基于关键点的类别级三维可形变目标姿态估计

为了解决类别级三维可形变目标姿态估计问题,基于目标的关键点,提出了一种面向类别的三维可形变目标姿态估计方法。该方法设计了一种基于关键点的端到端深度学习框架,框架以PointNet++为后端网络,通过特征提取、部位分割、关键点提取和基于关键点的姿态估计部分实现可形变目标的姿态估计,具有计算精度高、鲁棒性强等优势。同时,基于ANCSH方法设计了适用于K-AOPE网络的关键点标准化分层表示方法,该方法仅需提取目标少量的关键点即可表示类别物体。为了验证方法的有效性,在公共数据集shape2motion上进行测试。实验结果显示,提出的姿态估计方法(以眼镜类别为例)在旋转角上的误差分别为2.3°、3.1°、3.7°,平移误差分别为0.034、0.030、0.046,连接状态误差为2.4°、2.5°,连接参数误差为1.2°、0.9°,0.008、0.010。与ANCSH方法相比,所提方法具有较高的准确性和鲁棒性。     

Accurate constrained pose estimation for small objects

The paper proposes a novel method for accurate pose estimation of small objects. A range sensor is used to find object orientation that constraints a PnP solver. The method finds the global minimum of a cost function (as opposed to traditional PnP algorithms) in closed form. This approach was used to enable robust robotic operation with sub millimeter precision.     

基于点云采样权重估计的未知物体抓取位姿生成方法

针对机械臂在非结构环境中对未知物体抓取位姿生成困难及抓取稳定性差的问题,提出一种基于点云采样权重估计的抓取位姿生成方法.首先通过移动深度相机的方式拼接得到较完整的物体点云信息,并对物体的几何特性进行分析,有效避开物体不宜抓取的位置进行抓取位姿样本生成;然后结合几何约束条件实现抓取位姿搜索,并利用力封闭条件对样本稳定性进行评估;最后为了对实际的抓取位姿进行评价,根据其稳定性、夹取深度、夹取角度等设定抓取可行性指标,据此在工作空间输出最佳抓取位姿并完成指定的抓取任务.实验结果表明,采用所提方法能够高效生成大量且稳定的抓取位姿,并在仿真环境中有效实现机械臂对单个或多个随机摆放的未知物体的抓取任务.     

Head pose estimation in solving human-computer interaction problems

A system that estimates a user’s head pose by its video image is described. The system includes three basic algorithms, that is, segmentation, landmark detection, and motion direction estimation. The dynamics of change in geometric relations between facial landmarks on a sequence of frames is used to determine the direction of head motion. The change in the angle between the lines that connect the corners of the eyes and the tip of nose is shown to be similar to the yaw angle in their dynamics of change. The described system operates in real time (7 frames per second) and ensures a high precision estimate of motion direction (p = 0.95).     

跨阶段结构下的人体姿态估计

目的 基于图像的人体姿态估计是计算机视觉领域中一个非常重要的研究课题,并广泛应用于人机交互、监控以及图像检索等方面。但是,由于人体视觉外观的多样性、遮挡和混杂背景等因素的影响,导致人体姿态估计问题一直是计算机视觉领域的难点和热点。本文主要关注于初始特征对关节点定位的作用,提出一种跨阶段卷积姿态机（CSCPM）。方法 首先,采用VGG （visual geometry group）网络获得初步的图像初始特征,该初始特征既是图像关节点定位的基础,同时,也由于受到自遮挡和混杂背景的干扰难以学习。其次,在初始特征的基础上,构建多层模型学习不同尺度下的结构特征,同时为了解决深度学习中的梯度消失问题,在后续的各层特征中都串联该初始特征。最后,设计了多尺度关节点定位的联合损失,用于学习深度网络参数。结果 本文实验在两大人体姿态数据集MPII （MPII human pose dataset）和LSP （leeds sport pose）上分别与近3年的人体姿态估计方法进行了定性与定量比较,在MPII数据集中,模型的总检测率为89.1%,相比于性能第2的模型高出了0.7%;在LSP数据集中,模型的总检测率为91.0%,相比于性能第2的模型高出了0.5%。结论 实验结果表明,初始特征学习能够有效判断关节点的自遮挡和混杂背景干扰情况,引入跨阶段结构的CSCPM姿态估计模型能够胜出现有人体姿态估计模型。     

基于高分辨率网络的人体姿态估计方法

人体姿态估计在人机交互和行为识别应用中起着至关重要的作用,但人体姿态估计方法在特征图尺度变化中难以预测正确的人体姿态.为了提高姿态估计的准确性,将并行网络多尺度融合方法和生成高质量特征图的方法结合进行人体姿态估计(RefinedHRNet).在人体检测基础之上,采用并行网络多尺度融合方法在阶段内采用空洞卷积模块来扩大感...     

融合注意力特征的遮挡物体6D姿态估计

在机械臂视觉抓取过程中,现有的算法在复杂背景、光照不足、遮挡等条件下,难以对目标物体进行实时、准确、鲁棒的姿态估计。针对以上问题,提出一种基于关键点方法的融合注意力特征的物体6D姿态网络。首先,在跳跃连接（Skip Connection）阶段引入能够聚焦通道空间信息的卷积注意力模块（CBAM）,使编码阶段的浅层特征与解码阶段的深层特征进行有效融合,增强特征图的空间域信息和精确位置通道信息;其次,采用归一化损失函数以弱监督的方式回归每个关键点的注意力图,将注意力图作为对应像素位置上关键点偏移量的权重分数;最后,累加求和得到关键点坐标。实验结果证明,所提网络在LINEMOD数据集和Occlusion LINEMOD数据集上ADD(-S)指标分别达到了91.3%和46.3%。与基于关键点的逐像素投票网络（PVNet）相比ADD(-S)指标分别提升了5.0个百分点和5.5个百分点,验证了所提网络在遮挡场景下有更好的鲁棒性。     

基于改进LPP的头部姿态估计方法

针对无监督局部保持投影算法在头部姿态估计上的高误差性和对噪音的敏感性的问题,提出一种鲁棒的局部保持投影算法。其基本出发点是先对训练的头部姿态加以姿态标注,并获得各个头部姿态间的偏置距离,再对所有头部姿态样本点进行异常值的度量,训练出较好的线性映射矩阵。实验结果表明,改进的方法比传统的LPP在头部姿态估计上取得较好的效果。     

Human head pose estimation based on HF method

This Paper addresses the problem of head pose estimation. Driving assistance technology utilizes head pose estimation as an indicator for visual focus and mental attention of the driver. Head pose estimation detects head orientation with respect to the camera. Model based and appearance-based methods are the two approaches in head pose estimation. The first approach uses the facial features to create a face geometrical models whereas the second method only takes into consideration the entire face image. The proposed appearance-based method work is performed using Hough transform and random forest to classify ninety-three classes of Hough values in order to find the exact head pose. The performance of the proposed work is evaluated based on accuracy and the time taken to detect the head pose. The paper outperforms many other previous works.     

基于注意力机制的轻量型人体姿态估计

针对高分辨率人体姿态估计网络存在参数量大、运算复杂度高等问题,提出一种基于高分辨率网络（HRNet）的轻量型沙漏坐标注意力网络（SCANet）用于人体姿态估计。首先引入沙漏（Sandglass）模块和坐标注意力（CoordAttention）模块;然后在此基础上构建了沙漏坐标注意力瓶颈（SCAneck）模块和沙漏坐标注意力基础 （SCAblock）模块两种轻量型模块,在降低模型参数量和运算复杂度的同时,获取特征图空间方向的长程依赖和精确位置信息。实验结果显示,在相同图像分辨率和环境配置的情况下,在COCO（Common Objects in COntext）校验集上,SCANet模型与HRNet模型相比参数量降低了52.6%,运算复杂度降低了60.6%;在MPII（Max Planck Institute for Informatics）校验集上,SCANet模型与HRNet模型相比参数量和运算复杂度分别降低了52.6%和61.1%;与常见的人体姿态估计网络如堆叠沙漏网络（Hourglass）、级联金字塔网络（CPN）和SimpleBaseline相比,SCANet模型在拥有更少的参数量与运算复杂度的情况下,仍能实现对人体关键点的高准确度预测。     

基于姿态估计的实时跌倒检测算法

为了快速、准确地检测老年人跌倒事件的发生,给出一种基于姿态估计的实时跌倒检测算法.首先利用基于深度学习的人体姿态估计算法得到人体关节点的坐标;然后,通过计算人体质心点的下降速度、跌倒后颈部关节点的纵坐标值是否大于阈值、以及肩部和腰部关节点的相对位置关系来判断跌倒是否发生.所给出的跌倒检测算法利用单目相机进行检测,便于以嵌入式方式应用于机器人.实验结果表明,所提出算法与当前先进方法相比取得了较好的效果.     

基于点特征的位姿测量系统鲁棒性分析

从工程应用的角度出发,在P3P问题有唯一解的条件下,研究位姿误差对图像坐标检测误差的鲁棒性。通过简化条件下的理论推导并结合一般条件下的仿真实验,证明了由图像坐标检测误差引起的位姿误差随摄像机焦比的增大而减小。与主点坐标无关,随着目标模型中特征点间距离的增大而减小,随着测量距离的增大而增大等结论。这些结论对位姿测量系统的设计具有一定的指导意义。     

并联化高分辨网络的人体姿态估计方法

为降低高分辨网络模型因级联过多造成的参数冗余,设计一种用于人体姿态估计的并联化高分辨网络模型。通过适当的网络裁剪和并联化结构优化处理策略,对原始高分辨网络模型的消融实验结果进行分析,保留网络的前3个阶段,按第四阶段网络的分辨率构建等尺度的U型网络结构,在多尺度分支的第二阶段与第三阶段中分别并联该结构,保持原网络在同分辨率下的特征融合。针对人体姿态估计数据集MPII与MSCOCO进行检测,并联化高分辨网络的参数量仅为原网络的37%,浮点运算量降低30%。     

Instance segmentation based 6D pose estimation of industrial objects using point clouds for robotic bin-picking

3D object pose estimation for robotic grasping and manipulation is a crucial task in the manufacturing industry. In cluttered and occluded scenes, the 6D pose estimation of the low-textured or textureless industrial object is a challenging problem due to the lack of color information. Thus, point cloud that is hardly affected by the lighting conditions is gaining popularity as an alternative solution for pose estimation. This article proposes a deep learning-based pose estimation using point cloud as input, which consists of instance segmentation and instance point cloud pose estimation. The instance segmentation divides the scene point cloud into multiple instance point clouds, and each instance point cloud pose is accurately predicted by fusing the depth and normal feature maps. In order to reduce the time consumption of the dataset acquisition and annotation, a physically-simulated engine is constructed to generate the synthetic dataset. Finally, several experiments are conducted on the public, synthetic and real datasets to verify the effectiveness of the pose estimation network. The experimental results show that the point cloud based pose estimation network can effectively and robustly predict the poses of objects in cluttered and occluded scenes.     

基于全局姿态感知的轻量级人体姿态估计

人体姿态估计是近年来人机交互领域的热点话题.当前,常见的人体姿态估计方法集中在通过增加网络的复杂性来提高精度,却忽视了模型的效益问题,导致模型在实际应用中精度高但计算资源消耗巨大.针对这一问题设计了一个基于全局姿态感知的轻量级人体姿态估计模型,其在MSCOCO数据集上精度达68.2％AP,速度保持在255 fps,参数...     

基于单幅图像的人脸三维姿态估计

针对单张人像的三维姿态计算,结合面貌测量和射影几何的理论提出了一种方法：首先在人面部的平面区域内,选取眼角点,口角点,鼻翼点建立人脸模型;然后根据人脸平面上两个相互垂直的特征线投影到照片上的灭点位置,求出人脸平面的旋转方向。该方法特征点易于标定,且无需任何的辅助设备和先验知识,具有一定的实用性。     

基于堆叠沙漏网络改进的多人姿态估计

为解决多人姿态估计中小尺度关键点(手腕、脚踝等)定位准确率低的问题,采用自顶向上的方式,结合先进的人体目标检测模型YOLOv3,提出一种基于堆叠沙漏网络改进的多人姿态估计方法。在沙漏网络中融入通道混洗模块,加强不同尺度下多层特征之间的跨通道信息交流,提高被遮挡关键点的识别效果;使用注意力机制对沙漏网络原有的残差模块进行特征增强,抑制无用特征并提升有用特征,提高小尺度关键点的识别率。实验结果表明,在MPII数据集上的总体PCK@0.5达到了88.6%,在MSCOCO数据集上的AP@0.75相比原始网络提升了4.6%,验证了所提方法的有效性。     

一种基于图结构模型的人体姿态估计算法

针对单幅图片中人体姿态的估计问题,在图结构模型的基础上提出了一种新的人体姿态估计算法。算法提出了一个新的部位观测模型和一种新的减小部位状态空间的方法：（1）对人体不同部位采用不同尺寸的细胞单元计算HOG特征,并利用线性SVM进行分类,从而提出一种新的部位观测模型;（2）利用人体部位定位的先验分布确定部位定位区域,然后通过邻域归并和设置与部位模板的匹配度阈值进一步减小状态空间,从而提出了一种减小部位状态空间的方法。仿真实验结果表明所提算法与传统算法相比更加有效。     

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.     

Cooperative pose estimation of a fleet of robots based on interactive points alignment

Given a fleet of autonomous robots performing a cooperative task, such as rescue of people, it is crucial for the robots to share their relative position. If the site has not been explored before, auto localise each robot through landmarks is not possible. Moreover, not always the GPS information is available or it has the desirable accuracy. Our framework is composed of a fleet of robots that have 2D and 3D cameras, a human coordinator and a Human–Machine Interface. 3D-images are used to automatically align them and deduct the relative position between robots. 2D-images are used to reduce the alignment error in an interactive manner. A human visualises both 2D-images and the current automatic alignment and imposes a new alignment through the Human–Machine Interface. Since the information is shared through the whole fleet, robots can deduct the position of other ones that do not visualise the same scene. Practical evaluation shows that in situations that there is a large difference between images, the cooperative and interactive processes are crucial to achieve an acceptable result.