Light field imaging can record the intensity and direction information of light rays in space, which attracts extensive attention. However, the trade-off between angular and spatial resolution cannot be avoided due to the limitation of sensors in commercial light field cameras. To mitigate this problem, this paper proposes the view position prior-supervised light field angular super-resolution network with asymmetric feature extraction and spatial-angular interaction. First, there is a severe information asymmetry between the spatial and angular dimensions in light fields with sparse views. The asymmetric feature extraction block is proposed to extract spatial and angular features with different receptive fields in an asymmetric manner. As a result, more light field intrinsic features are extracted, which improves the utilization rate of the limited light field information. Second, the existing methods usually ignore the correlations among the newly synthesized views. The spatial-angular interaction module is proposed to collect the local and global information, build relations between any two points in the feature space, and reconstruct light field consistencies. Thus, the complete view correlations can be established. Last but not least, we investigate the impact of the given views on each viewpoint and propose a loss function based on the view position prior, which reduces the quality difference among the synthesized sub-aperture images and further improves the network performance. Comprehensive experiments demonstrate that our method can perform best on all datasets, and the depth estimation results present a new perspective to show the superiority of the proposed method.

Light field imaging can encode abundant scene information, including the intensity and direction of light rays, into 4D light field images. However, the limited number of sensors in commercial light field cameras leads to a trade-off between spatial and angular resolutions. In this paper, an angular super-resolution framework is proposed to synthesize new views and overcome hardware restrictions. First, light field intrinsic feature convolution is proposed to extract intrinsic information, i.e., scene content, complete view correlations, and epipolar structures. Consequently, spatial, angular, and cross-domain information can be preserved in the extracted features. Second, the spatial–angular and depth streams are proposed based on the light field intrinsic feature convolution to synthesize high angular resolution light fields. The spatial–angular stream utilizes the light field intrinsic information to improve the angular resolution, whereas the depth stream disentangles the geometric information from the extracted light field intrinsic features, which is used to warp the given sub-aperture images to the new view positions. Both streams can synthesize high-quality intermediate results, where the intrinsic and geometric information are utilized separately. Finally, a confidence-based stream fusion module is proposed to fuse the outputs from the two streams, achieving the joint employment of light field intrinsic and geometric information and solving the problem of insufficient information exploration in the current methods. We conduct a series of experiments to validate the effectiveness of each component in the framework and demonstrate that our method can achieve state-of-the-art performance in various scenes.

This article presents an efficient factorial-design-based method to calculate uncertainty propagation in the triangulation process in stereo-vision photogrammetry. A three-level full factorial numerical simulation experiment is executed by regarding 3-D measuring points as responses and image point pairs as factors. Hence, the covariance matrix of the 3-D measuring point coordinate can be computed from the responses of the factorial experiment with corresponding weights. The measurement uncertainties in ${x}$ -, ${y}$ -, and ${z}$ -axis directions as well as the principal axis direction are computed from the covariance matrix of the 3-D measuring point coordinate. The results calculated by the proposed method are compared with that of the Monte Carlo method, which shows that the proposed method can achieve equivalent precision with less computation cost. The spatial-variant uncertainty distribution of stereo-vision photogrammetry is analyzed using the proposed method. Measurement uncertainties under different parameters of stereo-vision photogrammetric system are discussed.

Light fields contain a wealth of information about real-world scenes and can be easily acquired by commercial light field cameras. However, the trade-off between angular and spatial resolution is inevitable. This paper proposes an end-to-end light field angular super-resolution network by exploiting structure and scene information to mitigate this problem. First, a Light Field ResBlock is proposed to explicitly extract epipolar plane image features and scene features. The epipolar plane image features with obvious directionality represent light field structure information that contains correlations among different views, such as the intensity and depth consistencies. The scene features express scene information that depicts local details and global information and describes individual objects in a scene. The exploitation of light field structure and scene information can gather the intrinsic characteristics of light fields. Then, 4D deconvolution is employed to upsample the angular resolution based on the extracted features related to light field structure and scene. The utilization of 4D deconvolution preserves the connections among sparse views and generates new views with high correlations. Last but not least, the refinement network further exploiting LF structure and scene information is employed to alleviate the artifacts induced by too little angular information during the early stage of the model. Experiments on two different angular super-resolution tasks demonstrate that our method can achieve the best angular resolution enhancement performance on different datasets.

As an extension of single/multiple stability notion, Lagrange stability is considered here. A class of switched inertial neural networks (SINNs) is studied on both continuous-time and discrete-time domain. Two kinds of activation functions are evolved for the network. Through characteristic function approach, matrix measure strategy, and theory of time scales, Lagrange stability of delayed continuous-/discrete-time SINNs with lurie-type and bounded-type activation functions are addressed, respectively. The results also show that the criteria corresponding to discrete-time network approach to those corresponding to continuous-time one as the graininess function tends to zero. Two examples are given to show the effectiveness of the main results.

We investigate kernel estimates in the functional nonparametric regression model when both the response and the explanatory variable (the covariate) are functional. The rates of almost complete and uniform almost complete convergence of the estimator are obtained under some mild $$\alpha $$-mixing functional sample. Finally, a simulation study is carried out to illustrate the finite sample performance of the estimator.

The solubility of two forms of imidacloprid, forms I and II, in seven pure organic solvents was measured using the analytical gravimetric method. It was shown that the solubility of the two forms of imidacloprid increases as the temperature increases. Meanwhile, the solubility at a low temperature follows the order methyl acetate > ethyl acetate > n-butyl acetate > methanol > ethanol > n-propanol > n-butanol and the order at relatively high temperatures is methyl acetate > ethyl acetate > methanol > n-butyl acetate > ethanol > n-propanol > n-butanol. Differential scanning calorimetry and solubility data show that the two forms are monotropically related and form I is the stable phase. Besides, the experimental solubility of forms I and II in the seven pure solvents is well correlated by the modified Apelblat model and lambda h model, respectively.

Solder bumps realize the mechanical and electrical interconnection between chips and substrates in surface mount components, such as flip chip, wafer level packaging and three-dimensional integration. With the trend to smaller and lighter electronics, solder bumps decrease in dimension and pitch in order to achieve higher I/O density. Automated and nondestructive defect inspection of solder bumps becomes more difficult. Machine learning is a way to recognize the solder bump defects online and overcome the effect caused by the human eye-fatigue. In this paper, we proposed an automated and nondestructive X-ray recognition method for defect inspection of solder bumps. The X-ray system captured the images of the samples and the solder bump images were segmented from the sample images. Seven features including four geometric features, one texture feature and two frequency-domain features were extracted. The ensemble-ELM was established to recognize the defects intelligently. The results demonstrated the high recognition rate compared to the single-ELM. Therefore, this method has high potentiality for automated X-ray recognition of solder bump defects online and reliable.

Flip chip technology has been widely used in IC packaging, and the combination of flip chip technology and solder joint interconnection technology has been utilized in the manufacturing of electronic devices universally. As the development of flip chip towards high density and ultra-fine pitch, the inspection of flip chips is confronted with great challenges. In this paper, we developed an intelligent system used for the detection of flip chips based on vibration. Thirty-four features including 18 time domain features and 16 frequency domain features were extracted from the raw vibration data. The support vector machine was employed to implement the recognition and classification of flip chips. In order to improve the classification accuracy of SVM, cross validation (CV) and genetic algorithm (GA) were utilized to optimize the parameters of SVM respectively. SVM, CV-SVM and GA-SVM were applied to classification separately and the results were obtained. By comparison, GA-SVM can recognize and classify the flip chips rapidly with high accuracy. Thus, GA-SVM is effective for the defect inspection of flip chips.

This paper is concerned on the global exponential synchronization in timescale sense for a class of nonautonomous recurrent neural networks (NRNNs) with discrete-time delays on time scales. Firstly, a timescale-type comparison result is given based on the induction principle of time scales. Then by the constructed comparison lemma, the theory of time scales and analytical techniques, several synchronization criteria for the driven and response NRNNs are obtained. Moreover, several examples are given to show the effectiveness and validity of the main results. The obtained synchronization criteria improve or extend some existing ones in the literature.

This paper deals with the passivity problem of inertial neural networks (INNs) with discrete delays on time scales. By one linear variable transformation, one passivity criterion is obtained based on the calculus of time scale and LMI techniques Furthermore, two particular criteria for the corresponding continuous and discrete time INNs are derived. Some numerical examples are given to show the validity of the obtained results. The time scale scheme offers results that hold for continuous-time systems, discrete-time systems, and the systems that involve on time intervals.

This paper proposes a novel technique to analyze the infrared vein patterns in the back of the hand for biometric purposes. The line-patterns are retrieved in a rotation and translation invariant manner to tackle broken line problem. A set of pair-wise angle relationships among lines are extracted to present the local structure of line pattern. In order to keep positional information as well, dynamic region division technique is proposed. Pair-wise angle relationships are simple, invariant to translation and rotation, robust to end-point erosion and segment error, and sufficient for discrimination. The matching experiment has shown encouraging results which implicate that line segments can provide sufficient information for vein recognition.

This paper presents a biometric system based on physiological hand vein patterns for personal identification. The paper proposes a framework to analyse the infrared vein patterns; and it provides solutions to two key stages of the framework: vein pattern image acquisition, and feature identification. We proposed and investigated two non-invasive image acquisition methods: far-infrared (FIR) thermography and near-infrared (NIR) imaging. Both imaging techniques are found capable of capturing the vein structures in the hand. Further analysis was carried out to identify the features of the vein pattern to provide strong discriminating power. The paper firstly utilizes the salient points features (SPF) of the vein patterns as a geometric representation of the shape of vein patterns. Experiments show SPF achieved a 0% of equal error rate (EER), which indicates the SPF can provide strong discriminating power for biometric purposes. Then, the paper proposed another novel feature type to analyze the vein patterns, the Line Edge Maps (LEM), which represents the vein patterns using a set of line segments. Experimental results showed that the LEM gives comparable matching accuracy to the one by SPF; and the LEM outperforms the SPF under the circumstances of object occlusion, which is more preferable in most of the applications. Additional experiments were carried out on identical twins, and the initial results show that the vein patterns are potentially genetically independent. Overall, the work in this paper shows the high potential of the vein patterns serving as a strong biometrics.

This paper describes a novel approach for personal verification by analyzing the face patterns formed by thermal imaging. A new feature set is proposed to represent the thermal face: the bifurcation points of the thermal pattern and the geographical gravity center of the thermal face region. We proposed to use the Modified Hausdorff Distance to measure the similarity between two feature vectors of the thermal face. Experimental results show that introduction of the geographical gravity center improves the accuracy performance significantly, and the Equal Error Rate (EER) achieved 6.7%. The work in this paper shows that the thermal face patterns can provide a reasonable level of discriminating power, and has the potential to be used in the context Of biometric applications especially when used in conjunction with other biometric modals.

This paper proposes a novel technique to analyze the infrared vein patterns in the back of the hand for biometric purposes. The technique utilizes the minutiae features extracted from the vein patterns for recognition, which include bifurcation points and ending points. Similar to fingerprints, these feature points are used as a geometric representation of the shape of vein patterns. Analysis of a database of infrared vein patterns shows a trend that for each hand vein pattern image, there are, on average, 13 minutiae points in each vein pattern image, including 7 bifurcation and 6 ending points. The modified Hausdorff distance algorithm is proposed to evaluate the discriminating power of these minutiae for person verification purposes. Experimental results show the algorithm reaches 0% of equal error rate (EER) on the database of 47 distinct subjects, which indicates the minutiae features of the vein pattern can be used to perform personal verification tasks. The paper also presents the preprocessing techniques to obtain the minutiae points as well as in-depth study on their tolerance to processing errors, such as loss of features and geometrical displacement.

In vein pattern biometrics, analysis of the shape of the vein pattern is the most critical task for person identification. One of best representations of the shape of vein patterns is the skeleton of the pattern. Many traditional skeletonization algorithms are based on binary images. In this paper, we propose a novel technique that utilizes the watershed algorithm to extract the skeletons of vein patterns directly from gray-scale images. This approach eliminates the segmentation stage, and hence prevents any error occurring during this process from propagating to the skeletonization stage. Experiments are carried out on a thermal vein pattern images database. Results show that watershed algorithm is capable of extracting the skeletons of the veins effectively, and also avoids any artifacts introduced by the binarization stage

This paper presents a novel technique to derive the filter parameters for removing signal dependent noise (SDN) in the image. In order to remove SDN, many de-noising algorithms rely on a priori knowledge of noise parameters, especially the variance sigman 2, and the gamma value gamma of the specific imaging technique. This paper proposes a technique to automatically derive the signal variance sigmaf 2 and use this parameter to construct the Local. Linear Minimum Mean Square Error (LLMMSE) filter without the need to know the values of sigman 2 and gamma. Two image instances of the same noisy scene are used to calculate the signal variance which is then used to construct the LLMMSE filter. Experiments with both the "Lena" image and real-life far-infrared (FIR) vein pattern images showed that the proposed technique can predict the signal variance consistently, and the constructed LLMMSE filter performs well in removing the signal dependent noise.

This paper investigates two infrared imaging technologies, far-infrared thermography and near-infrared imaging, to acquire hand vein pattern images for biometric purposes. The imaging principles for both technologies are studied in depth. Experiments involving data acquisition from various parts of hand, including the back of the hand, palm, and wrist are described using a population of 150 participants using both near and far infrared imaging techniques. Comparison and analysis of the data collected show that far-infrared thermography has difficulties in capturing vein images in the palm, and wrist. However, while it is more suitable for capturing the large veins in the back of the hand, it is sensitive to ambient conditions and human body condition and does not provide a stable image quality. On the other hand, near-infrared imaging produces good quality images when capturing vein patterns in the back of the hand, palm, and wrist. It is more tolerant to changes in environmental and body condition, but it also faces the problem of disruption due to skin features such as hairs and line patterns. An initial vein pattern biometric system is implemented. The results show that all the test subjects can be correctly identified.

In vein pattern biometrics, analysis of the shape of the vein pattern is the most critical task for person identification. One of best representations of the shape of vein patterns is the skeleton of the pattern. Many traditional skeletonization algorithms are based on binary images. In this paper, we propose a novel technique that utilizes the watershed algorithm to extract the skeletons of vein patterns directly from gray-scale images. This approach eliminates the segmentation stage, and hence prevents any error occurring during this process from propagating to the skeletonization stage. Experiments are carried out on a thermal vein pattern images database. Results show that watershed algorithm is capable of extracting the skeletons of the veins effectively, and also avoids any artifacts introduced by the binarization stage

Many biometrics, such as face, fingerprint and iris images, have been studied extensively for personal verification purposes in the past few decades. However, verification using vein patterns is less developed compared to other human traits. A new personal verification system using the thermal-imaged vein pattern in the back of the hand is presented in the paper. The system consists of five individual steps:Data Acquisition, Image Enhancement, Vein Pattern Segmentation, Skeletonization and Matching. Unlike most biometric systems that carry out comparisons based on a pre-selected feature set, this system directly recognizes the shapes of the vein pattern by measuring their Line-Segment Hausdorff Distance. Preliminary testing on a database containing 108 different images has been carried out and all the images are correctly recognized.