Three methods are proposed to enhance the targets in infrared images based on the analysis of the changes of the grey degree′s local mean contrast, local variance and local three dimonsional shape. Then the seed point is determined at the most possible location of the target by searching the point with the local maximum energy in the enhanced image. The seeded-region-grow algorithm is utilized to realize the separation of the target and the background and to extract the target from the background finally. Experimental results on a lot of real IR images with ocean and sky scenes show its effectiveness in detecting dot and spot targets in a single frame.

To detect moving point targets in low SNR, a new detection method——Multistage Velocity Matching Algorithm is presented. In the algorithm, target energy is accumulated by use of Velocity Matching Filter recursively. Compared with other detection methods, the new method′s computation is simple, and detection velocity is fast. Theory and performance of the method are analyzed, and simulation experiment proves that it is an effective method.

Most ATR methods deal with the single target. Actually, there are some targets arranged by some rules in spatial domain, which can be called "targets′ arrays". To recognize them, the spatial information among targets is used to make up the uncertainty of targets′ own features. A novel recognition approach of blob targets′ array at low resolution context (remote sensing images, infrared images, multi-spectrum images etc.) is presented. The method firstly utilizes the ARG model to customize the targets′ arrays and combine the spatial relationships of array nodes and the targets′ other simple attributes. By estimating the probability density distribution of the spatial relationships among targets and their own features, the arrays can be recognized. The method is different from the classic ATR method which is based on ARG model matching. Instead, the method extracts the probability features of spatial information from ARG model directly to recognize. The experiments show that the method is effective.

A method based on image processing to simulate far field laser spots is put forward and realized, which simulates a lot of spots under various conditions by processing the far field spot image under some certain conditions. Something imperfect are analyzed and pointed out, and the corresponding solutions are provided.

Firstly, the method of IR image generation by computer is reviewed. Then the specific simulation methods of simulating the target, the background, the atmospheric effect and the sensor effect are analyzed and compared, and the limitations are discussed. The advanced technologies of dynamic IR scene realization are introduced, such as the suspended and membrane resister technology and the laser diode technology. The developing tendency of dynamic IR scene realization is pointed out. Finally, the developing trend of the technique is expected.

The principles of the microwave radiometric imaging system are discussed. For there are many factors which influence microwave radiometric imaging quality, the relationships among the radiometer spatial resolution,the radiometer sensitivity, the sampling interval, the antenna scanning speed and the integration time are analyzed. Then some corresponding experiments are made to examine the influence of two main factors: the integration time and the sampling interval. And the result proves that these factors restrict and effect each other. When the parameters of microwave radiometric imaging system are chosen, different situation must be considered. Thus, perfect image can be obtained.

A hybrid RBF training method based on immune algorithm and least square method is proposed and applied in radar antenna scanning-style recognition system. By introducing the diversity control and immune memory mechanism in immune algorithm, the efficiency of the network training algorithm is improved and the immature convergence problem is restrained. Computer simulations demonstrate that the recognition system based on the proposed RBF network has a high recognition rate with concise structure.

To make a space-borne Fourier transform infrared spectrometer as an effective load on the FY 3 and FY4 satellites, the mechanical scanning methods of the Fourier Transform Spectrometers in the space-borne projects are compared and discussed, then a new method is put forward. It is based on the principle that when light passes through two parallel plane mirrors the optical path length depends on the incident angle. The new method correlates the optical path length changes between the two arms of the interferometer. This method requires only plane mirrors, and it is immune to tilt, wearing etc. Therefore this method is suitable for our project.

The light scattering characteristics of microsphere are studied by use of Mie scattering theory. The characteristics of polarization and distribution of scattered light and the effects of the radius of microsphere on light scattering and absorbing are also analyzed. The results show that the radius of microsphere has significant effect on light scattering and absorbing and the results may provide a reference for choosing appropriate micro-optical elements in practical experiment and further study of microsphere lenses arrays.

The computation of two-dimensional Fourier transform for point spread function of diffraction limited pupil can be transformed to the computation of one-dimensional Fourier and Hankel transform by use of the relationship between exponential function and Bessel function and property of an expansion of a optical symmetrical pupil function into azimuthal Fourier series. The detailed numerical procedure based on one-dimensional digital algorithm is completed on MATLAB. In comparison with the classic two-dimensional Fourier transform algorithm, the one-dimensional method of Fourier and Hankel transform can obtain intensity samples along arbitrary radial direction at the image plane. And compared with Fortran, Basic and C language etc., the MATLAB program has advantages of succinct language, great reductions of demand of mathematic basis and computer language, and it is a efficient tool for scientific and engineering research.

Active vibration-resistant techniques in interferometer are applied to eliminate the effects of environmental disturb, the mechano-electronic feedback method is one of the low-cost way to fulfill these techniques. The detection of vibrational signal is the key part of this method, in this case, a single photodiode is used to detect optical phase of the interference fields in order to obtain the vibrational signal. The optoelectronic signal concerning with the vibration is digitalized and analyzed by DSP, then a feedback voltage is produced to PZT phase shifter which can compensate the change of optical path difference due to vibration. The principles of each part of this vibration-resistant system are described including the circuit design. Experimental results indicate that this method is an effective way to control the vibration with amplitude-frequency product less than 100 wave·Hz

The designing formula of multilevel diffractive microlenses is presented. Body-of-revolutionary finite-difference time-domain method (BOR FDTD) is illustrated. The multilevel diffractive microlenses are analyzed by use of BOR FDTD. The effects of changing design parameters (including focal length, incident wavelength, refractive index of microlens material) on multilevel diffractive microlens are investigated. The results show that changing design parameters have effects on diffractive efficiency, airy diffraction disc radius and maximum electric magnitude of microlens. The qualitative analyses of microlenses are consistence with some of numerical results.

The polarization variation of a reflector light after a bundle of linear polarization light is incident on the ideal cube corner retroreflector is analyzed. Diffraction distribution about the orthogonal polarization quantities of a reflector light is calculated with Fresnel diffraction theory and the correlative diffraction diagram is drawn. Meanwhile, the comparision with the diffraction diagram without considering the polarization effect proves that they are different obviously. At the same time, the result provides the grounds for the theoretical design of cube corner reflector with different usage requirement.

Based on 3D IR scene, a virtual IR imaging guiding system model is established. The model consists of three modules, which are imaging system, signal processing and movement control. Analysis of the three modules is performed with emphasis on the movement relation and data exchange between the virtual system and 3D scene, and mathematic models are given. By use of time amplifying method, a way to resolve the conflict between the real-time requirement and simulation reality is discussed. At the end, simulation experiments are performed and the results show that this model is practical.

In order to improve CCD performance, the noise composition in CCD image is analyzed according to the operating principle of CCD, and CCD noise is categorized. Correspending measures are introduced to deal with different noise. Signal processing schemes of double-correlated sampling,dual slope integration and clamp sample are given to eliminate readout noise in this paper. As a result, the SNR of CCD is improved.

A design method for a new type of infrared remote control multi-channel switch system is introduced and the basic frame and operating principle of the system is discussed. A pair of encoder and decoder MC145206/145207 is chosen to encode and decode the instruction signal, and the instruction signal is carried to siglechip to realize multi-channel control by use of infrared remote control. An array of diode switches in combinatorial logic is chosen to encode the switch keys in the transmitting system to control the objects selectively.Applying singlechip in the receiving system, this method can make the best of singlechip′s resource of software and hardware and character the system flexible control, convenient extension and realized multi-channel,multi-function control.

The application of the neural network to remote sensing infrared chemical detection is introduced. The possibility that the back-propagation artificial neural network (BP-ANN) classifier is implemented to discriminate infrared spectra is approached. A data set of DMMP (dimethyl methylphosphonate) infrared spectra is used to perform the actual training and discrimination performence prediction of this classifier. The training results show that 95% samples in the data set can be correctly classified under certain conditions. The prediction results show that the neural network classifier which has been properly trained has possessed relatively high discrimination ability, and the discrimination rates more than 70% can be obtained from such classifiers.

A new type of infrared jammer testing system is introduced. Pyroelectric detector and fast respond infrared radiometer are used in the system to carry on the measurement, and effective data processing method is adopted to improve the test accuracy. Then the hardware and common principle of the system are analyzed, the demarcating and measuring principle of the system are discussed. A better effect in practical application is obtained.

The thermal effects of LD end-pumped Nd:YAG high-repetition-rate solid laser during the optical pumping are the important factor which affects the laser output characteristic and the system integrated performance. Starting with the thermal conduction function, the numerical value of temperature steady distributing in the laser crystal is found by the finite element method. And thermal lens effect is counted and analyzed. Then the separate parameters design of thermal insensitive cavity is processed to compensate the thermal lens effect by the propagation-circle and transform-circle graphic analysis and design method. The study establishes the base for optimizing the design of laser to improve the stabilization of laser output.

Since 1960s, the technological progress of low light level night vision technology has always been evaluated by "generation". According to usual understanding, the second generation is better than the first generation, and the third generation is superior to the second generation. In 20th century, during the developments of image intensifiers for low light level night vision (LLLNV), the first generation fiber optical cascade coupled image intensifier in 1960s, the second generation image intensifier with microchannel plate (MCP) in 1970s, the third generation image intensifier with GaAs negative electron affinity (NEA) photocathode in 1980s, appeared one and another. The new generation image intensifier or the fourth generation image intensifier has been continuously explored by scientists engaged in night vision technology. But, for the image intensifiers, what is the fourth generation, what can be called the fourth generation, it has some arguments in the night vision academic circle. The problems are as the following : whether or not it is appropriate and comprehensive for the image intensifiers to be divided by“generation”and evaluated also by“generation”;what is exactly when we evaluate an image intensifier by useof the performance or the technology ;and what is better and more comprehensive for expressing the performance of an image intensifier. All these problems have been paid attention by night vision academic circle. The present paper shows that the overall performance of low light level image intensifiers should be evaluated by the“figure of merit”, rather than by the concept of“generation”.

A fast and exact approach is presented to extract the corners of eyes on human faces automatically. It can locate the positions of eyes with its near and far corners exactly with over 98% accuracy. And it provises the basis of extracting the vital feature points on human face automatically and improves the accuracy of face recognition. Then, applying the results to the pretreatment of the face images, a method using the distance of two far corners of eyes to normalize the face images is presented. It could get more exacter and steadier outcomes than the method commonly using the distance of two irises.

The problem of enhancing and detecting target in a sequence of images recorded by a long-range infrared sensor is investigated. To highlight the difference between warm object and background, a new algorithm of infrared image enhancement based on local textural features is presented. To improve the speed and accuracy of target detection, the relative information among sequential images is used to detect and track the target at the same time. Experimental results show that the enhancing algorithm can highlight the warm objects and depress the background and improve the speed and accuracy of target detection.

An automated FFT-based technique is proposed for registration of remote sensing images having different spatial resolution and relative rotation and shift. The proposed method is an extension of the phase correlation technique. The main characteristic of the algorithm is to align two images automatically without requiring either control points or sensor′s parameters. Fourier scaling properties and rotational properties are used to find scale and rotational movement by log-polar coordinate transform. The phase correlation technique determines the translation movement. Experimental results show that satisfactory effect has been obtained by applying the (method.)

The full-search block-matching (FSBM) algorithm is an ordinary motion estimation method in image sequence stability. In traditional FSBM algorithm, every pair of pixels corresponding to the current and reference frame makes equal contribution to the matching result. Therefore, the performance of the traditional FSBM algorithm will tend to be influenced by various noise in the being stabilized image sequence. Aiming at the shortcoming, an improved FSBM algorithm is proposed. In the new matching rule, every pair of pixels is classified as matching pixel pair (set value as 1) or mismatching pixel pair (set value as 0). A threshold is selected to perform the above classification. The mismatching pixel pairs make no contribution to the result. Now the matching pixel pairs make equal contribution to the matching result as before, then multiply each by a weight that has relation with the position of the matching pixel pair and sum them up. The experiments show that this new method is promising.

The 3D computer vision is accomplished by a new vision system based on the combination of one camera and one mirror instead of binocular vision system. It is proved theoretically that if the mirror is placed in the certain conditions, the images that can be used to accomplish 3D stereo vision restoration is obtained from the combination. The point′s world-coordinates are calculated theoretically by use of the images, supposing the interior and exterior parameters of the combination are known. The calibration, parameters calculation and point′s 3D stereo restoration of the combination are also deduced theoretically. It is also proved that the essence of the combination′s 3D restoration is identical to the binocular vision system. Experiments are presented to contrast the performances of the two methods. The results show that the accuracies are nearly equal.