To solve the problem of tracking drifts or fail, a robust objects tracking algorithm based on geometric blur is proposed within the framework of online learning. Under the tracking-detection-learning mechanism, Lucas-Kanade algorithm is used to obtain the rough tracking estimation of the target. Based on the idea of geometric blur matching instead of traditional detection methods, the tracking drift is efficiently corrected. Then integrator is designed to compare the similarities between the previous frame and the results of the tracker and the detector. Their confidences are obtained by calculating normalized correlative coefficients between positive and negative samples and the detected region. An online learning is then developed to use the current result to update the tracker and the detector. Experimental results show that when applied to the fact moving target tracking under the condition of high background similarity, the proposed method performs well and outperforms other state-of-the-art methods with higher position accuracy.

This paper analyzed the impact of the atmosphere back-scattering on the active detecting system and built a corresponding physical model based on geometrical optics. A simulated calculation of several related factors was implemented and then the illation of the interfering ability of the back-scattering was validated by a long-distance detecting experiment outdoor. An appropriate result with the illation was achieved which can be helped for the design of system involving the transmitting of the laser in the atmosphere such as laser active detecting, range measurement with laser and so on.

In order to solve the impact of fast illumination change on moving object detection in a visual surveillance system, a new method of moving target detection is put forward. Through establishing illumination change model, the chromaticity difference model and brightness ratio model are used to eliminate the impact of fast illumination change. When fast illumination is changing fast, background pixels and moving object pixels are detected as foreground pixels. To separate moving object pixels and false foreground pixels from detected foreground pixels, chromaticity difference model and brightness ratio model are developed respectively to estimate the intensity difference and intensity ratio of false foreground pixels. The establishment of Chromaticity Difference Model (CDM) and Brightness Ratio Model (BRM) are based on the proposed illumination change model. The experimental results show that the method has good detection effect and real-time in the rapidly changing of illumination conditions.

In order to enhance target-background contrast in target detection, we use micropolarizer camera providing a division of focal plane to obtain four polarization status images. Based on the polarization status images, the Stokes vectors, degree of linear polarization, angle of polarization and polarization parameter F are calculated. The result shows that, the degree of linear polarization enhances target-background contrast by a large margin, comparing with the origin intensity image, and the polarization parameter enhances the clearness and expands information content of the image, meanwhile keeps the advantage of the degree of linear polarization image.

In order to improve the real-time and accuracy of the target detection, an improved target detection method based on spatiotemporal background subtraction is proposed. The spatiotemporal background model based on the color value and texture feature will be established by video sequence. Then the difference operation between the current data that carry both color intensity and texture feature and the background model will be implemented. When operation result satisfies the threshold, the pixel is labeled as target; otherwise it is considered background. Finally, we take advantage of the difference in image and background to update threshold and background model. Multiple video test results show that the method has low memory complexity, high accuracy, faster operation speed and robustness in the case of only processing gray image.

This paper establishes the optical tomography system based on single pixel detector, and it provides a new imaging method for the modern infrared imaging guidance technology in military. It designs conical scan periscope optics and modulating plate with 65 slits, making a good use of the high resolution of optical tomography system. Projection values of single dimension are obtained by the modulating plate scanning, which can rebuild the two-dimensional image of the testing object by the filter back projection accelerating algorithm. The single pixel detector with high frequency response is combined with hardware acceleration and accelerating algorithm of filter back projection, which can realize high-speed image acquisition. The paper simulates the whole optical system practically, and tomographic imaging system parameters with 65 slits are calculated. Studying the comparison of results between image reconstruction without noise and image reconstruction with noise, it is verified that imaging system has a good signal-to-noise ratio. This method plays an important role in developing novel target tracking system and provides deep foundation for deeper experimental research.

Tracking by Detection (TBD) is a widely used framework in object tracking. Most TBD algorithms focus on object`s appearance model, but hard to consider both fps and success rate. Point to these problem, a new and rapid tracking framework is imported which uses the On-line Sequential Extreme Learning Machine(OS-ELM) to update object`s appearance model incrementally. Due to the learning speed of elm is fast enough, classifier could be updated every frame, so the classifier is more suitable to object`s apparent variations. The result shows this algorithm realizes real time tracking, and the success rate is higher than other TBD algorithms.

Laser systems which operate in space environment must suffer irradiation of charging particles. Optical films are the weakest link in laser systems, and their performance and stability are easy to be affected by irradiation of charging particles. We prepared an anti-reflective film sample at 1 064 nm by Ion Beam Sputtering Deposition (IBSD), and the transmittance is 99.964 5% and the absorptance is 50 ppm at 1 064 nm. Then, used low energy protons and electrons (40 KeV) with fluencies of 1.8×1013 eV/cm2 to irradiate the sample for space charging particles simulation. The irradiation caused a transmittance decrease of 362 ppm and an absorptance increase of 5 ppm at 1 064 nm. After thermal annealing, the optical performance of the sample was recovered. We used SRIM simulation results to analyze the experimental results, and concluded that the vacancy damage is the reason for the optical performance degradation of the film at 1 064 nm.

ZnO-based Thin Film Transistors (ZnO-TFTs) were fabricated by Radio Frequency (RF) magnetron sputtering successfully, and the temperature dependence and influence mechanism of the electrical characteristics of ZnO-TFTs are investigated. With the increase of the test temperature in the temperature range from 27 ℃ to 210 ℃, the on/off current ratio and the threshold voltage of the ZnO-TFT decrease significantly, and the subthreshold swing increases obviously, and the carrier mobility increases firstly and then decreases gradually. The change of electrical properties is mainly due to the combination effect of the increase of carrier concentration, the generation of point defects, and the enhancement of interface scattering in the channel active layer caused by the temperature increase. In addition, when the device is instantaneously cooled to the initial temperature, there is a hysteresis in the electrical characteristics. The main reason is that the recombination process need take a long time to reach the initial state for the generated point defects and interstitial oxygen atoms in the active layer caused by high temperature on the heating stage.

A Bayesian wavelet speckle reduction algorithm for SAR image is developed under the non-homomorphic framework. We use Normal Inverse Gaussian (NIG) function for modeling backscattered signal in wavelet domain, and Gaussian function for speckle noise (i.e. signal-dependent noise). The estimation formula of noise-free signal is derived by Bayesian maximum a posteriori (MAP) criterion. With regarding to estimation of model parameters, we introduce Multiscale Local Coefficient of Variation (MLCV) as heterogeneity measure, the histogram of which can be well fitted by logarithmic normal distribution. Based on heterogeneity measure, each coefficient in wavelet sub-band is classified into one of several different heterogeneity scenes, and NIG model parameters are computed in each class through cumulants estimation method. Experiment results show that, compared with its counterpart algorithm in homomorphic framework and its counterpart algorithm in non-homomorphic framework without heterogeneity based classification, our method has obvious advantage in terms of both subjective and objective evaluation, and has obtained satisfactory de-speckled image. A classification method of wavelet coefficients is proposed by heterogeneity measure, which could provide a new means for the research of SAR image despeckling.

A detection algorithm of weapon concealed underneath person’s clothes based on Fuzzy C-means (FCM) clustering is presented, aiming at the problem of image fusion, for example, no protecting the personal privacy, no true color. The main idea of the algorithm is to segment and extract the shape of concealed weapon from the infrared image by using FCM clustering and mathematical morphology. And the weapon is embed into I component of visual image in HSI color space. At last, the image is converted into RGB color space. In this process, only the concealed weapon is participated in the fusion and the other parts of the human body are avoided, which protect the personal privacy. The fused image maintains true color of the visual image, and the concealed weapon is very clear. By comparison experiment, the results show that the fused image obtained from the proposed algorithm can preserve a large amount of information and have the better visual quality and objective evaluation index.

NVIDIA as the inventor of the GPU provides a library function CUFFT for computing Fast Fourier Transform (FFT). After several generations update of CUFFT, there is still promotion space and it is not suit for kernel fusing on GPU to reduce the memory access and increase the Instruction Level Parallelism (ILP). We develop our own custom GPU FFT implementation based on the well-known Cooley-Tukey algorithm. We analyze the relationship of coalesce memory access and occupancy of GPU and get the optimal configuration of thread block. The results show that the proposed method improved the computational efficiency by 1.27 times than CUFFT 6.5 for double complex data 512×512. And then it is used to the computation of OTF with kernel fusing strategy, and it improved the efficiency of computation about 1.5 times than conventional method using CUFFT.

A retrieval method based on the fusion of Bag of Visual Words (BoVW) and Gabor texture is presented for the high resolution remote sensing images. Remote sensing images have rich texture information and many local key points. But when an image contains lots of similar texture, the retrieval precision of BoVW will be reduced. The fusion of BoVW and Gabor texture combines the advantages of local feature and global feature, mid-level feature and low-level texture to improve image description. Experiment results show that the presented fusion method is superior to the traditional fusion method using Gabor texture and color moments. Retrieval performance of the fused features method is improved compared with that using single feature, and the improved performance depended on the suitable fusion weights. Experiment results indicate that the fused BoVW and Gabor texture is effective for high-resolution remote sensing image retrieval.

In order to overcome the defects of the traditional de-hazing algorithm based on dark channel, such as a long working time, high complexity, bad image quality, and so on, an improved algorithm is proposed. This proposed algorithm firstly uses edge detection and quad-tree algorithms to locate the interval of the atmospheric light A fast and uses the skyline algorithm to obtain the value of A precisely. Then, use the improved Wiener filter, which has the abilities of de-noising and edge preserving, to inhibit halo effect of the transmission. Finally, optimize the edge information of the transmission by dilation and erosion. A plenty of experimental results show that the improved algorithm has a better visual effect and a lower time consumption comparing with the traditional de-hazing algorithms.

Bundle adjustment which simultaneously refines motion and structure has the fault of low numerical stability, such as dependence on the initial value, slow convergence speed, and convergence divergence. In this paper, we propose a new multi-frame sequence motion estimation method for stereo visual localization. The method has the fast convergence speed, can converge to the global minimum, and greatly reduce the accumulated error. Stereo visual localization experiments with simulated data and outdoor intelligent vehicle show that our algorithm outperforms bundle adjustment in terms of run-time, accuracy, resistance to noise and dependence on the initial value.