Robust optical flow estimation via edge preserving filtering

It is known that optical flow estimation techniques suffer from the issues of ill-defined edges and boundaries of the moving objects. Traditional variational methods for optical flow estimation are not robust to handle these issues since the local filters in these methods do not hold the robustness near the edges. In this paper, we propose a non-local total variation NLTV-$L^1$ optical flow estimation method based on robust weighted guided filtering. Specifically, first, the robust weighted guided filtering objective function is proposed to preserve motion edges. The proposed objective function is based on the linear model which is computationally efficient and edge-preserving in complex natural scenarios. Second, the proposed weighted guided filtering objective function is incorporated into the non-local total variation NLTV-$L^1$ energy function. Finally, the novel NLTV-$L^1$ optical flow method is performed using the coarse-to-fine process. Additionally, we modify some state-of-the-art variational optical flow estimation methods by the robust weighted guided filtering objective function to verify the performance on Middlebury, MPI-Sintel, and Foggy Zurich sequences. Experimental results show that the proposed method can preserve edges and improve the accuracy of optical flow estimation compared with several state-of-the-art methods.     

基于线状特性的大数据地铁用户识别与定位方法

从MR原始数据特性出发研究数据结构与字段,采用基于线状特性分析MR大数据的地铁用户识别过滤方法和位置定位方法,该方法经过验证能够有效输出地铁用户数据并定位在地图上,解决人工地铁测试效率低且测试片面的问题,实现基于大数据的自动化地铁网络评估,提供海量的有效定位数据用于地铁评估与分析。     

Tissue-Specific Ferritin- and GFP-Based Genetic Vectors Visualize Neurons by MRI in the Intact and Post-Ischemic Rat Brain

(1) Background: Neurogenesis is considered to be a potential brain repair mechanism and is enhanced in stroke. It is difficult to reconstruct the neurogenesis process only from the histological sections taken from different animals at different stages of brain damage and restoration. Study of neurogenesis would greatly benefit from development of tissue-specific visualization probes. (2) Purpose: The study aimed to explore if overexpression of ferritin, a nontoxic iron-binding protein, under a doublecortin promoter can be used for non-invasive visualization of neurogenesis using magnetic resonance imaging (MRI). (3) Methods: Ferritin heavy chain (FerrH) was expressed in the adeno-associated viral backbone (AAV) under the doublecortin promoter (pDCX), specific for young neurons, in the viral construct AAV-pDCX-FerrH. Expression of the enhanced green fluorescent protein (eGFP) was used as an expression control (AAV-pDCX-eGFP). The viral vectors or phosphate-buffered saline (PBS) were injected intracerebrally into 18 adult male Sprague–Dawley rats. Three days before injection, rats underwent transient middle-cerebral-artery occlusion or sham operation. Animals were subjected to In vivo MRI study before surgery and on days 7, 14, 21, and 28 days after injection using a Bruker BioSpec 11.7 T scanner. Brain sections obtained on day 28 after injection were immunostained for ferritin, young (DCX) and mature (NeuN) neurons, and activated microglia/macrophages (CD68). Additionally, RT-PCR was performed to confirm ferritin expression. (4) Results: T2* images in post-ischemic brains of animals injected with AAV-pDCX-FerrH showed two distinct zones of MRI signal hypointensity in the ipsilesioned hemisphere starting from 14 days after viral injection—in the ischemic lesion and near the lateral ventricle and subventricular zone (SVZ). In sham-operated animals, only one zone of hypointensity near the lateral ventricle and SVZ was revealed. Immunochemistry showed that ferritin-expressing cells in ischemic lesions were macrophages (88.1%), while ferritin-expressing cells near the lateral ventricle in animals both after ischemia and sham operation were mostly mature (55.7% and 61.8%, respectively) and young (30.6% and 7.1%, respectively) neurons. RT-PCR confirmed upregulated expression of ferritin in the caudoputamen and corpus callosum. Surprisingly, in animals injected with AAV-pDCX-eGFP we similarly observed two zones of hypointensity on T2* images. Cellular studies also showed the presence of mature (81.5%) and young neurons (6.1%) near the lateral ventricle in both postischemic and sham-operated animals, while macrophages in ischemic lesions were ferritin-positive (98.2%). (5) Conclusion: Ferritin overexpression induced by injection of AAV-pDCX-FerrH was detected by MRI using T2*-weighted images, which was confirmed by immunochemistry showing ferritin in young and mature neurons. Expression of eGFP also caused a comparable reduced MR signal intensity in T2*-weighted images. Additional studies are needed to investigate the potential and tissue-specific features of the use of eGFP and ferritin expression in MRI studies.     

基于特征融合的中文简历解析方法研究

针对基于规则和统计的传统中文简历解析方法效率低、成本高、泛化能力差的缺点，提出一种基于特征融合的中文简历解析方法，即级联Word2Vec生成的词向量和用BLSTM（Bidirectional Long Short-Term Memory）建模字序列生成的词向量，然后再结合BLSTM和CRF（Conditional Random Fields）对中文简历进行解析（BLSTM-CRF）。为了提高中文简历解析的效率，级联包含字序列信息的词向量和用Word2Vec生成的词向量，融合成一个新的词向量表示；再由BLSTM强大的学习能力融合词的上下文信息，输出所有可能标签序列的分值给CRF层；再由CRF引入标签之间约束关系求解最优序列。利用梯度下降算法训练神经网络，使用预先训练的词向量和Dropout优化神经网络，最终完成对中文简历的解析工作。实验结果表明，所提的特征融合方法优于传统的简历解析方法。     

High-resolution ISAR imaging of fast rotating targets based on pattern-coupled Bayesian strategy for multiple measurement vectors

Very high resolution inverse synthetic aperture radar (ISAR) imaging of fast rotating targets is a complicated task. There may be insufficient pulses or may introduce migration through range cells (MTRC) during the coherent processing interval (CPI) when we use the conventional range Doppler (RD) ISAR technique. With compressed sensing (CS) technique, we can achieve the high-resolution ISAR imaging of a target with limited number of pulses. Sparse representation based method can achieve the super resolution ISAR imaging of a target with a short CPI, during which the target rotates only a small angle and the range migration of the scatterers is small. However, traditional CS-based ISAR imaging method generally faced with the problem of basis mismatch, which may degrade the ISAR image. To achieve the high resolution ISAR imaging of fast rotating targets, this paper proposed a pattern-coupled sparse Bayesian learning method for multiple measurement vectors, i.e. the PC-MSBL algorithm. A multi-channel pattern-coupled hierarchical Gaussian prior is proposed to model the pattern dependencies among neighboring range cells and correct the MTRC problem. The expectation-maximization (EM) algorithm is used to infer the maximum a posterior (MAP) estimate of the hyperparameters. Simulation results validate the effectiveness and superiority of the proposed algorithm.     

A two-layered cross coupling control scheme for a three-dimensional motion control system

A two-layered modeling and compensation scheme is proposed to reduce the contouring error of a three-dimensional motion control system. In the proposed scheme, the contouring error model of the three-dimensional motion control system is divided into two layers: the top layer and the bottom layer. The proposed multi-layered structure of the contouring error model presents more flexibility in the control system design because the cross coupling controllers in different layers can be designed separately. In this paper, a nonlinear PI controller and a position error compensator are designed in the bottom layer in order to achieve high contouring accuracy in the XY plane, while a unilateral compensator is designed in the top layer to further reduce contouring error in the three dimensional space. Finally, experiments are performed to verify the performance of the proposed two-layered modeling and compensation scheme. Experiment results show that the designed two-layered cross coupling controller can obtain higher contouring accuracy than traditional cross coupling controller both in the XY plane and in the XYZ space.     

Content adaptive video denoising based on human visual perception

In this paper, we propose content adaptive denoising in highly corrupted videos based on human visual perception. We introduce the human visual perception in video denoising to achieve good performance. In general, smooth regions corrupted by noise are much more annoying to human observers than complex regions. Moreover, human eyes are more interested in complex regions with image details and more sensitive to luminance than chrominance. Based on the human visual perception, we perform perceptual video denoising to effectively preserve image details and remove annoying noise. To successfully remove noise and recover the image details, we extend nonlocal mean filtering to the spatiotemporal domain. With the guidance of content adaptive segmentation and motion detection, we conduct content adaptive filtering in the YUV color space to consider context in images and obtain perceptually pleasant results. Extensive experiments on various video sequences demonstrate that the proposed method reconstructs natural-looking results even in highly corrupted images and achieves good performance in terms of both visual quality and quantitative measures.     

基扩展算法在信道参数估计中的应用

针对如何减少信道参数估计所需要的已知数据,文中在基于扩展算法的基础上,采用复指数基函数,通过获取部分信道响应数据来计算得到其他信道数据,从而减少信道的实际测量次数。重点研究了基扩展算法在不同基函数阶数和不同传输符号块采样点数的情况下,模型准确度的变化,给出了要使模型具有佳性能的参数设定方案。     

Experimental optimization of power-function-shaped drive pulse for stick-slip piezo actuators

Motion of a stick-slip piezo actuator is generally controlled by the parameters related to its mechanical design and characteristics of the driving pulses applied to piezoceramic shear plates. The goal of the proposed optimization method is to find the driving pulse parameters leading to the fastest and the most reliable actuator operation. In the paper the method is tested on a rotary stick-slip piezo actuating system utilized in an atomic force microscope.The optimization is based on the measurement of the actuator response to driving pulses of different shapes and repetition frequencies at various load forces. To provide it, a computer controlled testing system generating the driving pulses, and detecting and recording the corresponding angular motion response of the actuator by a position sensitive photo detector (PSPD) in real time has been developed. To better understand and interpret the experimental results, supportive methods based on a simple analytical model and numerical simulations were used as well.In this way the shapes of the single driving pulses and values of the load force providing the biggest actuator steps were determined. Generally, the maximal steps were achieved for such a combination of the pulse shapes and load forces providing high velocities at the end of the sticking mode of the actuator motion and, at the same time, lower decelerations during the slipping mode.As for the multiple driving pulses, the pulse shapes and values of repetition frequency ensuring the sticking mode of the actuator motion during the pulse rise time together with the maximum average angular rotor velocity were specified. In this way the effective and stable operation conditions of the actuator were provided.In principle, the presented method can be applied for the testing and optimization of any linear or angular stick-slip actuator.     

Style-Controllable Speech-Driven Gesture Synthesis Using Normalising Flows

Automatic synthesis of realistic gestures promises to transform the fields of animation, avatars and communicative agents. In off-line applications, novel tools can alter the role of an animator to that of a director, who provides only high-level input for the desired animation; a learned network then translates these instructions into an appropriate sequence of body poses. In interactive scenarios, systems for generating natural animations on the fly are key to achieving believable and relatable characters. In this paper we address some of the core issues towards these ends. By adapting a deep learning-based motion synthesis method called MoGlow, we propose a new generative model for generating state-of-the-art realistic speech-driven gesticulation. Owing to the probabilistic nature of the approach, our model can produce a battery of different, yet plausible, gestures given the same input speech signal. Just like humans, this gives a rich natural variation of motion. We additionally demonstrate the ability to exert directorial control over the output style, such as gesture level, speed, symmetry and spacial extent. Such control can be leveraged to convey a desired character personality or mood. We achieve all this without any manual annotation of the data. User studies evaluating upper-body gesticulation confirm that the generated motions are natural and well match the input speech. Our method scores above all prior systems and baselines on these measures, and comes close to the ratings of the original recorded motions. We furthermore find that we can accurately control gesticulation styles without unnecessarily compromising perceived naturalness. Finally, we also demonstrate an application of the same method to full-body gesticulation, including the synthesis of stepping motion and stance.