基于GFU和分层LSTM的组群行为识别研究方法

提出一种以"关键人物"为核心，使用门控融合单元（GFU，Gated Fusion Unit）进行特征融合的组群行为识别框架，旨在解决两个问题：①组群行为信息冗余，重点关注关键人物行为特征，忽略无关人员对组群行为的影响；②组群内部交互行为复杂，使用GFU有效融合以关键人物为核心的交互特征，再通过LSTM时序建模成为表征能力更强的组群特征.最终，通过softmax分类器进行组群行为类别分类.该算法在排球数据集上取得了86.7%的平均识别率.     

An adaptive image steganographic scheme based on Noise Visibility Function and an optimal chaotic based encryption method

Steganography is the science of hiding secret message in an appropriate digital multimedia in such a way that the existence of the embedded message should be invisible to anyone apart from the sender or the intended recipient. This paper presents an irreversible scheme for hiding a secret image in the cover image that is able to improve both the visual quality and the security of the stego-image while still providing a large embedding capacity. This is achieved by a hybrid steganography scheme incorporates Noise Visibility Function (NVF) and an optimal chaotic based encryption scheme. In the embedding process, first to reduce the image distortion and to increase the embedding capacity, the payload of each region of the cover image is determined dynamically according to NVF. NVF analyzes the local image properties to identify the complex areas where more secret bits should be embedded. This ensures to maintain a high visual quality of the stego-image as well as a large embedding capacity. Second, the security of the secret image is brought about by an optimal chaotic based encryption scheme to transform the secret image into an encrypted image. Third, the optimal chaotic based encryption scheme is achieved by using a hybrid optimization of Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) which is allowing us to find an optimal secret key. The optimal secret key is able to encrypt the secret image so as the rate of changes after embedding process be decreased which results in increasing the quality of the stego-image. In the extracting process, the secret image can be extracted from the stego-image losslessly without referring to the original cover image. The experimental results confirm that the proposed scheme not only has the ability to achieve a good trade-off between the payload and the stego-image quality, but also can resist against the statistics and image processing attacks.     

超高清电视系统关键技术浅析

本文对国内外的电视技术发展现状进行了充分的研究和分析,并对超高清电视系统的相关图像技术参数进行了分析和介绍。     

基于张量分解融合RGB-D图像的物体识别

为了充分利用RGB-D图像的深度图像信息，提出了基于张量分解的物体识别方法。首先将RGB-D图像构造成一个四阶张量，然后将该四阶张量分解为一个核心张量和四个因子矩阵，再利用相应的因子矩阵将原张量进行投影，获得融合后的RGB-D数据，最后输入到卷积神经网络中进行识别。RGB-D数据集中三组相似物体的识别结果表明，利用张量分解融合RGB-D图像的物体识别准确率高于未采用张量分解的物体识别准确率，并且单一错分实例的准确率最高可提升99%。     

Additive manufacturing of bipolar plates for hydrogen production in proton exchange membrane water electrolysis cells

Water electrolysis is a process that can produce hydrogen in a clean way when renewable energy sources are used. This allows managing large renewable surpluses and transferring this energy to other sectors, such as industry or transport. Among the electrolytic technologies to produce hydrogen, proton exchange membrane (PEM) electrolysis is a promising alternative. One of the main components of PEM electrolysis cells are the bipolar plates, which are machined with a series of flow distribution channels, largely responsible for their performance and durability. In this work, AISI 316L stainless steel bipolar plates have been built by additive manufacturing (AM), using laser powder bed fusion (PBF-L) technology. These bipolar plates were subjected to ex-situ corrosion tests and assembled in an electrolysis cell to evaluate the polarization curve. Furthermore, the obtained results were compared with bipolar plates manufactured by conventional machining processes (MEC). The obtained experimental results are very similar for both manufacturing methods. This demonstrates the viability of the PBF-L technology to produce metal bipolar plates for PEM electrolyzers and opens the possibilities to design new and more complex flow distribution channels and to test these designs in initial phases before scaling them to larger surfaces.     

Rapid screening of DON contamination in whole wheat meals by Vis/NIR spectroscopy and computer vision coupling technology

This work intends to develop an online experimental system for screening of deoxynivalenol (DON) contamination in whole wheat meals by visible/near-infrared (Vis/NIR) spectroscopy and computer vision coupling technology. Spectral and image information of samples with various DON levels was collected at speed of 0.15 m s⁻¹ on a conveyor belt. The two-type data were then integrated and subjected to chemometric analysis. Discriminant analysis showed that samples could be classified by setting 1000 μg kg⁻¹ as the cut-off value. The best correct classified rate obtained in prediction was 93.55% based on fusion of spectral and image features, with reduced prediction uncertainty as compared to single feature. However, quantification of DON by quantitative analysis was not successful due to poor model performance. These results indicate that, although not accurate enough to provide conclusive result, this coupling technology could be adopted for rapid screening of DON contamination in cereals and feeds during processing.     

Sparsity-inspired image reconstruction for electrical capacitance tomography

We present a new image reconstruction method for Electrical Capacitance Tomography (ECT). ECT image reconstruction is generally ill-posed because the number of measurements is small whereas the image dimensions are large. Here, we present a sparsity-inspired approach to achieve better ECT image reconstruction from the small number of measurements. Our approach for ECT image reconstruction is based on Total Variation (TV) regularization. We apply an efficient Split-Bregman Iteration (SBI) approach to solve the problem. We also propose three metrics to evaluate image reconstruction performance, i.e., a joint metric of positive reconstruction rate (PRR) and false reconstruction rate (FRR), correlation coefficient, and a shape and location metric. The results on both synthetic and real data show that the proposed TV-SBI method can better preserve the edges of images and better resolve different objects within reconstructed images, as compared to a representative state-of-the-art ECT image reconstruction algorithm, Projected Landweber Iteration with Linear Back Projection initialization (LBP-PLI).     

Change detection in synthetic aperture radar images based on unsupervised artificial immune systems

In this paper, we propose a novel change detection method for synthetic aperture radar images based on unsupervised artificial immune systems. After generating the difference image from the multitemporal images, we take each pixel as an antigen and build an immune model to deal with the antigens. By continuously stimulating the immune model, the antigens are classified into two groups, changed and unchanged. Firstly, the proposed method incorporates the local information in order to restrain the impact of speckle noise. Secondly, the proposed method simulates the immune response process in a fuzzy way to get an accurate result by retaining more image details. We introduce a fuzzy membership of the antigen and then update the antibodies and memory cells according to the membership. Compared with the clustering algorithms we have proposed in our previous works, the new method inherits immunological properties from immune systems and is robust to speckle noise due to the use of local information as well as fuzzy strategy. Experiments on real synthetic aperture radar images show that the proposed method performs well on several kinds of difference images and engenders more robust result than the other compared methods.     

ECISER: Efficient Clip-art Image SEgmentation by Re-rasterization

Clip-art image segmentation is widely used as an essential step to solve many vision problems such as colorization and vectorization. Many of these applications not only demand accurate segmentation results, but also have little tolerance for time cost, which leads to the main challenge of this kind of segmentation. However, most existing segmentation techniques are found not sufficient for this purpose due to either their high computation cost or low accuracy. To address such issues, we propose a novel segmentation approach, ECISER, which is well-suited in this context. The basic idea of ECISER is to take advantage of the particular nature of cartoon images and connect image segmentation with aliased rasterization. Based on such relationship, a clip-art image can be quickly segmented into regions by re-rasterization of the original image and several other computationally efficient techniques developed in this paper. Experimental results show that our method achieves dramatic computational speedups over the current state-of-the-art approaches, while preserving almost the same quality of results.