2.5D Elastic graph matching

In this paper, we propose novel elastic graph matching (EGM) algorithms for face recognition assisted by the availability of 3D facial geometry. More specifically, we conceptually extend the EGM algorithm in order to exploit the 3D nature of human facial geometry for face recognition/verification. In order to achieve that, first we extend the matching module of the EGM algorithm in order to capitalize on the 2.5D facial data. Furthermore, we incorporate the 3D geometry into the multiscale analysis used and build a novel geodesic multiscale morphological pyramid of dilations/erosions in order to fill the graph jets. We show that the proposed advances significantly enhance the performance of EGM algorithms. We demonstrate the efficiency of the proposed advances in the face recognition/verification problem using photometric stereo.     

Detecting end-effectors on 2.5D data using geometric deformable models: Application to human pose estimation

End-effectors are usually related to the location of limbs, and their reliable detection enables robust body tracking as well as accurate pose estimation. Recent innovation in depth cameras has re-stated the pose estimation problem. We focus on the information provided by these sensors, for which we borrow the name 2.5D data from the Graphics community. In this paper we propose a human pose estimation algorithm based on topological propagation. Geometric Deformable Models are used to carry out such propagation, implemented according to the Narrow Band Level Set approach. A variant of the latter method is proposed, including a density restriction which helps preserving the topological properties of the object under analysis. Principal end-effectors are extracted from a directed graph weighted with geodesic distances, also providing a skeletal-like structure describing human pose. An evaluation against reference methods is performed with promising results. The proposed solution allows a frame-wise end-effector detection, with no temporal tracking involved, which may be generalized to the tracking of other objects beyond human body.     

Learning templates from fuzzy examples in structural patternrecognition

Fuzzy-Attribute Graph (FAG) was proposed to handle fuzziness in the pattern primitives in structural pattern recognition. FAG has the advantage that we can combine several possible definitions into a single template, and hence only one matching is required instead of one for each definition. Also, each vertex or edge of the graph can contain fuzzy attributes to model real-life situations. However, in our previous approach, we need a human expert to define the templates for the fuzzy graph matching. This is usually tedious, time-consuming and error-prone. In this paper, we propose a learning algorithm that will, from a number of fuzzy examples, each of them being a FAG, find the smallest template that can be matched to the given patterns with respect to the matching metric.     

Linear response algorithms for approximate inference in graphical models

Belief propagation (BP) on cyclic graphs is an efficient algorithm for computing approximate marginal probability distributions over single nodes and neighboring nodes in the graph. However, it does not prescribe a way to compute joint distributions over pairs of distant nodes in the graph. In this article, we propose two new algorithms for approximating these pairwise probabilities, based on the linear response theorem. The first is a propagation algorithm that is shown to converge if BP converges to a stable fixed point. The second algorithm is based on matrix inversion. Applying these ideas to gaussian random fields, we derive a propagation algorithm for computing the inverse of a matrix.     

自适应骨骼中心的人体行为识别算法

目的 基于3维骨架的行为识别研究在计算机视觉领域一直是非常活跃的主题,在监控、视频游戏、机器人、人机交互、医疗保健等领域已取得了非常多的成果。现今的行为识别算法大多选择固定关节点作为坐标中心,导致动作识别率较低,为解决动作行为识别中识别精度低的问题,提出一种自适应骨骼中心的人体行为识别的算法。方法 该算法首先从骨骼数据集中获取三维骨架序列,并对其进行预处理,得到动作的原始坐标矩阵;再根据原始坐标矩阵提取特征,依据特征值的变化自适应地选择坐标中心,重新对原始坐标矩阵进行归一化;最后通过动态时间规划方法对动作坐标矩阵进行降噪处理,借助傅里叶时间金字塔表示的方法减少动作坐标矩阵时间错位和噪声问题,再使用支持向量机对动作坐标矩阵进行分类。论文使用国际上通用的数据集UTKinect-Action和MSRAction3D对算法进行验证。结果 结果表明,在UTKinect-Action数据集上,该算法的行为识别率比HO3D J2算法高4.28%,比CRF算法高3.48%。在MSRAction3D数据集上,该算法比HOJ3D算法高9.57%,比Profile HMM算法高2.07%,比Eigenjoints算法高6.17%。结论 本文针对现今行为识别算法的识别率低问题,探究出问题的原因是采用了固定关节坐标中心,提出了自适应骨骼中心的行为识别算法。经仿真验证,该算法能有效提高人体行为识别的精度。     

Median graph computation for graph clustering

In this paper, we are interested in the problem of graph clustering. We propose a new algorithm for computing the median of a set of graphs. The concept of median allows the extension of conventional algorithms such as the k-means to graph clustering, helping to bridge the gap between statistical and structural approaches to pattern recognition. Experimental results show the efficiency of the new median graph algorithm compared to the (only) existing algorithm in the literature. We also show its effective use in clustering a set of random graphs and in a content-based synthetic image retrieval system.     

基于深度信息的压缩感知人脸检测跟踪

传统的压缩感知跟踪是基于彩色视频图像序列中的目标跟踪, 但在跟踪过程中可能会受到光照变化和旋转遮挡因素的影响, 从而导致复杂环境下跟踪结果的鲁棒性不足. 为了获得稳定的跟踪结果, 提出了一种基于深度信息的压缩感知人脸检测跟踪算法. 首先, 根据改进的质心分割算法确定首帧深度图中人脸的跟踪位置. 其次, 根据深度信息计算出深度图中每一点对应的平均曲率并形成平均曲率图. 然后, 基于平均曲率图随机提取压缩特征; 最后, 通过压缩降维, 目标邻域搜索, 迭代更新特征模板, 计算出平均曲率图中下一帧人脸的最优跟踪位置, 实现人脸跟踪. 实验结果表明, 将人脸的深度信息和压缩感知特征相结合在光照变化和旋转遮挡情况下具有很好的鲁棒性, 能更好的实现复杂背景下对多姿态人脸的跟踪.     

基于人体特征三维人体模型的骨架提取算法

实现骨骼动画的一个前提是获取人体模型的骨架,现有的骨架提取算法不是计算复杂度高,就是提取准确度不高,或者需要手工干预.提出一种基于人体特点和黄金比例律的人体模型骨架提取算法,首先对模型进行精简,然后根据人体的特点与黄金比例律确定模型关节点的大概位置,在此基础上对模型进行分割.由于人体存在个体差异且姿势也可能不一致,采用测地距离方法对关节点的位置进行修正,确定其位置.与现有的算法相比,本方法效率高,同时实验显示本算法具有更好的骨架提取效果.     

Algorithm recognition for programming tutoring based on flow graph parsing

Plan-recognition with template matching shows reasonable performance for recognizing general control structures. However, problems containing well-defined algorithms such as sorting and searching are difficult to recognize by the template-based methodology alone, because an algorithm is often highly optimized and therefore hard to divide into smaller meaningful units. Based on this observation, we propose an algorithm recognition methodology to augment the plan-recognition approach.Our algorithm recognition approach is based on the flow graph parsing which performs a partial recognition of programs. The methodology is extended to completely understand algorithm implementations by providing the information of the program's goal, a specification of programming assignment in tutoring environment. Utilization of goal information is two-fold; extending the role of transformation rules to represent algorithm-specific information and looking for salient graph parts of algorithm plan to determine which has been used to implement the goal. Preliminary evaluation was performed on students' programs containing sort algorithm.     

基于时空权重姿态运动特征的人体骨架行为识别研究

人体行为识别在视觉领域的广泛应用使得它在过去的几十年里一直都是备受关注的研究热点.近些年来,深度传感器的普及以及基于深度图像实时骨架估测算法的提出,使得基于骨架序列的人体行为识别研究越来越吸引人们的注意.已有的研究工作大部分提取帧内骨架不同关节点的空间域信息和帧间骨架关节点的时间域信息来表征行为序列,但没有考虑到不同关节点和姿态对判定行为类别所起作用是不同的.因此本文提出了一种基于时空权重姿态运动特征的行为识别方法,采用双线性分类器迭代计算得到关节点和静止姿态相对于该类别动作的权重,确定那些信息量大的关节点和姿态;同时,为了对行为特征进行更好的时序分析,本文引入了动态时间规整和傅里叶时间金字塔算法进行时序建模,最后采用支持向量机完成行为分类.在多个数据集上的实验结果表明,该方法与其它一些方法相比,表现出了相当大的竞争力,甚至更好的识别效果.     

Graph embedding and extensions: a general framework for dimensionality reduction

A large family of algorithms - supervised or unsupervised; stemming from statistics or geometry theory - has been designed to provide different solutions to the problem of dimensionality reduction. Despite the different motivations of these algorithms, we present in this paper a general formulation known as graph embedding to unify them within a common framework. In graph embedding, each algorithm can be considered as the direct graph embedding or its linear/kernel/tensor extension of a specific intrinsic graph that describes certain desired statistical or geometric properties of a data set, with constraints from scale normalization or a penalty graph that characterizes a statistical or geometric property that should be avoided. Furthermore, the graph embedding framework can be used as a general platform for developing new dimensionality reduction algorithms. By utilizing this framework as a tool, we propose a new supervised dimensionality reduction algorithm called marginal Fisher analysis in which the intrinsic graph characterizes the intraclass compactness and connects each data point with its neighboring points of the same class, while the penalty graph connects the marginal points and characterizes the interclass separability. We show that MFA effectively overcomes the limitations of the traditional linear discriminant analysis algorithm due to data distribution assumptions and available projection directions. Real face recognition experiments show the superiority of our proposed MFA in comparison to LDA, also for corresponding kernel and tensor extensions     

基于MobileNetV3与ST-SRU的危险驾驶姿态识别

针对危险驾驶行为引起的交通安全事故频发的现状,提出一种基于MobileNetV3和ST-SRU的危险驾驶姿态识别系统.首先,修改MobileNetV3的网络结构使其适用于人体姿态估计任务,输出关节点的热力图和偏移量图,用来估计J个关节点的二维坐标位置;其次,定义ST-SRU骨架动作识别算法,利用动作的骨架序列数据对动作进行分类.实验结果表明:MobileNetV3姿态估计算法在自建的AI Challenger上肢姿态数据集上测得PCP值(percentage correct parts)达到95.6%,测试1 000次用时仅为5.03 s;利用自建的危险驾驶行为数据集将训练好的姿态估计和动作识别模型移植到嵌入式平台,实现了实时的危险驾驶姿态识别系统.     

3D human pose estimation in motion based on multi-stage regression

3D human pose estimation in motion is a hot research direction in the field of computer vision. However, the performance of the algorithm is affected by the complexity of 3D spatial information, self-occlusion of human body, mapping uncertainty and other problems. In this paper, we propose a 3D human joint localization method based on multi-stage regression depth network and 2D to 3D point mapping algorithm. First of all, we use a single RGB image as the input, through the introduction of heatmap and multi-stage regression to constantly optimize the coordinates of human joint points. Then we input the 2D joint points into the mapping network for calculation, and get the coordinates of 3D human body joint points, and then to complete the 3D human body pose estimation task. The MPJPE of the algorithm in Human3.6 M dataset is 40.7. The evaluation of dataset shows that our method has obvious advantages.     

Recognition of shapes by attributed skeletal graphs

In this paper, we propose a framework to address the problem of generic 2-D shape recognition. The aim is mainly on using the potential strength of skeleton of discrete objects in computer vision and pattern recognition where features of objects are needed for classification. We propose to represent the medial axis characteristic points as an attributed skeletal graph to model the shape. The information about the object shape and its topology is totally embedded in them and this allows the comparison of different objects by graph matching algorithms. The experimental results demonstrate the correctness in detecting its characteristic points and in computing a more regular and effective representation for a perceptual indexing. The matching process, based on a revised graduated assignment algorithm, has produced encouraging results, showing the potential of the developed method in a variety of computer vision and pattern recognition domains. The results demonstrate its robustness in the presence of scale, reflection and rotation transformations and prove the ability to handle noise and occlusions.     

基于图持续学习的时序数据分析

随着可穿戴设备大规模进入生活, 基于动作传感器产生的时序数据来人体行为识别已成为该领域的研究热点. 然而目前的方法无法发现多个传感器数据在时空中相互作用的关系. 此外, 传统神经网络在学习新任务时, 由于学习的新任务参数会覆盖掉旧任务参数, 这会引起“灾难性遗忘”问题. 为解决这两个问题, 本文提出了一种基于图注意力网络与生成式回放持续学习机制融合方法的人体行为识别算法. 该算法通过卷积神经网络与图注意力网络提取时序特征, 使得模型能够同时关注时间与空间特征, 同时, 采用了基于生成式数据重放策略的情景记忆持续学习方法, 通过条件变分自编码器记忆历史数据分布来解决灾难性遗忘问题. 最后, 通过在多个公开数据集上与不同的基线算法对比, 实验结果表明本文所提算法可以在取得较高的准确率的同时, 缓解灾难性遗忘问题.     

Human activity recognition using quasiperiodic time series collected from a single tri-axial accelerometer

The current generation of portable mobile devices incorporates various types of sensors that open up new areas for the analysis of human behavior. In this paper, we propose a method for human physical activity recognition using time series, collected from a single tri-axial accelerometer of a smartphone. Primarily, the method solves a problem of online time series segmentation, assuming that each meaningful segment corresponds to one fundamental period of motion. To extract the fundamental period we construct the phase trajectory matrix, applying the technique of principal component analysis. The obtained segments refer to various types of human physical activity. To recognize these activities we use the k-nearest neighbor algorithm and neural network as an alternative. We verify the accuracy of the proposed algorithms by testing them on the WISDM dataset of labeled accelerometer time series from thirteen users. The results show that our method achieves high precision, ensuring nearly 96 % recognition accuracy when using the bunch of segmentation and k-nearest neighbor algorithms.     

An Optimal Parallel Co-Connectivity Algorithm

In this paper we consider the problem of computing the connected components of the complement of a given graph. We describe a simple sequential algorithm for this problem, which works on the input graph and not on its complement, and which for a graph on n vertices and m edges runs in optimal O(n+m) time. Moreover, unlike previous linear co-connectivity algorithms, this algorithm admits efficient parallelization, leading to an optimal O(log n)-time and O((n+m)log n)-processor algorithm on the EREW PRAM model of computation. It is worth noting that, for the related problem of computing the connected components of a graph, no optimal deterministic parallel algorithm is currently available. The co-connectivity algorithms find applications in a number of problems. In fact, we also include a parallel recognition algorithm for weakly triangulated graphs, which takes advantage of the parallel co-connectivity algorithm and achieves an O(log² n) time complexity using O((n+m²) log n) processors on the EREW PRAM model of computation.     

Action recognition based on global optimal similarity measuring

Human action recognition based on the 3D skeleton is an important yet challenging task, because of the instability of skeleton joints and great variations in action length. In this paper we propose a novel method that can effectively deal with unstable joints and significant temporal misalignment. Action recognition is elegantly formulated as a sequence-matching problem on a pre-constructed weighted graph, which can encodes any spatio-temporal features and the transition probabilities between action elements. To classify any input sequence of actions, a global optimal matching algorithm based on dynamic programming is introduced, which can deal with temporal misalignment without pre-segmentation, The weighted graph is constructed in training stage. The proposed approach is evaluated on two benchmark datasets captured by a single depth sensor. Experimental results show that our approach can achieve superior performance to most state-of-the-art algorithms.     

Isometric 3D Shape Partial Matching Using GD-DNA

Isometric 3D shape partial matching has attracted a great amount of interest, with a plethora of applications ranging from shape recognition to texture mapping. In this paper, we propose a novel isometric 3D shape partial matching algorithm using the geodesic disk Laplace spectrum (GD-DNA). It transforms the partial matching problem into the geodesic disk matching problem. Firstly, the largest enclosed geodesic disk extracted from the partial shape is matched with geodesic disks from the full shape by the Laplace spectrum of the geodesic disk. Secondly, Generalized Multi-Dimensional Scaling algorithm (GMDS) and Euclidean embedding are conducted to establish final point correspondences between the partial and the full shape using the matched geodesic disk pair. The proposed GD-DNA is discriminative for matching geodesic disks, and it can well solve the anchor point selection problem in challenging partial shape matching tasks. Experimental results on the Shape Retrieval Contest 2016 (SHREC’16) benchmark validate the proposed method, and comparisons with isometric partial matching algorithms in the literature show that our method has a higher precision.     

Approximation techniques for response-time analysis of static-priority tasks

We consider sporadic tasks with static priorities and constrained deadlines to be executed upon a uniprocessor platform. Pseudo-polynomial time algorithms are known for computing worst-case response times for this task model. Some applications require to evaluate efficiently upper bounds of response times. For this purpose, we propose parametric algorithms that allow to make a tradeoff between quality of results and computational effort according to an input accuracy parameter. In this paper, we present a parametric polynomial-time algorithm for computing upper bounds of worst-case response times, that is based on an improved fptas (Fully Polynomial Time Approximation Scheme). Then, we show that our bound does not achieve constant error bound in comparison with the exact worst-case response time. However, using the resource augmentation technique, we obtain a performance guarantee that allows to define a compromise between our response-time bound and processor capacity requirements. The algorithm average behavior is then analyzed through numerical experimentations.