Aiming at the shortcoming of adopting the Mie scattering theory to calculate non-spherical particles, the T-matrix method was used to calculate millimeter-wave extinction performance of randomly oriented cylinder particle under the conditions of different effective volume-equivalent-sphere radius, complex refractive index, and wave-length and aspect ratio. The results show that the extinction efficiency factor peak of cylinder particle enhances with the increasing of the real part of complex refractive index. With the increasing of the virtual part of complex refractive index,the extinction efficiency factor peak of cylinder particle becomes smaller, and its absorption efficiency factor rises. With the increasing of millimeter-wave wavelength, the extinction efficiency factor peak of cylinder particle remains unchanged, and the positions of peak value moves toward larger particles. The change of aspect ratio mainly influences the scattering performance, but has little impact on absorption performance. The conclusions are obtained which can provide solutions for cylinder particle to improve its extinction performance and lay the theoretical foundation to design millimeter-wave interference smokescreen.

Based on the principle of Geometrical Optics (GO), a Flyeye concentrator with uniformity irradiance distribution on the solar cell in Concentrator Photovoltaic (CPV) system is presented. This Flyeye concentrator is composed of four surfaces, that is, refractive surface, mirror surface, freeform surface, and transmission surface. The discrete points of concentrator contour are solved according to edge-ray principle, ray reversibility principle and Fermat’s principle. Finally, the solid model is constructed by using 3D software. The sun light is used to simulate the incident light by ray tracing software. Results show that the proposed Flyeye concentrator is designed well by this method with a uniformity of more than 98%, Concentration Ratio (CR) of more than 31× and average thickness of 9.9 mm.

The 6 n mile visible pilot lamp system is designed based on the theory of atmospheric transmission, photometry and application of geometry optics. LEDs are arranged in linear order and through analyzing the luminescence characteristics of LED, the appropriate system of light collecting, collimating and dodging are designed and optimized so that the design can distribute the light in vertical direction within 10 degrees and in horizontal direction within 3 degrees in a uniform manner in 6 n mile away. The design can be applied to help ships with different drafts to choose the right channel.

With the rapid development of spectral analysis technology in the information age, spectral instrument becomes a preferred access to information in various fields for its performance like high precision, low intrusion and small form factor. With the basic theory of optical design as guidance, echelle grating as key part, high resolution and wide detection wavelength as design target, a small size echelle spectrograph based on Czerny-Turner optical structure is designed. The structure of the system parameters is obtained based on the theoretical computation, and the optical system is simulated by the optical design software Zemax. The design results show that the theoretical resolution of the system, which works in the spectrum range from 200 nm to 800 nm, is better than 0.1 nm.

The ommatidia arranged at the edge have great off-axis aberrations in artificial compound eye structure with wide Field of View (FOV) and small volume. A new compound eye structure is presented, where the ommatidia are designed as freeform prisms. With the light passing through the freeform surface and folding in the ommatidium, the off-axis aberrations are well corrected. Therefore, the image quality of the ommatidia arranged at the edge is improved,and the FOV of the structure is enlarged. To image 180° FOV, this structure contains 5 groups of ommatidia with a particular FOV correspondingly. The model of each group is designed and the ray trace is completed based on the optical simulation software codeV. The results demonstrate that the compound eyes structure has high image quality for the entire 180° FOV. On the one hand, the modulation transformation function MTF is greater than 0.2 at the frequency of 100lp/mm, and on the other hand, the distortion is less than 30%.

As LLE algorithm is apt to be influenced by the deflection angle and the selection of the proximal points, a method of Double-layer LLE (DLLE) human ear recognition is put forward. And a further algorithm combined with Gabor is proposed with the name of GDLLE (Gabor and DLLE). The DLLE algorithm calculate the Euclidean distance between sample point and the center of the sample points first, then take all of the K-class sample points whose Euclidean distance are the minimum as neighbor points of LLE, extract the discriminate features, and classify them by the nearest neighbor classifier finally. The experiments on USTB3 ear image database reveal that the DLLE method proposed in this paper can reduce the impacts of the deflection angle and the selection of the proximal points on LLE algorithm, and recognition rate is further improved with the combination of Gabor wavelet.

In order to realize automatic detection, identification and display lettered characters on the workpiece,automatic recognition system of workpiece is designed and achieved. The designed system can solve the problem that workpiece characters are difficultly recognized in the situation of the light abrupt changing. According to effects of different light sources combinations for character recognition performance, a novel algorithm of feature extraction is proposed to adapt light mutations. Firstly, the character details are enhanced by using homomorphic filtering operator.Then, contour and global features are respectively gained by utilizing adaptive threshold algorithm based on LOG while texture feature map is obtained by using local binary pattern (LBP). Finally, two feature in the same image position are combined by “OR” operator to gain light insensitive fusion feature. In the industrial circumstance, the extensive measurement results demonstrate that the recognition accuracy of detecting system reaches 94.72% and the average execution time is 230 ms. Compared to the SIFT and Bayesian algorithms, our system’s accuracy better satisfied the requirements of automatic recognition in industrial applications and the method has been used for industrial measurement.

The RETE matching algorithm was a classical algorithm in the rule-based reasoning system, but when the number of rules and facts increased in the knowledge base, the generated intermediate match information greatly increased too, resulting to the large time complexity and space complexity, severely reduced the reasoning efficiency. To address this issue, this paper compared several improvement strategies of RETE algorithm, and optimization algorithm was proposed based on RETE cost model. The algorithm can automatically find the optimal RETE topology, reduce intermediate nodes, and greatly reduce RETE algorithm's time complexity and space complexity. The experiment shows that the running cost of optimized RETE algorithm is only about half the time than before optimization, and the reasoning efficiency is improved.

The traditional method based on the histogram of oriented gradients and Support Vector Machine causes large amount of computation. To deal with the problem, a novel method called the contour feature of head-shoulders combined with neural network is proposed. The head-shoulder model is relatively stable and the contour feature can be used as a basis for human identification. There are two main parts in the paper. Firstly, the head-shoulder model was extracted by edge detection and mean shift algorithm. Then Fourier descriptors with PCA dimensionality reduction were calculated according to contours of the head-shoulder model. Combined with neural network classifier, the initial human identification was completed. Secondly, several models of human head-shoulders from aim pictures which have been identified as non-person with RGB hair mode and the mean-shift algorithm were clustered and classified them again. The experiment result shows that, the detection accuracy and speed are improved compared with the conventional algorithms,and it performances well when shelters occur.

An endoscopic ultrasound imaging system based on coded excitation and pulse compression is designed, which can improve the image quality and compensate the deficiency of small echo and high attenuation due to the underdeveloped manufacture technics of the catheter type probe. This system includes probe drive unit, data processing and cache unit, and interactive unit and display unit. In order to suppress the sidelobe caused by the pulse compression,three sidelobe suppression filters are designed. The simulation results show that, for the 4-bit Barker code, the sidelobe can be effectively suppressed by using the mismatch filtering methods. Furthermore, the mainlobe and the peak sidelobe level are able to achieve 0.2 mm and -55 dB with the spike filter method. Finally, an endoscopic ultrasound imaging experimental system was set up, using the target lines as sample, achieved the images with 1 024×1 024 image size and 25 frames per secend speed, which proved the feasibility of the system and the correctness of the above conclusions.

As medical CT images always have blurred edges, low contrast, and noise pollution, etc., a new method based on Short-time Fractional Fourier Transform (STFrFT) and infinite-norm is proposed to detect edges of CT images. The imaginary part of STFrFT domain represents the local feature of non-stationary signal. At first, transform the non-stationary signal to STFrFT domain by short-time fractional Fourier transform. Then, obtain the vector infinite norm of STFrFT domain. Two dimensional time-frequency plane is mapped to a one-dimensional signal which shows the edges of the non-stationary signal. Experiment results demonstrate that the proposed method reveals a remarkable performance in detecting the edges of CT images, as well as weak edges. Compared with the results of differential operators and wavelet transform, the proposed approach performs better in the edge detection of CT images and is robust with noise.

In Liquid Crystal Display (LCD) backlight local dimming technology, to ensure the resulting image quality and luminance, accurate pixel compensation is necessary after each region backlight level been determined. Low pass filter was used to simulate the process of Light-emitting Diode (LED) backlight diffusion to get the LED light spread function and the actual backlight level of each pixel, and then according it to compensate the pixel. On the direct-backlight LCD,specific images were used to verify the accuracy of the method. This method with common test platform has a high precision and a simple system. The experimental validation results show that the average relative error is 6.41% when the number of the direct-backlight region is 32. The diffusion method was used in the dynamic local dimming LCD TV. The energy-saving rate reaches 27.06%, and the screen appears no tailing phenomena.

In order to eliminate or suppress the noise (especially speckle noise) in frequency domain optical coherence tomography (FD-OCT) imaging system which will degrade the images quality, a speckle denoise method based on wave atom threshold de-noising algorithm is presented. Wave atom transform is a novel two-dimensional multi-scale transform,and it meets the curve-wave parabolic proportion scale relationship and anisotropy. Moreover, wave atom is suit to mode at any of the local direction to the axial direction of the anisotropic with model sparse expand. FD-OCT images of the human eye fundus tissue and the skin of finger tip are taken as test image samples. The comparative studies with the proposed method, traditional wavelet threshold algorithm and fast curvelet transform algorithm are given. The results show that the wave atom-based threshold de-noising method can successfully despeckle the noise in OCT images but maintain the detail edges of testing images. It is shown that the proposed method is better than the traditional wavelet threshold method and the fast curvelet transform.

For the problem that image segmentation link of steel tube recognition and counting system is affected easily by uneven illumination, and deficiency that some objects’ mistaken segmentation is caused by segmentation after image’s enhancing, a multi-threshold S-F (Segmentation-Fusion) image segmentation method is proposed. According to improved Otsu multi-threshold method, the morphology algorithm and image fusion technology are applied to extract steel tube objects. Experimental results show that the steel tube image segmentation effect of this method is obviously superior to traditional methods’ under uneven illumination circumstance, and the proposed method is free of the effect of illumination quality and well-adapted, which can be applied in objects recognition of the machine vision field.

The most popular online high-precision Automatic Optical Inspection (AOI) detection system for PCB bare board adopts the inspection method based on contour vectorization and comparison, and arc detection is the key of realizing the contour vectorization. A new fast arc detection method is proposed. Firstly, segmentation primitives are obtained from contours by Douglas-Peucker polyline simplification algorithm and their normal angles are calculated.Then, the sum of difference between normal angles of each two adjacent primitives is calculated, which is used to build the normal angle transformation curve with the position of mid-point of primitive. Finally, DP algorithm is used again to detect the line segments in normal angle transformation curve to implement arc detection. The proposed algorithm simplifies the arc detection into two times line segment detection, and the complexity is O(n). Experimental results show that the proposed algorithm owns higher accuracy and lower cost time in comparison with the same module of Halcon software.