Search

Nonlinear and nonconvex optimization problem (NNOP) is a challenging problem in control theory and applications. In this paper, a novel mixture varying-gain dynamic learning network (MVG-DLN) is proposed to solve NNOP with inequality constraints. To do so, first, this NNOP is transformed into some equations through Karush–Kuhn–Tucker (KKT) conditions and projection theorem, and the neuro-dynamics function can be obtained. Second, the time varying convergence parameter is utilized to obtain a faster convergence speed. Third, an integral term is used to strengthen the robustness. Theoretical analysis proves that the proposed MVG-DLN has global convergence and good robustness. Three numerical simulation comparisons between FT-FP-CDNN and MVG-DLN substantiate the faster convergence performance and greater robustness of the MVG-DLN in solving the nonlinear and nonconvex optimization problems.

Image indexing plays a significant role in approximate K -nearest neighbor image retrieval. In the past decades, the most commonly used methods in image indexing are unsupervised-based. We recognize that unsupervised-based methods cannot benefit from supervised information. In this paper, supervised learning to index method is proposed to realize approximate K -nearest neighbor image retrieval. For designing the supervised learning to index algorithm, we first need to establish the supervision information of an image that should be quantified to the codeword in the codebook. Thus, we propose an image relabeling algorithm to relabel the database images so that the supervised codebook learning can be realized. Second, compared with only distance-based unsupervised image indexing method, a better idea of supervised learning to index is incorporating the distance-based similarities with the classification probabilities of samples at the image indexing stage. In order to achieve this goal, a classifier is also trained based on the relabeled images, and we finally build a constraint optimization objective function to learn the codebook and the classifier simultaneously. The proposed method is verified on the MNIST, the CIFAR-10, and the ImageNet data sets. Experimental results demonstrate the effectiveness of the proposed method.

The accurate estimation of camera internal and external parameters is one of the keys to high-precision camera calibration. At present, the mainstream Zhang's calibration method requires the use of a checkerboard to achieve high-precision calibration result. Therefore, the precise extraction of the pixel coordinates at the corner points on the checkerboard is one of the key technologies. The traditional methods are to improve the corner extraction algorithm, but ignore the imaging process of the calibration images. It is difficult to detect and eliminate the high-light regions in the process of calibrating image, and the existence of high-light regions will cause low accuracy of camera calibration. Therefore, a camera calibration method based on optimal polarization information is proposed in this paper. The method first uses a Stokes variable linear feedback through a CCD camera equipped with a polarizer to obtain a checkerboard calibration image at an optimal polarization angle in different spatial positions. Then use the sub-pixel level detection algorithm and Gaussian fitting method to locate the corner points in the image and to solve the sub-pixel coordinates of the corner points precisely. At last, the two-dimensional pixel coordinates of the corner points in each checkerboard image are extracted to perform camera calibration. The experimental results show that the coordinates of corner points of calibration plate image based on the optimal polarization angle have about 1 pixel difference compared with the coordinates of the calibration plate corner points under natural illumination, and the slopes between the parallel lines of the fitted are closer, meet the high precision calibration requirement of camera.

Structured light projection is widely applied in 3D shape measurement. Affected by uneven illumination, the highlight is produced on high-reflectivity surfaces, and the fringe pattern on the surface will be covered by it completely. To solve this problem, a polarization-based method is proposed to remove the image highlight. The experiment result proved that the image highlights are removed and the SNR of polarization image is improved. Meanwhile, a normalized weighting algorithm is applied in this method to restore the surface information in the highlighted area. The experiment results show that most of the fringe pattern is restored.

It is a challenge task to reconstruct images from compressed sensing measurement due to its implicit ill-posed property. In this paper, we propose an image reconstruction algorithm for compressed sensing image application based on the adaptive dictionary, which is learned from the reconstructed image itself. The sparsity level is enhanced since the sparse coding of overlapping image patches emphasizes the local image features; accordingly the quality of the reconstructed image is also improved. In addition, Batch-OMP algorithm, linearization technique and dynamic updating sparse coding algorithm are used to reduce the computational complexity of our proposed algorithm. Numerical experiments are conducted on several test images with a variety of sampling ratios. The results demonstrate that our proposed algorithm can efficiently reconstruct images from compressed sensing measurements and achieve more than 3 dB gain averagely over the current leading CS reconstruction algorithm.

Currently, Hadamard transform spectral imager with the digital micro-mirror device (DMD) as the optical modulator is one of the most popular research topics in international spectral imaging field. The physical models of DMD-based Hadamard transform spectral imager have been experimentally demonstrated to be feasible. However, the research about how to process the recovered spectral data of this type of spectral imager to reach the standard of engineering applications is quite rare. This paper discusses the spectral data processing method, mainly including the correction for the recovered spectral data and how to decode the interlaced encoding pixels. Eventually, the processing methods proposed is proved to be effective and feasible by experiments.

An indirect illumination pattern with light-emitting diodes (LED) and diffuse transmissive free-form surface was proposed, and in this paper, we designed an ultra-compact rotational symmetric lens with double free-form surfaces based on the edge-ray principle. Based on the Lambertian characteristic of LEDs and ideal diffuse transmissive surface (e.g. frosted glass), a mathematical model was constructed. The coordinates of profile curve of the transmitting part are calculated by Matlab according to the Snell’s law and iterative method at first. Then the TIR lens model is established with Solidworks. The diffuse transmissive free-form surface is constructed by solving a series of equations numerically. The inner surface was elliptic and the outer surface was diffuse transmissive free-form. Furthermore, to keep the optical system compact and provide enough space for the LED, the size of the designed lens model was also greatly reduced. Finally, the results show better uniformity and higher efficiency comparing with the single diffuse transmissive free-form surfaces and flats, and the size of the lens was dramatically reduced. The Monte-Carlo ray-tracing result shows an illumination circular area with a clear cut-off at R = 150 mm at the target plane in a distance H = 600 mm for the near-field illumination. The uniformity within the target illumination area is greater than 86%.

The third-order nonlinearity parameter C/A for biological media is studied theoretically and experimentally. Based on the theory of fluid dynamics, the generalized Burgers equation is extended to the third order for biological media. A perturbation solution of the equation is carried out and the formula for C/A for biological media is derived. The sound pressure of the third harmonics changing with distance in water is measured and compared with the theoretical results. An experimental set-up is established for measurement of C/A for biological media and the values of the acoustic nonlinearity parameter C/A are measured for several porcine samples. The results show that the values of C/A for the measured biological tissues are in the range of 60.0–120.0 while the values B/A for the measured biological tissues are in the range 6.5–11.0. The maximum relative deviation for C/A among the measured biological tissues is 81.2% while it is only 66.2% for B/A. Therefore, C/A may be more appropriate for tissue characterization compared with B/A.

Video transcoding is one of the key techniques to provide heterogeneous multimedia applications with Universal Multimedia Access (UMA) service. Recently, compressed domain (DCT, Discrete Cosine Transform) transcoding architectures have been proposed to achieve fast transcoding, in which DCT domain Inverse Motion Compensation (IMC) is the most important module to reduce computational complexity. The problem of IMC has been studied for integer-pixel precision motion vectors (MVs), and, subsequently, for half-pixel precision MVs in which extra filtering, hence, extra computation is introduced. Since most current video coding standards, such as H.263, MPEG-2, etc., use half-pixel precision MVs to achieve better quality, reducing computational complexity of IMC for half-pixel MVs is important for real time fast transcoding video streams encoded by these standards. In this paper, we propose a novel half-pixel filter for IMC, which simplifies the extract operations by integrating the interpolation and translation operations into one single step. Compared to other existing half-pixel filtering algorithms, the proposed filter does not introduce any distortion and drift errors. Experimental results demonstrate that the proposed filter achieves faster transcoding than other DCT domain transcoders, and almost the same video quality as that of pixel domain transcoders.

The damage mechanism of a cracked material due to laser beam thermal shock is an important problem when the interactions of high power laser beam with materials are studied. The transient thermal stress intensity factors (TSIFs) for a center crack in an infinite plate subjected to laser beam thermal shock are investigated. When the crack is in the heat affected zone, the compressive thermal stress causes the crack surface to come into contact with each other over a certain contact length, but the high tensile stresses around the crack tip tend to open the crack. In this case, the problem may be treated as a pair of embedded cracks problem with a smooth closure condition of the center crack. The TSIFs and the crack contact lengths are calculated with different laser beam spatial shapes. The TSIFs induced by thermal shock are in marked different from those induced by general mechanical loading.