结合个性化建模和深度数据的三维人体姿态估计

利用深度传感器估计三维人体姿态是计算机视觉领域的一个重要问题,在人机交互、虚拟现实和动画设计等领域有重要的应用价值.针对该问题的主流方法是自底向上的方法,这类方法一般采用分类、回归或检索技术,可以直接从深度数据中估计三维肢体姿态,在人机交互中得到了很广泛的应用.但是这类方法依赖于大规模的姿态数据库,而且结果不够精确.本文提出一种结合个性化人体建模和深度数据的三维姿态估计方法,首先对运动对象建立三维虚拟人模型,然后利用该个性化的虚拟人模型与深度数据之间的点匹配关系构造姿态优化的目标函数,通过迭代优化目标函数,估计出与深度数据相吻合的三维姿态.与传统方法相比,本文方法不需要任何姿态数据库.实验表明,本文方法得到的结果更加精确.     

Multi-contour initial pose estimation for 3D registration

Reliable manipulation of everyday household objects is essential to the success of service robots. In order to accurately manipulate these objects, robots need to know objects’ full 6-DOF pose, which is challenging due to sensor noise, clutters, and occlusions. In this paper, we present a new approach for effectively guessing the object pose given an observation of just a small patch of the object, by leveraging the fact that many household objects can only keep stable on a planar surface under a small set of poses. In particular, for each stable pose of an object, we slice the object with horizontal planes and extract multiple cross-section 2D contours. The pose estimation is then reduced to find a stable pose whose contour matches best with that of the sensor data, and this can be solved efficiently by cross-correlation. Experiments on the manipulation tasks in the DARPA Robotics Challenge validate our approach. In addition, we also investigate our method’s performance on object recognition tasks raising in the challenge.     

Algorithms for matching 3D line sets

Matching two sets of lines is a basic tool that has applications in many computer vision problems such as scene registration, object recognition, motion estimation, and others. Line sets may be composed of infinitely long lines or finite length line segments. Depending on line lengths, three basic cases arise in matching sets of lines: 1) finite-finite, 2) finite-infinite, and 3) infinite-infinite. Case 2 has not been treated in the literature. For Cases 1 and 3, existing algorithms for matching 3D line sets are not completely satisfactory in that they either solve special situations, or give approximate solutions, or may not converge, or are not invariant with respect to coordinate system transforms. In this paper, we present new algorithms that solve exactly all three cases for the general situation. The algorithms are provably convergent and invariant to coordinate transforms. Experiments with synthetic and real 3D image data are reported.     

Robust 3D object registration without explicit correspondence using geometric integration

3D vision-guided manipulation of components is a key problem of industrial machine vision. In this paper, we focus on the localization and pose estimation of known industrial objects from 3D measurements delivered by a scanning sensor. Since local information extracted from these measurements is unreliable due to noise, spatially unstructured measurements and missing detections, we present a novel objective function for robust registration without using correspondence information, based on the likelihood of model points. Furthermore, by extending Runge–Kutta-type integration directly to the group of Euclidean transformation, we infer object pose by computing the gradient flow directly on the related manifold. Comparison of our approach to existing state of the art methods shows that our method is more robust against poor initializations while having comparable run-time performance.     

SoftPOSIT: Simultaneous Pose and Correspondence Determination

The problem of pose estimation arises in many areas of computer vision, including object recognition, object tracking, site inspection and updating, and autonomous navigation when scene models are available. We present a new algorithm, called SoftPOSIT, for determining the pose of a 3D object from a single 2D image when correspondences between object points and image points are not known. The algorithm combines the iterative softassign algorithm (Gold and Rangarajan, 1996; Gold et al., 1998) for computing correspondences and the iterative POSIT algorithm (DeMenthon and Davis, 1995) for computing object pose under a full-perspective camera model. Our algorithm, unlike most previous algorithms for pose determination, does not have to hypothesize small sets of matches and then verify the remaining image points. Instead, all possible matches are treated identically throughout the search for an optimal pose. The performance of the algorithm is extensively evaluated in Monte Carlo simulations on synthetic data under a variety of levels of clutter, occlusion, and image noise. These tests show that the algorithm performs well in a variety of difficult scenarios, and empirical evidence suggests that the algorithm has an asymptotic run-time complexity that is better than previous methods by a factor of the number of image points. The algorithm is being applied to a number of practical autonomous vehicle navigation problems including the registration of 3D architectural models of a city to images, and the docking of small robots onto larger robots.     

利用2D射影变换求解空间复杂结构的3D不变量

3D不变量作为不随姿态、视点等成像条件变化而变化的特征参量，可以广泛应用于计算机视觉的多重领域．通过分析2D射影变换矩阵求解的多种可能性，由单纯基于点集对应的思路扩展到利用点集、线集、点、线组合等其它方法，从而拓宽了建立两射影平面对应关系的应用条件．由此提出了一种基于多种点线组合构造虚元素的方法，结合实元素和虚元素可以巧妙提取空间复杂结构下的多种3D不变量，以用于目标识别和描述当中．实验结果验证了方法的有效性。     

3D video components generation using object tracking technique

2D-to-3D conversion that would be a solution of the lack of 3D contents has been a worthy and challenging research field. In this paper, we propose a computer interactive conversion method to capture components which is used to generate 3D sequences. First, we divide the key frame into foreground and background, and then label the objects by convenient computer interactive operation. Depth information of objects is labeled after segmentation. Second, we use object tracking technique which synthesizes the advantages of kernel-based mean shift tracker and contour tracker to accomplish object depth capture for non-key frame. Finally, all the 3D information is prepared to render 3D sequences. After all, we propose our future work direction: a 2D-to-3D system which can generate 3D sequence interactively.     

Pose estimation from multiple cameras based on Sylvester’s equation

In this paper, we introduce a method to estimate the object’s pose from multiple cameras. We focus on direct estimation of the 3D object pose from 2D image sequences. Scale-Invariant Feature Transform (SIFT) is used to extract corresponding feature points from adjacent images in the video sequence. We first demonstrate that centralized pose estimation from the collection of corresponding feature points in the 2D images from all cameras can be obtained as a solution to a generalized Sylvester’s equation. We subsequently derive a distributed solution to pose estimation from multiple cameras and show that it is equivalent to the solution of the centralized pose estimation based on Sylvester’s equation. Specifically, we rely on collaboration among the multiple cameras to provide an iterative refinement of the independent solution to pose estimation obtained for each camera based on Sylvester’s equation. The proposed approach to pose estimation from multiple cameras relies on all of the information available from all cameras to obtain an estimate at each camera even when the image features are not visible to some of the cameras. The resulting pose estimation technique is therefore robust to occlusion and sensor errors from specific camera views. Moreover, the proposed approach does not require matching feature points among images from different camera views nor does it demand reconstruction of 3D points. Furthermore, the computational complexity of the proposed solution grows linearly with the number of cameras. Finally, computer simulation experiments demonstrate the accuracy and speed of our approach to pose estimation from multiple cameras.     

Alignment of continuous video onto 3D point clouds

We propose a general framework for aligning continuous (oblique) video onto 3D sensor data. We align a point cloud computed from the video onto the point cloud directly obtained from a 3D sensor. This is in contrast to existing techniques where the 2D images are aligned to a 3D model derived from the 3D sensor data. Using point clouds enables the alignment for scenes full of objects that are difficult to model; for example, trees. To compute 3D point clouds from video, motion stereo is used along with a state-of-the-art algorithm for camera pose estimation. Our experiments with real data demonstrate the advantages of the proposed registration algorithm for texturing models in large-scale semi-urban environments. The capability to align video before a 3D model is built from the 3D sensor data offers new practical opportunities for 3D modeling. We introduce a novel modeling-through-registration approach that fuses 3D information from both the 3D sensor and the video. Initial experiments with real data illustrate the potential of the proposed approach.     

Semantic part segmentation method based 3D object pose estimation with RGB-D images for bin-picking

3D object pose estimation for grasping and manipulation is a crucial task in robotic and industrial applications. Robustness and efficiency for robotic manipulation are desirable properties that are still very challenging in complex and cluttered scenes, because 3D objects have different appearances, illumination and occlusion when seen from different viewpoints. This article proposes a Semantic Point Pair Feature (PPF) method for 3D object pose estimation, which combines the semantic image segmentation using deep learning with the voting-based 3D object pose estimation. The Part Mask RCNN ispresented to obtain the semantic object-part segmentation related to the point cloud of object, which is combined with the PPF method for 3D object pose estimation. In order to reduce the cost of collecting datasets in cluttered scenes, a physically-simulated environment is constructed to generate labeled synthetic semantic datasets. Finally, two robotic bin-picking experiments are demonstrated and the Part Mask RCNN for scene segmentation is evaluated through the constructed 3D object datasets. The experimental results show that the proposed Semantic PPF methodimproves the robustness and efficiency of 3D object pose estimation in cluttered scenes with partial occlusions.     

3D SSD tracking with estimated 3D planes

We present a tracking method where full camera position and orientation is tracked from intensity differences in a video sequence. The camera pose is calculated based on 3D planes, and hence does not depend on point correspondences. The plane based formulation also allows additional constraints to be naturally added, e.g., perpendicularity between walls, floor and ceiling surfaces, co-planarity of wall surfaces etc. A particular feature of our method is that the full 3D pose change is directly computed from temporal image differences without making a commitment to a particular intermediate (e.g., 2D feature) representation. We experimentally compared our method with regular 2D SSD tracking and found it more robust and stable. This is due to 3D consistency being enforced even in the low level registration of image regions. This yields better results than first computing (and hence committing to) 2D image features and then from these compute 3D pose.     

Linear N-point camera pose determination

The determination of camera position and orientation from known correspondences of 3D reference points and their images is known as pose estimation in computer vision and space resection in photogrammetry. It is well-known that from three corresponding points there are at most four algebraic solutions. Less appears to be known about the cases of four and five corresponding points. We propose a family of linear methods that yield a unique solution to 4- and 5-point pose determination for generic reference points. We first review the 3-point algebraic method. Then we present our two-step, 4-point and one-step, 5-point linear algorithms. The 5-point method can also be extended to handle more than five points. Finally, we demonstrate our methods on both simulated and real images. We show that they do not degenerate for coplanar configurations and even outperform the special linear algorithm for coplanar configurations in practice     

基于深度学习的二维人体姿态估计综述

人体姿态估计是计算机视觉领域的一个基础且具有挑战的任务,人体姿态估计对于描述人体姿态、描述人体行为等至关重要,是行为识别、行为检测等计算机视觉任务的基础.近年来,随着深度学习的发展,基于深度学习的人体姿态估计算法展现出了极其优异的效果.从单人人体姿态估计、自顶向下的多人人体姿态估计和自底向上的多人人体姿态估计这3种主流的人体姿态估计方式,介绍近年来基于深度学习的二维人体姿态估计算法的发展,并讨论目前二维人体姿态估计所面临的困难和挑战.最后,对人体姿态估计未来的发展做出展望.     

SDF-2-SDF Registration for Real-Time 3D Reconstruction from RGB-D Data

We tackle the task of dense 3D reconstruction from RGB-D data. Contrary to the majority of existing methods, we focus not only on trajectory estimation accuracy, but also on reconstruction precision. The key technique is SDF-2-SDF registration, which is a correspondence-free, symmetric, dense energy minimization method, performed via the direct voxel-wise difference between a pair of signed distance fields. It has a wider convergence basin than traditional point cloud registration and cloud-to-volume alignment techniques. Furthermore, its formulation allows for straightforward incorporation of photometric and additional geometric constraints. We employ SDF-2-SDF registration in two applications. First, we perform small-to-medium scale object reconstruction entirely on the CPU. To this end, the camera is tracked frame-to-frame in real time. Then, the initial pose estimates are refined globally in a lightweight optimization framework, which does not involve a pose graph. We combine these procedures into our second, fully real-time application for larger-scale object reconstruction and SLAM. It is implemented as a hybrid system, whereby tracking is done on the GPU, while refinement runs concurrently over batches on the CPU. To bound memory and runtime footprints, registration is done over a fixed number of limited-extent volumes, anchored at geometry-rich locations. Extensive qualitative and quantitative evaluation of both trajectory accuracy and model fidelity on several public RGB-D datasets, acquired with various quality sensors, demonstrates higher precision than related techniques.     

3D free-form surface registration and object recognition

A new technique to recognise 3D free-form objects via registration is proposed. This technique attempts to register a free-form surface, represented by a set of % MathType!MTEF!2!1!+-% feaafeart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaGOmamaala% aabaGaaGymaaqaaiaaikdaaaGaamiraaaa!38F8!\[2\frac{1}{2}D\] sensed data points, to the model surface, represented by another set of % MathType!MTEF!2!1!+-% feaafeart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaGOmamaala% aabaGaaGymaaqaaiaaikdaaaGaamiraaaa!38F8!\[2\frac{1}{2}D\] model data points, without prior knowledge of correspondence or view points between the two point sets. With an initial assumption that the sensed surface be part of a more complete model surface, the algorithm begins by selecting three dispersed, reliable points on the sensed surface. To find the three corresponding model points, the method uses the principal curvatures and the Darboux frames to restrict the search over the model space. Invariably, many possible model 3-typles will be found. For each hypothesized model 3-tuple, the transformation to match the sensed 3-tuple to the model 3-tuple can be determined. A heuristic search is proposed to single out the optimal transformation in low order time. For realistic object recognition or registration, where the two range images are often extracted from different view points of the model, the earlier assumption that the sensed surface be part of a more complete model surface cannot be relied on. With this, the sensed 3-tuple must be chosen such that the three sensed points lie on the common region visible to both the sensed and model views. We propose an algorithm to select a minimal non-redundant set of 3-tuples such that at least one of the 3-tuples will lie on the overlap. Applying the previous algorithm to each 3-tuple within this set, the optimal transformation can be determined. Experiments using data obtained from a range finder have indicated fast registration for relatively complex test cases. If the optimal registrations between the sensed data (candidate) and each of a set of model data are found, then, for 3D object recognition purposes, the minimal best fit error can be used as the decision rule.     

Global feature space neural network for active computer vision

We advance new active computer vision algorithms based on the Feature space Trajectory (FST) representations of objects and a neural network processor for computation of distances in global feature space. Our algorithms classify rigid objects and estimate their pose from intensity images. They also indicate how to automatically reposition the sensor if the class or pose of an object is ambiguous from a given viewpoint and they incorporate data from multiple object views in the final object classification. An FST in a global eigenfeature space is used to represent 3D distorted views of an object. Assuming that an observed feature vector consists of Gaussian noise added to a point on the FST, we derive a probability density function for the observation conditioned on the class and pose of the object. Bayesian estimation and hypothesis testing theory are then used to derive approximations to the maximum a posterioriprobability pose estimate and the minimum probability of error classifier. Confidence measures for the class and pose estimates, derived using Bayes theory, determine when additional observations are required, as well as where the sensor should be positioned to provide the most useful information.     

Uncalibrated multi-view multiple humans association and 3D pose estimation by adversarial learning

Multimedia Tools and Applications - Multiple human 3D pose estimation is a useful but challenging task in computer vison applications. The ambiguities in estimation of 2D and 3D poses of multiple...     

利用角度签名进行3维表面配准

3维表面的配准在3维物体重建、场景检测和物体识别过程中起着重要的作用。为此提出了一种新的3维表面表示方法——角度签名(angle signature),并将其用于3维表面配准。该表示方法将表面的局部几何信息表示成为1维的向量,具有对刚体变换的不变性。由于其简洁的表示方式,可以实现表面的快速配准。此外,该方法较其他3维表面的表示方法具有更强的鲁棒性。在实际应用中,为了提高表面配准的速度,首先筛选出特征点,然后利用特征点寻找表面之间的对应关系,从而将刚体变换的参数求出,实现表面的配准。实验结果表明,采用角度签名实现物体表面配准具有较快的速度和较高的精度。