To realize testing of surface crack defect in X80 pipeline, an entirely non-contact and non-destructive detection method of using pulsed laser to excite ultrasound and using continuous laser to receive ultrasound is proposed, and laser ultrasonic testing system is built. By doing the A-scan and B-scan of pipe surface, adopting amplitude normalization, filtering, differential and other methods of signal processing to extract wave signal, propagation characters of ultrasonic surface wave in the pipe are explored. The result show that the reflection occurs in a part of the ultrasonic surface wave when encountering crack defect, and another part propagates along defect edge to the crack bottom, generating mode conversion and changing into transverse wave to propagate to the pipe inwall. Then, it is reflected back again to the crack bottom, generating again the mode conversion and changing into surface wave to propagate along the defect edge. Defect characteristics can be obviously identified, the maximum error of location and depth is 4%, and the resolution is 0.01 mm. Experiment results show that the laser ultrasonic testing system can effectively detect surface crack in the pipe.

In order to detect the surface defects of TFT-LCD panel, an Automated Optical Inspection (AOI) experimental system was built and a defect detection algorithm was proposed based on two-dimensional (2D) DFT. The lines associated with high-energy frequency components in the spectrum that represented linear texture could be detected by Hough Transform (HT), and the corresponding neighborhood of high-energy frequency components were set to zero, and then a spatial domain image was reconstructed using the two-dimensional IDFT to remove the background of directional linear texture. Moreover, a simple threshold algorithm was used to discriminate the defects from the background. When calculating the slope-angle of high-energy frequency components in Fourier domain, two key points of how to select the high-energy threshold in frequency domain and how to extract the peaks from the angle histogram automatically were solved by using mathematical statistics. To preserve the defect, whose direction is near to directional-textures in the reconstructed image, all frequency components lying within a small annulus with the DC as the center in Fourier domain must be retained. The experiments on a variety of surface defects such as fibers, stains and scratches in TFT -LCD panel testified the effectiveness and robustness of the proposed method.

In order to get the rotation angle accurately and quickly by using the image registration, the Binary Robust Invariant Scalable Keypoints (BRISK) are improved. Firstly, we can get the origin key-points by BRISK, and set a threshold to the Hamming distance between the key-points to pick out the matching pairs. Afterwards, further screen the matching pairs by the minimum Hamming distance between them. Finally, combining the matching points respectively with OTSU, the affine matrix and feature points of the slope of the straight line the median value, the rotation angle and the time it costs are measured. Experiments show that the secondary hamming distance to get the matching pairs can improve the accuracy of the matching point, the error of rotation angle measured by these three methods is not more than 1 degree, and the time they cost is less than 1 second. Besides, it keeps certain robustness to noise, loss of outline, illumination variation and so on.

By “fast and cold tapered technology”, a home-made Microstructure Fiber (MF) is tapered to 8 mm, 10 mm tapered length while keeping d/Λ unchanged. Numerical simulations by multi-pole method show that untapered MF has a single Zero Dispersion Wavelength (ZDW) at 1 129 nm, while after tapering an additional ZDW appears at tapered waist. The two ZDWs of 8 mm and 10 mm-tapered-MF locate at 806 nm/2 456 nm and 637 nm/1 164 nm, respectively. When pumped by ultrafast pulse with center wavelength at 800 nm and average energy of 0.45 W, 8 mm-tapered-MF generates supercontinuum at range of 378 nm～1 632 nm and 1 777 nm～2 450 nm with 20 dB flatness. In 10 mm-tapered-MF, redshift of Raman soliton and supercontinuum is hindered by its second ZDW at 1 164 nm. However, the blue shift efficiency of bump energy is enhanced. When pump power reaches 0.5 W, up to 70.5% of pump energy is unconverted to 395 nm~ 475 nm.

Since the PMF’s polarization axis alignment method based on image processing of PMF’s end, has been researched and applied more and more, manual focus predominates in the image-processing alignment system. Consequently, an autofocus system is built, and an autofocus algorithm based on the imaging characteristic of alignment system is presented. The variance function is selected to evaluate image definition, with a focusing window selection method based on local variance maximum to enhance its sensitivity. Additionally, a comprehensive strategy based on iterative fitting is designed to search for image focus. The experiment results indicate that this algorithm can realize the focusing of PMF and its gasket both fast and stably.

In order to realize the bending stress of gear measurement, based on the analysis of Fiber Bragg Grating (FBG) strain sensing model, an optical fiber dynamic testing system design scheme of gear bending stress was put forward. And then the coupled mode theory was used to analyze the spectrum characteristics of FBG, according to which the optical fiber grating sensor parameter was determined. Based on gear bending stress distribution rule, the set of optical fiber probe and its transmission way were designed. What’s more, the fiber grating demodulation system was designed based on F-P filter. The simulation results verified the effectiveness and rationality of demodulation system. Finally, verified by theoretical analysis, the measurement scheme has the advantages of simple structure, resistance to electromagnetic interference, suitable for distributed measurement, etc.

A novel method to control the pulsed light extinction ratio automatically, which combines a double electrode series structure EOM with “scanning & step-tracking” algorithm, is proposed. It aims to address the problem that extinction ratio of pulsed light is affected by bias voltage shifting in a Brillouin Optical-fiber Time-domain Reflectometry (BOTDR) based on distributed optical fiber sensing system. Based on the method, some experiments are conducted and important conclusions are deduced. The results show that it could keep the extinction ratio of pulsed light above 50 dB steadily. The fluctuation of extinction ratio is only 1.3 dB, maximum value of extinction ratio is 52.4 dB and the minimum value is 51.1 dB. It is indicated that the proposed method will have important significance for enhancing the performance of pulsed light source in BOTDR optical fiber sensing system and high power laser driving device.

In order to improve the reliability of a two-axis Fast Steering Mirror (FSM) system with minimum hardware consumption, a fault detection method based on observer was developed. The dynamics model of the two-axis FSM was established firstly, and then the state-space form of the FSM was adopted. An observer for fault detection was designed based on the state-space form. Then the residual vector could be generated by comparing the output of the system with its estimation provided by the observer. Finally, the residual evaluation function was adopted, and a fault can be detected by comparing the residual evaluation function with a threshold. Experimental studies on a prototype system show that the sensor fault signal can be detected by the residual vectors timely and accurately. Meanwhile, the consumption of the hardware and space is decreased.

In the development of the laser communication ground optical terminal, the dynamic stability of secondary mirror structure system plays an important role in ensuring the high image quality and stable optical axis orientation. By means of Finite Element Method (FEM), the dynamic stability of secondary mirror structure system has been analyzed by applying variable offset distances and preloads. The natural frequency of secondary mirror structure system increases with the offset distances, presenting a maximum around 20mm at variable preloads. While, the curvilinear relationship between natural frequency and preloads shows a monotonic increasing function. The results provide a theoretical reference to improve the torsional stiffness and the dynamic stability of secondary mirror structure system, which helps to optimize the structure design of secondary mirror system of laser communication ground optical terminal.

To solve the problem of poor robustness and low effectiveness of target tracking in complex scenes, a target tracking algorithm based on adaptive multi-feature fusion in tracking-by-detection framework is proposed. Features are extracted from the sub-images extracted by dense sampling, and the target appearance models are established respectively. The response of each model is obtained with regularized least squares classifier. The final response is achieved by weighted sums of the responses, in which the weights are updated by solving a regression equation. It helps to obtain accurate and stable detection scores by enhancing local discrimination. Experimental results show that the algorithm outperforms other state-of-the-art tracking algorithms in tracking accuracy and robustness in most complex scenes.

Phase diversity (PD) can not only be used as wave-front sensor but also as image restoration technique. However, its computations have been perceived as being too burdensome and it is not to satisfy the real-time requirement. On the principle of minimum data communication between CPU and GPU, we analyze the process of PD algorithm and the experiment result shows that the part of optimization is appropriate to be realized on GPU. Firstly, the process of PD is analyzed and then task partition is implemented on dual GPUs. Under the experiment condition of wavefront aberration of PV=2λ, after 50 iterations, the execution time of single GPU and dual GPUs are 53 ms and 45 ms respectively for image size of 256 pixels×256 pixels.

Principal Component Analysis (PCA) can effectively extract global features from images and has advantages of dimension reduction. During the dimension reduction process, because of the comparatively concentration of eigenvalues, the dimension is still larger than the best. To solve this problem, this paper presents the optimal-sample PCA (OS-PCA) for dimension reduction. By choosing the training samples and optimizing the covariance matrix, OS-PCA achieves the purpose of further dimension reduction. Because Discrete Cosine Transform (DCT) has robustness of light, as well as Local Binary Pattern (LBP) is effective in describing local texture features, the paper combines DCT and LBP features to make up for the limitations of OS-PCA in facial expression representation. In order to utilize the advantages of collaboration features and classifiers, this paper constructs a facial expression recognition model, which is based on three layers of the optimal integration of multiple classifiers. Firstly, facial images are preprocessed. This step includes the detection of face from images and normalization. Then the OS-PCA, DCT and LBP features are delivered into the model. Finally, based on the best match combination between single classifier and single feature, the model completes the optimal integration of multiple features and multiple classifiers. Via voting mechanism, the model makes adaptive decisions for images that are still different to get the final recognition result. Experiments show that OS-PCA is more effective than PCA in dimension reduction. On the JAFFE and CK database, recognition rates are higher than 95% and 96%, and the proposed model shows brilliant time performance.

To overcome the problem that traditional Iterative Closest Point (ICP) can’t properly handle the situation of outliers, noise and missing data appeared in the process of image processing which would result in low accuracy of image registration. An enhanced Sparse Iterative Closest Point (SICP) to register the 3D point cloud is proposed, and sparse ICP address these problems by formulating the registration optimization using sparsity inducing norms. What’s more, a fast head segmentation algorithm was proposed to segment the user’s head in depth image. Based on the proposed fast face segmentation algorithm and sparse ICP, a new 3D face modeling system is put forward. The experimental results demonstrate the effectiveness of the proposed algorithm