plane based on optical butting, an error analysis-based algorithm of overlapping pixels was proposed. Firstly, an imaging model for focal plane based on optical butting is established according to the imaging principle of camera. Then, the reason causing imaging gaps is analyzed which combined with telecentric and non-telecentric optical system. Combined with the decrease of MTF in the vignetting distortion caused by the reflector, the formula for computing the numbers of overlapping pixels between adjacent CCDs is obtained. Finally, an example is given. Experimental results show that the proposed algorithm can eliminate the imaging gap.

Catadioptic deep ultraviolet lithography objective took the mirror as active free-form surface to compensate thermal aberration. Mirror in compliance with the requirements of the system optical path had a rectangular aperture. For the purpose of describing the active deformation of mirror, it’s necessary for orthogonal polynomials in rectangle zone to fit the deformation data. For high accuracy of fitting surface deformation, it was needed to obtain higher order polynomial expressions. According to the properties of orthogonal polynomials, rectangular orthogonal polynomials formula was derived and the first 36 polynomial expressions were listed. And the polynomials were applied to fit the deformation data of active free-surface. The results showed that the minimum fitting residual error RMS is 0.024 nm when the deformation is 2 nm~5 nm in RMS. According to the characteristic of fitting residual map, the structure design of the active free surface could be improved.

In order to improve the detection accuracy of dim target in infrared image with complex background, an infrared small target detection algorithm based on multi-feature tightness difference decision coupling improved Top-Hat transform. Firstly, multi-scale structure elements was obtained by segmenting the single structural element, and the gray change map was defined according to the gray difference between the small target and its surrounding background, and the target decision factor was constructed by calculating the mean and variance of this map, so the new Top-Hat transformation was formed by embedding it and multi-scale structure elements into classical Top-Hat transform. Then the multi-feature tightness difference model was established to extract candidate regions that contain the real weak and small targets were extracted by combining the gray intensity, contrast and structure information. Finally, the pipeline filtering pattern was introduced to eliminate the suspicious objects in the candidate region and keep the real dim target. The experimental data show that this algorithm had higher detection precision to completely check out the dim target with better ROC curve under the complex background.mapping; pipeline filtering mode

Aiming at the problem of low detection precision and poor effect of single frame infrared dim target detection algorithm, a novel algorithm of single frame infrared small target detection based on improved Sobel is proposed. According to the characteristic of Sobel operator edge detection, firstly, the image median filtering can improve the dim small target signal-noise ratio, and then use the Sobel operator and the proposed detection template to deal with the image convolution, finally detect the infrared dim small target after median filtering. The effectiveness of the improved algorithm is analyzed theoretically and compared with the centroid detection algorithm and the detection algorithm based on local mean and Top-Hat. Experimental results show that the proposed algorithm is more accurate than other detection algorithms in the experiment, and the detection effect is more obvious.

For the problems of low-resolution and poor visual effect of infrared cloud images, a super-resolution method based on structural group sparse representation was proposed. In consideration of the self similarity of infrared image, a structural group sparse representation model was first established. In the training stage, the Gauss mixture model is used to study the prior information of the image structure group, and then to cluster it, using principal component analysis to get a compact classification dictionary. In the reconstruction phase, the best matching dictionary of each structure group is selected, adaptively reweighted l1-norm sparsity is introduced to effectively obtain sparse coefficient. Experimental results demonstrate that our method can achieve better reconstruction effect in both subjective visual effect and objective evaluation criteria compared with ScSR, Zeyde and NARM methods．

A trajectory measurement of combination vehicles is proposed based on machine vision, which can be used to calculate steering characteristics parameters. Firstly, the land marks composed of several horizontal and vertical identification lines are designed and pasted in experimental field. Secondly, the cameras installed on the vehicle capture the driving video streams, and the intersection coordinate can be calculated by the threshold segmentation and the curve fitting of the horizontal and vertical identification lines. Finally, the global coordinates of the cameras can be calculated by the geometrical deduction, and then the trajectory of the point on the vehicle can be calculated. At the end of the paper, we compare our method with GPS trajectory measurement. The experimental results suggest that the trajectory measurement based on machine vision possesses the advantages of high speed, high precision and high validity and provides a new solution for the measurement of curve possibility for combination vehicles.

Person re-identification, identifying the same person in database from non-overlapping camera views, is a challenging task. To reduce the influence of foggy weather on person re-identification, dark channel prior is used to remove haze from input image first. Then, local maximal occurrence representation and metric learning is used to identify the same person’s images which remove haze. Experimental results show that the recognition rate of haze removed achieving 41.75% rank-1 and 81.26% rank-10, is higher than the recognition rate without haze removing which achieve 35.64 % rank-1and 46.75% rank-10.

To meet the needs of finger interaction system and the recognition of tilt fingertips, a fast and accurate method is described to detect the position of fingertips and real-time recognize tilt fingertip gesture. This method uses YCbCr color-space-segmentation-algorithm to devise complexion roughly. After that, preprocessing regions are segmented carefully by the concept of "proportion of perimeter and acreage" to eliminate interference region other than the hand skin color. In addition, obtain the finger profile by binomial method of least squares fitting algorithm. Besides, the improved convex algorithms are employed to complete the fingertips detection and correct tilt angle. Finally, fingertips are recognized by means of partial scan method. Experimental results demonstrate that the proposed method can realize the recognition refers to easy tilt fingertips of zero to night, and reach the high recognition rate of 95.7%. Apart from this, stable performance are achieved．

To overcome the deficiency of high complexity performance in video emotion recognition, we propose a novel Local Binary Pattern Moment method based on Temporal-Spatial for feature extraction of dual-modality emotion recognition. Firstly, preprocessing is used to obtain the facial expression and posture sequences. Secondly, TSLBPM is utilized to extract the features of the facial expression and posture sequences. The minimum Euclidean distances are selected by calculating the features of the testing sequences and the marked emotion training sets, and they are used as independent evidence to build the Basic Probability Assignment (BPA). Finally, according to the rules of Dempster-Shafer evidence theory, the expression recognition result is obtained by fused BPA. The experimental results on the FABO expression and posture dual-modality emotion database show the Temporal-Spatial Local Binary Pattern Moment feature of the video image can be extracted quickly and the video emotional state can be effectively identified. What’s more, compared with other methods , the experiments have verified the superiority of fusion.

A compact binary feature for RGB-D face description and recognition is proposed. First, different from traditional hand-craft feature, we learned the compact binary feature from the training set using unsupervised learning method. Then, in order to make full use of the contextual information, we use the pixel difference vectors as the input. Finally, considering the smoothness of the depth image, we extract different size of pixel difference vectors from every block of RGB and depth image. This work demonstrates that the proposed method is highly discriminable and is robust to facial occlusion and illumination. And recognition rates are comparatively high on two publicly available RGB-D Kinect database.

In order to realize the automation and high efficiency labeling of round bales, a robot labeling system of round bales has been constructed. A image recognition method of round bale end face under complex background has been studied and a combined recognition algorithm is proposed. Firstly, the radius range of round bale is determined, and the watershed algorithm is used to segment the adhesion image. In order to prevent the leakage and error identification of round bale end face caused by over-segmentation, true roundness threshold and angle threshold are proposed to recognize the over-segmented images further precisely. Then the pixel coordinate centers of round bale end faces are obtained using the ellipse fitting method, and the conversion relationship between the pixel coordinates and world coordinates is completed by the calibration based on the Interpolation of Delaunay Triangles. Finally, the labeling experimental system is set up. Experimental results show that the speed of this labeling system is 20 roots per minute,and the accuracy rate is 99.8%. The system can meet the actual production needs of enterprises, and it can provide certain reference for technology development and practical application of round bale end face automatic labeling at home and abroad.

The flaw in most depth videos is the temporal inconsistency of the depth value of static region, which decreases encoding performance and the quality of rendering. To solve the problem, this paper proposes a method to enhance the temporal consistency of depth video based on the motion intensity of scene. Firstly, we applied block histogram difference (BH) to measure the relative motion intensity of each depth frame, and selected a segment of video as the source for refinement adaptively according to BH value. Secondly, we detected motion region and made a segmentation for each frame of corresponding color video，then refined the depth value which had changed variously in static regions using the accurate temporal consistency information of color video. Finally, we applied the weighted mode temporal filtering on refined depth video to generate well optimized depth video further. Experiment results show that proposed algorithm can save encoding bite rate ranging from 17.48% to 31.75%, while it improves the subjective quality of rendered virtual views.

The automated detection of diabetic retinopathy exudates has great importance for early diagnosis of diabetic retinopathy. Aiming to reduce the residual error caused by ineffective image enhancement and the false detection caused by incomplete removal of interference regions existing in common morphology-based exudates detection methods, an automated method based on mathematical morphology is proposed, which mainly improves the preprocessing of fundus images and the detection of interference regions like optic disc. In the image preprocessing step, the proposed method corrects the brightness of the image in the HSV color space, and then adopts multi-scale top-hat transform to enhance the image. Afterwards, a novel method is used to localize the center of optic disc according to the edge and brightness characteristics of image, and then the optic disc region is segmented by Chan-Vese level set model. Furthermore, other interference regions including bright border and optical artifacts reflection are detected and removed. Finally, the exudates are precisely segmented by background estimation and morphological reconstruction. From the testing results on the new public dataset of e-ophtha EX, the proposed method achieves sensitivity of 91.7% , specificity of 94.6% on the exudate level and sensitivity of 100%, specificity of 88.6% and accuracy of 95.1% on the image level.

Abnormal event detection plays an important role in intelligent video surveillance. A new abnormal behavior detection algorithm is presented by fusing spatiotemporal features. We first extract SI as the feature representation in the spatial domain. Then, by combining the high precision optical flow algorithm with social force model, we calculate the interaction force as the feature representation in the temporal domain. A novel motion feature descriptor, i.e., Histogram of Interaction Force (HOIF) is proposed, which is combined with SI as temporal-spatial features to be input to the Support Vector Machine (SVM) to identify the crowd events. The effectiveness of the proposed algorithm is put to test on the UMN dataset, and the experimental results indicate that the presented method offers more reliable performance than some existing algorithms in terms of accuracy and robustness.

Optimal scanning algorithm produces contour on the image which is very rich in color, the algorithm will adopt non-linear calibration and eliminate the contour of image. Firstly, find out the relationship between the gray level and pixel data of the optimal scanning algorithm, and get the reason of contour with the traditional pulse-width modulation (PWM) gray-scale strategy comparison. Then according to the human eyes have different feelings on different gray error images, the gray error is classified as acceptable error and unacceptable error. Next, rearrange all possible optimal scanning gray weights by MATLAB, and exclude composition comprising an unacceptable error. Therefore, the rest combinations have good linearity. Finally, calculate the data transmission frequency at different display resolutions, experimental results show that the number of contour lines from 13 becomes 0 when the weight values of 8 bits turn to 208:104:52:24:14:8:3:1 for 256 level gray-scale images. The contour lines disappear by OLED-on-silicon microdisplay, image quality is improved, and in ultra-high resolution, system clock frequency is lower than the tradition field scanning method, it has a great advantage in reducing system cost.

For the speckle noise in laser display, a speckle reduction method by the spectral broadening is proposed, which is based on the external cavity semiconductor laser system. Firstly, by appropriately tuning the feedback ratio, the state of coherence collapse is obtained using the optical-feedback-induced self-mixing interference and can be harnessed for the realization of the spectrally broadband emission. Then the optimized broadband laser source is utilized in the speckle measurement system. Significant speckle reduction can be achieved by the spectral broadening due to the low temporal coherence properties of lights. Experimental results show the performance of the proposed approach. The extraordinarily high spectral bandwidth of about 6 nm with multimode emission is achieved in the coherence collapsed state. In free-space reflection geometry, the speckle contrast has been successfully reduced down to 0.049 with this broadband laser source.

For solving the problem that vehicle laser radar cannot measure range and speed simultaneously, we proposed a method of modulating amplitude of transmitting laser by pseudo-random noise code. At the receiving end, we measure target range by calculating correlation function of laser echoes and the local modulating codes, and target speed by analyzing spectrum of beat signal between laser echoes and local reference signal. Firstly, we show the schematic diagram, and discuss the inherent problem existing in the pseudo-random noise code modulated laser radar. Then, we proposed a novel spectrum analysis method for calculating beat signal. Finally, we verified the feasibility of the proposed method by numerical simulations. The results show that maximum speed that can be measured can be extended to 360 km/h without reducing the range measurement performance of current laser radar.