Search

How to effectively and accurately detect target information in the formed images has long been an area that we attach great importance to. For example, we not only need the imaging results not to be affected by the weather and other environments, but also hope that the edge information of the imaging results is very clear and the image resolution is very high. Therefore, a single sensor cannot achieve this effect at the same time. In this case, the fusion rule (nonsubsampled contourlet transform) proposed in this paper solves the problem of insufficient imaging information of a single sensor and removes redundant information. The target is pre-screened to make the target area more accurate. Through spatial morphological filtering, false alarm rate is reduced to improve the success rate of target matching.

In this paper, a terahertz Coded-Aperture imaging system with phase modulation (TCAIS-PM) is presented. In TCAIS-PM, the coded aperture mask implements reflection and phase modulation of the terahertz (THz) beam. Considering that the THz image has a priori property of smooth regions, we introduced the total variation (TV) to the reconstruction approach. With the signal echo model, the image reconstruction problem can be transformed to constrained TV Regularization optimization problem. An Augmented Lagrange Method-Alternating Direction Method of Multipliers (ALM-ADMM) method is developed, which combines the multiplier method, the idea of alternating minimizing variables and second iterative process. In the end, simulations results are given to demonstrate the effectiveness and efficiency of the proposed imaging scheme and corresponding reconstruction method.

Owing to the explosion effect and local ship damage caused by multi-round missile attacks, it becomes very difficult to accurately recognize the specific ship target. To solve this problem, this paper properly combines the convolutional neural network (CNN) with the support vector machine, and proposes a new recognition method for locally damaged ship targets. By migrating the deep convolutional neural network to locally damaged ship targets recognition the proposed technique is able to learn the robust features of locally damaged ship targets, which can avoid difficult manual designed feature extraction processes. When the target characteristics of the ship vary on account of the explosion effect and local ship damage, it performs better than other methods. Since it is difficult to obtain the target samples of military ships, our method introduces transfer learning into the target recognition of ships. By training the CNN with the samples of general object and using damaged ship samples to fine-tuning characteristic network parameters and classifiers, the ship recognition capacity of our method still achieves excellent performance. The experimental results show that its recognition rate can reach 97%. And it is robust to the explosion effect and local ship damage, which change the target characteristics of the ship.

Virtual cell can be foreseen as an effective way to solve the challenges in the ultra-dense network (UDN) which manifests as one of key technologies for 5G system. However, with the increasing density of network, the interference scenario becomes more complex, and user centric virtual cells are more likely to be overlapped. This situation made the multi-dimensional cooperation much difficult due to the coupled nature of different resource dimensions and the complexity of the optimum solution. In this paper, a novel scheme is proposed to tackle the two challenges highlighted above by virtue of virtual cell forming approach and subsequent resource allocation. The proposed scheme exploits multi-dimensional cooperation gain in the virtual cell forming stage by virtue of a graph edge weight design, which captures the contribution from interference fair power allocation, MU-MIMO operation and cooperative transmission. Based on the edge weight, virtual cells are formed by a clustering algorithm named Pairwise Maximum Weights (PMW). The proposed scheme allows the advanced power allocation and MU-MIMO operation in the subsequent resource allocation when users are jointly scheduled. The simulation result demonstrates that the proposed algorithm can increase the system performance effectively at a low complexity compared with the reference scheme.

Digital image watermarking is a useful solution to the problem of information security, copyright and network security. In this paper, a new watermarking algorithm is propose based on Singular Value Decomposition (SVD) and discrete wavelet transform (DWT). The algorithm uses a gray image as a watermark, increasing embedded information capacity. The algorithm can satisfy the transparence and robustness of the watermarking system very well. The experimental result based on this algorithm have shown that the watermarking is robust to the common signal processing techniques including JPEG compressing, adding noise, low pass filter, median filter, cropping and so on.

In this paper, we propose a new approach to find a large number of reliable point features between high distorted stereo fisheye images, which is very valuable in many applications, such as image-based metrology, geometry estimation and 3D reconstruction. In first step the affine invariant regions are matched by classical approaches then distinctive points within the region can be matched based on the corresponding affine transformations in normalized frames. In addition, two distinctive point matching algorithms are proposed, one is based on the intensity similarity measure, and the others is based on the geometric consistency constraint. Experiment results show that the number of the final point correspondences increased greatly with high accuracy and reliability. This electronic document is a “live” template. The various components of your paper [title, text, heads, etc.] are already defined on the style sheet, as illustrated by the portions given in this document.

Compared with conventional perspective camera, data noise for omnidirectional camera is more serious due to its non-linear projection. For linear estimation of motion parameters with noise corrupted data, it is necessary to establish sufficient number of correspondences and use a appropriate robust tech- nique. In this paper, our interests concentrate on two parts. First, we propose a new method to establish a large number of reliable point correspondences between omnidirectional images. Classical affine invariant region detecting and matching algorithms are initially employed, then more distinctive point correspondences inside or near the matched regions are extended. Our second work focuses on the RANSAC based linear estimation technique. Based on the simulation experiment we give a strategy to grad- ually select more and more inliers from a noise-corrupted data set, by which the initial guess of motion parameters improved in precision and stability. Experiment results with true data show that our method works well in real applications and makes the performance of linear estimates of motion parameters improved greatly.