In order to study the anti-disturbance characteristic of the Micro Air Vehicle (MAV) with flexible aerodynamic shape, a hovering MAV with flexible aerodynamic shape was designed, and the finite element method was used to conduct modal analysis of its vertical airway. Then, its vertical lift and hovering were achieved. The preliminary three-dimensional visual inspection of the aircraft was completed in an airflow disturbance experimental device, which is based on binocular stereo vision. The aircraft’s motion trace and morphological changes of the aerodynamic shape under airflow disturbance were recorded using stroboscopic imaging technique, which provided an effective extraction method for studying the unsteady aerodynamic characteristic of the flexible aerodynamic shape and the anti-disturbance mechanism of hovering MAV.

By calculating the interference fringe density of aspheric surface reflecting optical wave and a spherical optical wave, a new method for determining asphericity and the best-fit spherical surface of aspheric surface has been discovered. This method applies computer digital analysis technology to directly compute interference fringe density that the fringes are produced by interfering with aspheric wave and spherical wave in different radius. The spherical surface that is of the least fringe density in the most densities above is best-fit spherical surface of aspheric. The prime advantage of this method is that the radius of the best-fit spherical surface, the asphericity, the density and positions of interference fringe can be well determined not to require any analytical calculation for aspheric surface function.

In order to improve the accuracy and usefulness of CCD-based measurement of Flame Radiant Energy (FRE) and apply it to the boiler combustion control system, a new method of obtaining FRE is proposed. According to the principle of CCD imaging and radiation heat transfer theory, the mechanism model of the FRE is derived. Then based on this model and a large number of experimental data, we can apply nonlinear least square method to identify those unknown parameters, so the function of FRE, RGB and the flame area can be determined. The experimental results show that the maximum relative error of the fitting model is 1.38%, and the correlation coefficient of the true values and the fitting values is 0.9978. The accuracy of the model meets the industrial application requirements with strong engineering practical importance and extensive application prospect.

In order to realize the high precision 3D measurement for the objects with different color under different illumination, the phase composition, phase coding and global phase unwrapping methods are researched. First, three different frequency sinusoidal waveforms are selected to be projected to the object after 3~8 phase shift. The equivalent frequency is acquired by the phase composition. The phase coding according to the three different frequencies is performed based on the equivalent frequency. Through the global phase unwrapping method, the global phase can be acquired, further to compute the high precision 3D coordinate. The precision testing method and comparison between the proposed method and the Gray Code method are introduced at last. Experimental results indicate that the precision of the Gray Code method is 0.04 mm; and the precision of the proposed method is 0.02 mm. The proposed method is high precision and better than the Gray Code method because it can solve the problem of the objects under different illumination.

A fast ellipse center computation algorithm is proposed for optical power-meter to make it highly-accurate and fast, considering that the spots of optical power-meter may be deformed, partial absent or double reflected and the goal of the spot calculation is just to obtain the centroid. The edge of the measured spot is obtained by bi-value morphology. With improved RANSAC method, the false edge points are eliminated for ellipse fitting. And the center of the ellipse, which is also the center of the measured spot, is directly calculated from the fitted coefficient of the ellipse. Experimental result shows that, when applied to optical power-meter, the proposed algorithm is fast and as accurate as the traditional Hough Transform – 0.01D and it is suitable for practical application.

In order to investigate the problem whether resolving power for aberration image of light beam can be improved by increasing the amplification ratio of optical device, the shape distortion of small light beam passing through compressible mixing layer has been researched. The experimental aberration image of small light beam has been obtained using CCD camera imaging system, then a two-dimensional biorthogonal wavelet transform and Canny edge test technique have also been adopted to treat with the aberration image. Using these techniques, the edge of zero-order diffraction spot has been identified clearly. In addition, comparing perimeter of beam with its area, it is more suitable for using area to describe image distortion than for its perimeter.

To improve the automation level of bullet management, a novel bullet counting system based on UV electro-optical technology is designed. The system is composed of the upper PC, main node and three counting nodes. In the counting node, UV radiation of bullet flame is detected by ‘solar blind’ UV detector. Probability of false alarm of the system is low because background radiation of UV is very weak. Bullet signal is converted to digital signal and counted by STM32, a MCU with low power cost. Counting nodes can be extended easily because they have same the structures. Different from counting nodes, main node has a USART interface to communicate with PC. Main node and counting nodes are connected with CAN bus. The bullet counting system based on UV detection technology can detect bursts of bullets. It’s a novel, reliable and accurate method and works well at shooting ranges.

In order to realize rapid detection for the shootings effectively, a cartridge case compared system of the 3Dtopography is established. We compared the indentation of the firing pin which remained on the cartridge cases todetermine whether the cartridge cases fired from the same gun. First, according to the prior knowledge that tail of cartridge case is circle, we calculated the center of mass which is the indentation of firing pin, rotated a cartridge case around the center of circle, and made the connections of center of mass of two firing pins to center of circle parallel. Then, the cross-correlation measure was calculated based on voxel. In order to improve the matching accuracy, iterative algorithm on adaptive variable coefficients was proposed for calculation of cross-correlation. Experimental results indicate that when the two cartridges are fired from the same gun, the maximum of cross-correlation measure is improved more than 30%, and the average is more than 12%. The result of comparison is stable and reliable; which can satisfy the requirement of the bullets comparison system.

Automatic identification technology of digital display instrument has significant meaning in improving its using efficiency and environment adaptability. As to strokes missing (or character missing) in the image acquisition process, a digital automatic recognition algorithm is put forward. Two-dimensional entropy threshold is introduced in image binarization segmentation to reduce the influence of low or uneven illumination. At the same time, multiple segmentation and coding matching method is used to identify the missing characters and numbers. After a variety of instrument recognition experiments, the results show that this method can meet the time and accuracy requirements in digital display instruments, and has better adaptability to situation like image tilt, zoom and so on.

The influence and origin of quadrature phase shift on the measured rectilinear displacement measurement error was investigated using a homodyne laser interferometer with a quadrature output. Common nonlinearities, including zero-shift, different amplifier gains and phase shift are three typical errors for displacement measurements. A simple method was used to determine and correct the zero shift and unequal gains, and the influence of the phase-shift was described in detailed experimentally, including the effective way to correct phase error. The effectiveness of the method was validated by comparing the displacement error before and after phase shift error-corrected. Moreover, the method can be applied to assess standard machine and adjust interferometer.

A parameter estimator of the heavy-tailed Rayleigh prior model was proposed for the multi-look averaged amplitude Synthetic Aperture Radar (SAR) image. First, the multiplicative model of SAR image was changed into the additive one according to the logarithmic transformation. Then, the expectation and variance of the log-transformed original image were derived using the theory of negative-order moments. Lastly, the expectation and variance of the log-transformed speckle were obtained based on the Gaussian approximation of the log-transformed speckle. The proposed estimator is easy to be implemented in the form of analytical expressions. Parameter estimation experiments on Monte Carlo simulations and real SAR images demonstrate that the proposed estimator can efficiently estimate the parameters of the heavy-tailed Rayleigh prior model from the multi-look averaged amplitude SAR image.

The similarity metrics of voxel-based image registration are usually Normalized Mutual Information (NMI), and its good registration properties can make images to achieve sub-voxel registration. However, the local extrema and the narrow capture range of global maximum for multi-modal images are easy to cause registration to fail. The structural similarity function has been used to assess image quality. It may reflect the difference of the visual effect and structural information between images, and is associated with the statistical distribution of the voxel gray. When the spatial location between images is changed, the structural similarity also will be changed. We modify this function to make it available for image registration. Simulation results demonstrate that the registration curves of this modified structural similarity (MSSIM) used as a new registration metric show a good convex upward function, and have no significant local extrema. Moreover, the global maximum is located exactly. Especially, the capture range of the global maximum is wide, and hence its robustness is strong. In addition, it is sensitive to strong noise and its operation speed is slow. The metric MSSIM can achieve sub-pixel registration accuracy for three-dimensional image registration even if a 10-parameter affine transformation.

A Fully Affine Invariant Feature (FAIF) detector based on affine Gaussian scale-space has been put forward. Affine Gaussian scale-space is difficult to be built up. FAIF transforms the affine Gaussian scale-space into scale space to cope with the complexity of affine Gaussian scale-space construction. Covariance matrix of an image patch is used to measure the isotropy of the patch. Anisotropic patches are transformed into isotropic ones by rotating and squeezing. Finally, the fully affine invariant key points are detected on isotropic patches. Experimental results indicate that FAIF has the ability to cope with large view angle and scale changes. Moreover, sufficient matches have been detected by FAIF even in images of 3D scenes. Compared with the state- of- the- art, FAIF is the best.

For Chang’e-3 landing mission in the 2nd stage of Chang’e project, high-resolution images were necessary. So a lunar satellite images super-resolution reconstruction algorithm via using compressed sensing was presented. The images from Apollo project, CE-1, CE-2 and tests in the 2nd stage of Chang’e project were used in extracting patches and the dictionaries Ah and Al were built with joint training. Through solving optimization problem via Regularized Orthogonal Matching Pursuit algorithm, the sparse representation for each low-resolution image patch with respect to Al was obtained, and the representation coefficients were applied to Ah in order to generate the corresponding high-resolution image patch. At the end of experiment, the high-resolution image which satisfied the reconstruction constraint was achieved. Numerical experiments demonstrated the effectiveness of the proposed algorithm. Moreover, the proposed algorithm outperforms traditional methods in terms of visual quality, the Peak Signal to Noise Ratio (PSNR) and Root Mean Square Error (RMSE).

Through discussion of the characteristics of complex diffusion, a novel adaptive complex diffusion algorithm based on diffusion parameter and iterative step is proposed to overcome the disadvantages that the diffusion parameter k and iterative step Δt are constant. This new algorithm used the gradient information of image’s real part to control the diffusion strength to realize different diffusion velocity in different gradient region, and increased the iterative step with increasing iteration number, so this adaptive complex diffusion can achieve better denoising effect in less diffusion time. Numerical experiments results show that this algorithm can remove the noise while preserving more image details and gain higher peak signal noise ratio in a shorter time.

Non-negative Matrix Factorization (NMF) consists in factorizing a nonnegative data matrix by the product of two-rank nonnegative matrixes. It has been successfully applied as data analysis technique in hyperspectral unmixing. However, the direct application of the standard NMF algorithm to the decomposition of mixed pixels will result to the problem of local minimum and slow convergence. The basic theory of NMF algorithm is introduced firstly. Then, the endmember matrix was initialized by the automatic morphological endmember extraction method, so the endmembers selected would be close to the real endmembers. The NMF algorithm is extended by incorporating the nonnegativity and sparseness constraining to unmix hyperspectral data and make sure the error is as small as possible. The measurement of sparseness is implemented by non-smooth NMF and NMF with sparseness constraints algorithms respectively. The optimization results is got by continuous iterative. The repeated iterative calculation is included in one iterative more than once. The experimental result proves the efficiency of the approach.

Image compression for low bit rates will loss edge characteristic seriously, which will influence subsequent recognition and comprehension to the image. Hence, how to preserve the edges of the image when compressing the image becomes an urgent problem. In order to solve this problem, an image compression algorithm is proposed. This algorithm first extracted the edges of the image and encoded the edges based on the compression of wavelet Contourlet transform, and then reconstructed the edge preserved image by the edge preservation rule. The experiment results demonstrated that our algorithm could preserve the edges more completely, and keep the value of image contrast index higher. As a result, our algorithm is propitious to enhance the degree of comprehending the reconstructed image content under low bit rates reconstruction condition.

In the wavefront coding researches, defocus limits and the parameters of the cubic phase plates determine thequality of the images and difficulty of restoration. We used the accurate Modulation Transfer Function (MTF) of thewavefront coding imaging system which was obtained through Wigner function to construct a novel function and discussed the relationships between defocus paramount and surface parameters. Through this method, we got the theoretical defocusing limits and the practical defocusing ranges for different systems. At the same time, by comparing the MTF function of defocus system with that of focus system, we discussed the global consistency of the defocusing MTF and tested the selection of surface parameters for different requirements of depth of focus. Finally, CODE V and mathematic calculation were used to examine the defocus limits of an optical system respectively. Meanwhile, the surface parameter for a certain system was offered.

A novel optical encryption system based on non coherent superposition is proposed. In this system, the original image is hidden into three Phase Only Masks (POMs). One of the phase masks is generated by computer while another two are obtained analytically. For decryption, the diffraction filed of the three POMs is superposed by utilizing the beam splitters. The intensity of the superposition field, which is exactly the original image, could be recorded by CCD camera. Compared with the earlier interference-based method, our approach is simple and easy to be realized by optical elements. Moreover, the silhouette problem existing in the earlier method with two POMs can be resolved in our method, thus the security of the method is further improved. Simulation results are presented to verify the validity of the proposed approach.

Microscanning is an effective method for reducing spectrum aliasing and raising spatial resolution. It is widely used in IRFPA imaging system. Based on introducing microscanning imaging technique, the imaging process and the direct reconstructed algorithm of 2×2 microscanning are analyzed. A weighted interpolated reconstruction algorithm based on pixel-correlation is proposed and the interpolated image is corrected with microscanning low-resolution images. Experiments indicate that RPSNR of new algorithm reconstructed image is 1.82 dB higher than the direct reconstructed image and eMSE is 34% lower, so it can effectively improve the microscanning imaging effects.

A dynamic global backlight dimming algorithm is proposed based on the analysis of the Cumulative Distribution Function (CDF) of 12 local images histogram. For each image, the proposed algorithm decided Backlight Luminance (BL) dynamically from maximum CDF luminance ratio of 12 local regions instead of the value of the whole image. Ratios of backlight luminance are defined between 0.6 and 0.9, in order to preserve quality of images and decrease power consumption. The experimental results show that the improved method significantly improves the image quality of Liquid Crystal Display (LCD) and substantially reduces the power consumption, which also meets requirement of the PSNR rule.

Through the analysis of the disadvantages of early ear recognition methods, Locally Linear Embedding (LLE) algorithm is employed to deal with ear recognition. However, LLE algorithm depends heavily on the number K of close neighbors points. Usually, K is large in order to get a good dimensionality reduction effect, while the computational complexity increases. In order to weaken the algorithm’s dependence on K, the LLE algorithm was improved, which made the sample set be more evenly distributed. Even if a smaller number K was selected, the improved LLE algorithm could still get a good dimension reduction effect, and expand the range of K to some extent. Ear recognition contrast experiments of the improved LLE algorithm and the original LLE algorithm show that the improved LLE can obtain a higher recognition rate, thus verify the efficiency of the improved LLE algorithm.

In order to address the problem that Local Phase Quantization (LPQ) method couldn’t discriminate among the sub-patterns based on their different contribution when describing the image feature. A method for face recognition called as Adaptively Weighted Local Phase Quantization (AWLPQ) is proposed. At first, the face images are divided into several sub-images and the feature fetch is based on the LPQ method. And then proposed algorithm employs an adaptively weighting map to weight the sub-patterns based on their information entropy which is defined as the contribution to describe the whole face images. Experiments on the FERET face database show that the proposed method is effective. In addition, in order to solve the problem of high dimension in AWLPQ, Neighbor Preserving Embedding (NPE) is applied for dimension reduction. The experimental results indicate that the method gains both relative robustness and good recognition accuracy.

Since the Pyramid Histogram of Oriented Gradients (PHOG) has a poor performance in describing the shape of faces with noise or abrupt intensity changes, a face recognition algorithm based on improved PHOG is proposed. Firstly, we improve the PHOG feature used in the clear outline face recognition to describe the further refinement of the local structure of the face. Then the noise is restrained through the improved normalized method. Finally, KPCA is used to project improved PHOG feature into the more expressive kernel space to further select the feature with strong descriptive ability and the nearest method is adopted for classification. The experimental results show that the characteristics combined improved PHOG with KPCA has obvious advantages in face recognition, and the experimental results on ORL, FERET and YALE face database can achieve high face recognition rate up to 98%，95% and 98.67%. It is shown that theproposed method has better effect on noise suppression and improving the recognition rate.