Unconstrained monocular 3D human pose estimation by action detection and cross-modality regression forest

This work addresses the challenging problem of unconstrained 3D human pose estimation (HPE) from a novel perspective. Existing approaches struggle to operate in realistic applications, mainly due to their scene-dependent priors, such as background segmentation and multi-camera network, which restrict their use in unconstrained environments. We therfore present a framework which applies action detection and 2D pose estimation techniques to infer 3D poses in an unconstrained video. Action detection offers spatiotemporal priors to 3D human pose estimation by both recognising and localising actions in space-time. Instead of holistic features, e.g. silhouettes, we leverage the flexibility of deformable part model to detect 2D body parts as a feature to estimate 3D poses. A new unconstrained pose dataset has been collected to justify the feasibility of our method, which demonstrated promising results, significantly outperforming the relevant state-of-the-arts. © 2013 IEEE.

Single camera pose estimation using Bayesian filtering and Kinect motion priors

Traditional approaches to upper body pose estimation using monocular vision rely on complex body models and a large variety of geometric constraints. We argue that this is not ideal and somewhat inelegant as it results in large processing burdens, and instead attempt to incorporate these constraints through priors obtained directly from training data. A prior distribution covering the probability of a human pose occurring is used to incorporate likely human poses. This distribution is obtained offline, by fitting a Gaussian mixture model to a large dataset of recorded human body poses, tracked using a Kinect sensor. We combine this prior information with a random walk transition model to obtain an upper body model, suitable for use within a recursive Bayesian filtering framework. Our model can be viewed as a mixture of discrete Ornstein-Uhlenbeck processes, in that states behave as random walks, but drift towards a set of typically observed poses. This model is combined with measurements of the human head and hand positions, using recursive Bayesian estimation to incorporate temporal information. Measurements are obtained using face detection and a simple skin colour hand detector, trained using the detected face. The suggested model is designed with analytical tractability in mind and we show that the pose tracking can be Rao-Blackwellised using the mixture Kalman filter, allowing for computational efficiency while still incorporating bio-mechanical properties of the upper body. In addition, the use of the proposed upper body model allows reliable three-dimensional pose estimates to be obtained indirectly for a number of joints that are often difficult to detect using traditional object recognition strategies. Comparisons with Kinect sensor results and the state of the art in 2D pose estimation highlight the efficacy of the proposed approach.

Fast Pose Estimation with Parameter Sensitive Hashing

Example-based methods are effective for parameter estimation problems when the underlying system is simple or the dimensionality of the input is low. For complex and high-dimensional problems such as pose estimation, the number of required examples and the computational complexity rapidly becme prohibitively high. We introduce a new algorithm that learns a set of hashing functions that efficiently index examples relevant to a particular estimation task. Our algorithm extends a recently developed method for locality-sensitive hashing, which finds approximate neighbors in time sublinear in the number of examples. This method depends critically on the choice of hash functions; we show how to find the set of hash functions that are optimally relevant to a particular estimation problem. Experiments demonstrate that the resulting algorithm, which we call Parameter-Sensitive Hashing, can rapidly and accurately estimate the articulated pose of human figures from a large database of example images.

Database Indexing Methods for 3D Hand Pose Estimation

Estimation of 3D hand pose is useful in many gesture recognition applications, ranging from human-computer interaction to automated recognition of sign languages. In this paper, 3D hand pose estimation is treated as a database indexing problem. Given an input image of a hand, the most similar images in a large database of hand images are retrieved. The hand pose parameters of the retrieved images are used as estimates for the hand pose in the input image. Lipschitz embeddings of edge images into a Euclidean space are used to improve the efficiency of database retrieval. In order to achieve interactive retrieval times, similarity queries are initially performed in this Euclidean space. The paper describes ongoing work that focuses on how to best choose reference images, in order to improve retrieval accuracy.

3D Hand Pose Estimation by Finding Appearance-Based Matches in a Large Database of Training Views

Ongoing work towards appearance-based 3D hand pose estimation from a single image is presented. A large database of synthetic hand views is generated using a 3D hand model and computer graphics. The views display different hand shapes as seen from arbitrary viewpoints. Each synthetic view is automatically labeled with parameters describing its hand shape and viewing parameters. Given an input image, the system retrieves the most similar database views, and uses the shape and viewing parameters of those views as candidate estimates for the parameters of the input image. Preliminary results are presented, in which appearance-based similarity is defined in terms of the chamfer distance between edge images.

Multiplicative Kernels: Object Detection, Segmentation and Pose Estimation

Object detection is challenging when the object class exhibits large within-class variations. In this work, we show that foreground-background classification (detection) and within-class classification of the foreground class (pose estimation) can be jointly learned in a multiplicative form of two kernel functions. One kernel measures similarity for foreground-background classification. The other kernel accounts for latent factors that control within-class variation and implicitly enables feature sharing among foreground training samples. Detector training can be accomplished via standard SVM learning. The resulting detectors are tuned to specific variations in the foreground class. They also serve to evaluate hypotheses of the foreground state. When the foreground parameters are provided in training, the detectors can also produce parameter estimate. When the foreground object masks are provided in training, the detectors can also produce object segmentation. The advantages of our method over past methods are demonstrated on data sets of human hands and vehicles.

Articulated Pose Estimation in a Learned Smooth Space of Feasible Solutions

A learning based framework is proposed for estimating human body pose from a single image. Given a differentiable function that maps from pose space to image feature space, the goal is to invert the process: estimate the pose given only image features. The inversion is an ill-posed problem as the inverse mapping is a one to many process. Hence multiple solutions exist, and it is desirable to restrict the solution space to a smaller subset of feasible solutions. For example, not all human body poses are feasible due to anthropometric constraints. Since the space of feasible solutions may not admit a closed form description, the proposed framework seeks to exploit machine learning techniques to learn an approximation that is smoothly parameterized over such a space. One such technique is Gaussian Process Latent Variable Modelling. Scaled conjugate gradient is then used find the best matching pose in the space of feasible solutions when given an input image. The formulation allows easy incorporation of various constraints, e.g. temporal consistency and anthropometric constraints. The performance of the proposed approach is evaluated in the task of upper-body pose estimation from silhouettes and compared with the Specialized Mapping Architecture. The estimation accuracy of the Specialized Mapping Architecture is at least one standard deviation worse than the proposed approach in the experiments with synthetic data. In experiments with real video of humans performing gestures, the proposed approach produces qualitatively better estimation results.

Pose estimation of a humanoid robot using images from a mobile external camera

NOGUEIRA, Marcelo B. ; MEDEIROS, Adelardo A. D. ; ALSINA, Pablo J. Pose Estimation of a Humanoid Robot Using Images from an Mobile Extern Camera. In: IFAC WORKSHOP ON MULTIVEHICLE SYSTEMS, 2006, Salvador, BA. Anais... Salvador: MVS 2006, 2006.

A Multiresolution Image Alignment Technique Based on Direct Methods for Pose Estimation of Aerial Vehicles

In this paper, we seek to expand the use of direct methods in real-time applications by proposing a vision-based strategy for pose estimation of aerial vehicles. The vast majority of approaches make use of features to estimate motion. Conversely, the strategy we propose is based on a MR (Multi-Resolution) implementation of an image registration technique (Inverse Compositional Image Alignment ICIA) using direct methods. An on-board camera in a downwards-looking configuration, and the assumption of planar scenes, are the bases of the algorithm. The motion between frames (rotation and translation) is recovered by decomposing the frame-to-frame homography obtained by the ICIA algorithm applied to a patch that covers around the 80% of the image. When the visual estimation is required (e.g. GPS drop-out), this motion is integrated with the previous known estimation of the vehicles' state, obtained from the on-board sensors (GPS/IMU), and the subsequent estimations are based only on the vision-based motion estimations. The proposed strategy is tested with real flight data in representative stages of a flight: cruise, landing, and take-off, being two of those stages considered critical: take-off and landing. The performance of the pose estimation strategy is analyzed by comparing it with the GPS/IMU estimations. Results show correlation between the visual estimation obtained with the MR-ICIA and the GPS/IMU data, that demonstrate that the visual estimation can be used to provide a good approximation of the vehicle's state when it is required (e.g. GPS drop-outs). In terms of performance, the proposed strategy is able to maintain an estimation of the vehicle's state for more than one minute, at real-time frame rates based, only on visual information.

On-board and Ground Visual Pose Estimation Techniques for UAV Control

In this paper, two techniques to control UAVs (Unmanned Aerial Vehicles), based on visual information are presented. The first one is based on the detection and tracking of planar structures from an on-board camera, while the second one is based on the detection and 3D reconstruction of the position of the UAV based on an external camera system. Both strategies are tested with a VTOL (Vertical take-off and landing) UAV, and results show good behavior of the visual systems (precision in the estimation and frame rate) when estimating the helicopter¿s position and using the extracted information to control the UAV.

Visual Tracking, Pose Estimation, and Control for Aerial Vehicles

El principal objetivo de esta tesis es dotar a los vehículos aéreos no tripulados (UAVs, por sus siglas en inglés) de una fuente de información adicional basada en visión. Esta fuente de información proviene de cámaras ubicadas a bordo de los vehículos o en el suelo. Con ella se busca que los UAVs realicen tareas de aterrizaje o inspección guiados por visión, especialmente en aquellas situaciones en las que no haya disponibilidad de estimar la posición del vehículo con base en GPS, cuando las estimaciones de GPS no tengan la suficiente precisión requerida por las tareas a realizar, o cuando restricciones de carga de pago impidan añadir sensores a bordo de los vehículos. Esta tesis trata con tres de las principales áreas de la visión por computador: seguimiento visual y estimación visual de la pose (posición y orientación), que a su vez constituyen la base de la tercera, denominada control servo visual, que en nuestra aplicación se enfoca en el empleo de información visual para controlar los UAVs. Al respecto, esta tesis se ocupa de presentar propuestas novedosas que permitan solucionar problemas relativos al seguimiento de objetos mediante cámaras ubicadas a bordo de los UAVs, se ocupa de la estimación de la pose de los UAVs basada en información visual obtenida por cámaras ubicadas en el suelo o a bordo, y también se ocupa de la aplicación de las técnicas propuestas para solucionar diferentes problemas, como aquellos concernientes al seguimiento visual para tareas de reabastecimiento autónomo en vuelo o al aterrizaje basado en visión, entre otros. Las diversas técnicas de visión por computador presentadas en esta tesis se proponen con el fin de solucionar dificultades que suelen presentarse cuando se realizan tareas basadas en visión con UAVs, como las relativas a la obtención, en tiempo real, de estimaciones robustas, o como problemas generados por vibraciones. Los algoritmos propuestos en esta tesis han sido probados con información de imágenes reales obtenidas realizando pruebas on-line y off-line. Diversos mecanismos de evaluación han sido empleados con el propósito de analizar el desempeño de los algoritmos propuestos, entre los que se incluyen datos simulados, imágenes de vuelos reales, estimaciones precisas de posición empleando el sistema VICON y comparaciones con algoritmos del estado del arte. Los resultados obtenidos indican que los algoritmos de visión por computador propuestos tienen un desempeño que es comparable e incluso mejor al de algoritmos que se encuentran en el estado del arte. Los algoritmos propuestos permiten la obtención de estimaciones robustas en tiempo real, lo cual permite su uso en tareas de control visual. El desempeño de estos algoritmos es apropiado para las exigencias de las distintas aplicaciones examinadas: reabastecimiento autónomo en vuelo, aterrizaje y estimación del estado del UAV. Abstract The main objective of this thesis is to provide Unmanned Aerial Vehicles (UAVs) with an additional vision-based source of information extracted by cameras located either on-board or on the ground, in order to allow UAVs to develop visually guided tasks, such as landing or inspection, especially in situations where GPS information is not available, where GPS-based position estimation is not accurate enough for the task to develop, or where payload restrictions do not allow the incorporation of additional sensors on-board. This thesis covers three of the main computer vision areas: visual tracking and visual pose estimation, which are the bases the third one called visual servoing, which, in this work, focuses on using visual information to control UAVs. In this sense, the thesis focuses on presenting novel solutions for solving the tracking problem of objects when using cameras on-board UAVs, on estimating the pose of the UAVs based on the visual information collected by cameras located either on the ground or on-board, and also focuses on applying these proposed techniques for solving different problems, such as visual tracking for aerial refuelling or vision-based landing, among others. The different computer vision techniques presented in this thesis are proposed to solve some of the frequently problems found when addressing vision-based tasks in UAVs, such as obtaining robust vision-based estimations at real-time frame rates, and problems caused by vibrations, or 3D motion. All the proposed algorithms have been tested with real-image data in on-line and off-line tests. Different evaluation mechanisms have been used to analyze the performance of the proposed algorithms, such as simulated data, images from real-flight tests, publicly available datasets, manually generated ground truth data, accurate position estimations using a VICON system and a robotic cell, and comparison with state of the art algorithms. Results show that the proposed computer vision algorithms obtain performances that are comparable to, or even better than, state of the art algorithms, obtaining robust estimations at real-time frame rates. This proves that the proposed techniques are fast enough for vision-based control tasks. Therefore, the performance of the proposed vision algorithms has shown to be of a standard appropriate to the different explored applications: aerial refuelling and landing, and state estimation. It is noteworthy that they have low computational overheads for vision systems.

Fast human pose estimation using 3D Zernike descriptors

Markerless video-based human pose estimation algorithms face a high-dimensional problem that is frequently broken down into several lower-dimensional ones by estimating the pose of each limb separately. However, in order to do so they need to reliably locate the torso, for which they typically rely on time coherence and tracking algorithms. Their losing track usually results in catastrophic failure of the process, requiring human intervention and thus precluding their usage in real-time applications. We propose a very fast rough pose estimation scheme based on global shape descriptors built on 3D Zernike moments. Using an articulated model that we configure in many poses, a large database of descriptor/pose pairs can be computed off-line. Thus, the only steps that must be done on-line are the extraction of the descriptors for each input volume and a search against the database to get the most likely poses. While the result of such process is not a fine pose estimation, it can be useful to help more sophisticated algorithms to regain track or make more educated guesses when creating new particles in particle-filter-based tracking schemes. We have achieved a performance of about ten fps on a single computer using a database of about one million entries.