Camera Viewpoint Research Articles

Directional pedestrian counting with a hybrid map-based model

Detection-based pedestrian counting methods produce results of considerable accuracy in non-crowded scenes. However, the detection-based approach is dependent on the camera viewpoint. On the other hand, map-based pedestrian counting methods are performed by measuring features that do not require separate detection of each pedestrian in the scene. Thus, these methods are more effective especially in high crowd density. In this paper, we propose a hybrid map-based model that is a new directional pedestrian counting model. Our proposed model is composed of direction estimation module with classified foreground motion vectors, and pedestrian counting module with principal component analysis. Our contributions in this paper have two aspects. First, we present a directional moving pedestrian counting system that does not depend on object detection or tracking. Second, the number and major directions of pedestrian movements can be detected, by classifying foreground motion vectors. This representation is more powerful than simple features in terms of handling noise, and can count the moving pedestrians in images more accurately.

Towards automated visual surveillance using gait for identity recognition and tracking across multiple non-intersecting cameras

Despite the fact that personal privacy has become a major concern, surveillance technology is now becoming ubiquitous in modern society. This is mainly due to the increasing number of crimes as well as the essential necessity to provide secure and safer environment. Recent research studies have confirmed now the possibility of recognizing people by the way they walk i.e. gait. The aim of this research study is to investigate the use of gait for people detection as well as identification across different cameras. We present a new approach for people tracking and identification between different non-intersecting un-calibrated stationary cameras based on gait analysis. A vision-based markerless extraction method is being deployed for the derivation of gait kinematics as well as anthropometric measurements in order to produce a gait signature. The novelty of our approach is motivated by the recent research in biometrics and forensic analysis using gait. The experimental results affirmed the robustness of our approach to successfully detect walking people as well as its potency to extract gait features for different camera viewpoints achieving an identity recognition rate of 73.6 % processed for 2270 video sequences. Furthermore, experimental results confirmed the potential of the proposed method for identity tracking in real surveillance systems to recognize walking individuals across different views with an average recognition rate of 92.5 % for cross-camera matching for two different non-overlapping views.

A Review of Camera Viewpoint Automation in Robotic and Laparoscopic Surgery

Complex teleoperative tasks, such as surgery, generally require human control. However, teleoperating a robot using indirect visual information poses many technical challenges because the user is expected to control the movement(s) of the camera(s) in addition to the robot’s arms and other elements. For humans, camera positioning is difficult, error-prone, and a drain on the user’s available resources and attention. This paper reviews the state of the art of autonomous camera control with a focus on surgical applications. We also propose potential avenues of research in this field that will support the transition from direct slaved control to truly autonomous robotic camera systems.

Rate-constrained 3D surface estimation from noise-corrupted multiview depth videos.

Transmitting compactly represented geometry of a dynamic 3D scene from a sender can enable a multitude of imaging functionalities at a receiver, such as synthesis of virtual images at freely chosen viewpoints via depth-image-based rendering. While depth maps—projections of 3D geometry onto 2D image planes at chosen camera viewpoints-can nowadays be readily captured by inexpensive depth sensors, they are often corrupted by non-negligible acquisition noise. Given depth maps need to be denoised and compressed at the encoder for efficient network transmission to the decoder, in this paper, we consider the denoising and compression problems jointly, arguing that doing so will result in a better overall performance than the alternative of solving the two problems separately in two stages. Specifically, we formulate a rate-constrained estimation problem, where given a set of observed noise-corrupted depth maps, the most probable (maximum a posteriori (MAP)) 3D surface is sought within a search space of surfaces with representation size no larger than a prespecified rate constraint. Our rate-constrained MAP solution reduces to the conventional unconstrained MAP 3D surface reconstruction solution if the rate constraint is loose. To solve our posed rate-constrained estimation problem, we propose an iterative algorithm, where in each iteration the structure (object boundaries) and the texture (surfaces within the object boundaries) of the depth maps are optimized alternately. Using the MVC codec for compression of multiview depth video and MPEG free viewpoint video sequences as input, experimental results show that rate-constrained estimated 3D surfaces computed by our algorithm can reduce coding rate of depth maps by up to 32% compared with unconstrained estimated surfaces for the same quality of synthesized virtual views at the decoder.

View invariant action recognition using generalized 4D features

In this paper, we propose a method to recognize human actions independently of viewpoints by developing 4D space–time features which can generalize the information from a finite number of views in training phase so as to show a satisfactory performance in arbitrary testing views. This 4D space–time interest points (4D-STIPs, [x,y,z,t]) are extracted using 3D space volumes reconstructed from images of a finite number of different views. Since the proposed features are constructed using volumetric information, the features for arbitrary 2D space viewpoint in testing can be generated by projecting the 3D space volumes and 4D-STIPs on corresponding test image planes. This enables action recognition in any camera viewpoint even after training with images from only a finite number of views. We also propose the variant of 3D space–time interest points, which take into account the simultaneous gradient variation in all 3 dimensions to focus on the motion of important spatial corner points. 3D space volumes and 4D-STIPs can be projected to arbitrary viewpoints for training each action to get generalization capability of the classifier. With these projected features, we construct motion history images and non-motion history images which encode the moving and non-moving parts of an action respectively. After reducing the feature dimension, the final features are learned by support vector data description method. In experiments, we train the models using IXMAS dataset constructed from five views and test them with a new SNU dataset made for evaluating the generalization performance for arbitrary view videos.

Analysis of Using Metric Access Methods for Visual Search of Objects in Video Databases

This article presents an approach to object retrieval that searches for and localizes all the occurrences of an object in a video database, given a query image of the object. Our proposal is based on text-retrieval methods in which video key frames are represented by a dense set of viewpoint invariant region descriptors that enable recognition to proceed successfully despite changes in camera viewpoint, lighting, and partial occlusions. Vector quantizing these region descriptors provides a visual analogy of a word - a visual word. Those words are grouped into a visual vocabulary which is used to index all key frames from the video database. Efficient retrieval is then achieved by employing methods from statistical text retrieval, including inverted file systems, and text-document frequency weightings. Though works in the literature have only adopted a simple sequential scan during search, we investigate the use of different metric access methods (MAM): M-tree, Slim-tree, and D-index, in order to accelerate the processing of similarity queries. In addition, a ranking strategy based on the spatial layout of the regions (spatial consistency) is fully described and evaluated. Experimental results have shown that the adoption of MAMs not only has improved the search performance but also has reduced the influence of the vocabulary size over test results, which may improve the scalability of our proposal. Finally, the application of spatial consistency has produced a very significant improvement of the results.

Unsupervised visual hull extraction in space, time and light domains

This paper presents an unsupervised image segmentation approach for obtaining a set of silhouettes along with the visual hull (VH) of an object observed from multiple viewpoints. The proposed approach can deal with mostly any type of appearance characteristics such as texture, similar background color, shininess, transparency besides other phenomena such as shadows and color bleeding. Compared to more classical methods for silhouette extraction from multiple views, for which certain assumptions are made on the object or scene, neither the background nor the object appearance properties are modeled. The only assumption is the constancy of the unknown background for a given camera viewpoint while the object is under motion. The principal idea of the method is the estimation of the temporal evolution of each pixel over time which provides a stability measurement and leads to its associated background likelihood. In order to cope with shadows and self-shadows, an object is captured under different lighting conditions. Furthermore, the information from the space, time and lighting domains is exploited and merged based on a MRF framework and the constructed energy function is minimized via graph cut. Experiments are performed on a light stage where the object is set on a turntable and is observed from calibrated viewpoints on a hemisphere around the object. Real data experiments show that the proposed approach allows for robust and efficient VH reconstruction of a variety of challenging objects.

Toward Standardized Mapping for Left Atrial Analysis and Cardiac Ablation Guidance.

In catheter-based cardiac ablation, the pulmonary vein ostia are important landmarks for guiding the ablation procedure, and for this reason, have been the focus of many studies quantifying their size, structure, and variability. Analysis of pulmonary vein structure, however, has been limited by the lack of a standardized reference space for population based studies. Standardized maps are important tools for characterizing anatomic variability across subjects with the goal of separating normal inter-subject variability from abnormal variability associated with disease. In this work, we describe a novel technique for computing flat maps of left atrial anatomy in a standardized space. A flat map of left atrial anatomy is created by casting a single ray through the volume and systematically rotating the camera viewpoint to obtain the entire field of view. The technique is validated by assessing preservation of relative surface areas and distances between the original 3D geometry and the flat map geometry. The proposed methodology is demonstrated on 10 subjects which are subsequently combined to form a probabilistic map of anatomic location for each of the pulmonary vein ostia and the boundary of the left atrial appendage. The probabilistic map demonstrates that the location of the inferior ostia have higher variability than the superior ostia and the variability of the left atrial appendage is similar to the superior pulmonary veins. This technique could also have potential application in mapping electrophysiology data, radio-frequency ablation burns, or treatment planning in cardiac ablation therapy.

Automatic Viewpoint Planning in Teleoperation of a Mobile Robot

Camera viewpoint has significant impact on operators situation awareness in teleoperation. This paper presents a method for automatic optimal positioning of a single camera for a remotely navigated...

Technology and Operational Safety - Pros and Cons for Police Officers [Leading Edge

The point of view (POV) camera is a compact portable device carried by police officers that captures video/audio records of operational situations. However, this technology may create serious repercussions in post-incident investigations, if the differences between human and camera perception are not understood. In addition, POV cameras pose specific issues for officer safety. Data integrity is a crucial issue for the legal process after the fact, as well as for reviewers understanding how to interpret the data correctly in the context of the original situation. The purpose is not to challenge the integrity of the equipment, but to remind all parties that it has limitations. POV camera recordings can be great memory refreshers for the officers involved and offer valuable insights for reviewers, but the key lies in understanding that they do not definitively explain every incident.

Autonomous vision-based detection of non-stellar objects flying in formation with camera point of view

The μ-advanced stellar compass (μASC) is a camera system which autonomously can provide attitude information based on the visible stars. An extended functionality of the μASC, called the vision-based sensor (VBS), enables the camera system to render rendezvous and docking navigation information for a formation flying spacecraft. The VBS system facilitates different operation modes, enabling i.e. inertial line-of-sight (LoS) solutions for any spacecraft reflecting more light than an Mv7 star. This paper will focus on the LoS solutions of the VBS system, with a thorough analysis of inflight performance and future mission aspects. The captured data used for the analysis presented in this paper, includes centroids from the VBS camera head unit (CHU) dedicated for inertial LoS operation and from the two standard startracker CHUs. By combining the data from these three CHUs, a total of 300 hours of operation distributed over the entire year is included in the thorough data analysis and performance characterisation.

애니메이션 영상화면에 나타난 카메라의 구도적 관점 (비언어 커뮤니케이션을 중심으로)

This study is focused on a compositional viewpoint of camera appears in animation feature films through nonverbal communication approach; analyzing the approach would reveal more effective interaction between receivers` and actants` communication. The content of study includes semiological analysis of one of Hayao Miyazaki`s master pieces, Howl`s Moving Castle, in terms of descriptive structure and interpretation of screen. The semiological analysis methods in this study composes with sequence analysis, actant analysis, disjunction and conjunction analysis, and shot in media semiology analysis. Through these analyses, a conclusion has been drawn that a characteristic of interactive symbol connects audience and feature films. This study applies nonverbal communication in compositional viewpoint of camera to bring more effective approach for communication among actants, animation feature films, and receivers.

Dynamic motion analysis of dart throwers motion visualized through computerized tomography and calculation of the axis of rotation

We used a dynamic three-dimensional (3D) mapping method to model the wrist in dynamic unrestricted dart throwers motion in three men and four women. With the aid of precision landmark identification, a 3D coordinate system was applied to the distal radius and the movement of the carpus was described. Subsequently, with dynamic 3D reconstructions and freedom to position the camera viewpoint anywhere in space, we observed the motion pathways of all carpal bones in dart throwers motion and calculated its axis of rotation. This was calculated to lie in 27° of anteversion from the coronal plane and 44° of varus angulation relative to the transverse plane. This technique is a safe and a feasible carpal imaging method to gain key information for decision making in future hand surgical and rehabilitative practices.

Scene invariant multi camera crowd counting

Automated crowd counting has become an active field of computer vision research in recent years. Existing approaches are scene-specific, as they are designed to operate in the single camera viewpoint that was used to train the system. Real world camera networks often span multiple viewpoints within a facility, including many regions of overlap.This paper proposes a novel scene invariant crowd counting algorithm that is designed to operate across multiple cameras. The approach uses camera calibration to normalise features between viewpoints and to compensate for regions of overlap. This compensation is performed by constructing an ‘overlap map’ which provides a measure of how much an object at one location is visible within other viewpoints. An investigation into the suitability of various feature types and regression models for scene invariant crowd counting is also conducted. The features investigated include object size, shape, edges and keypoints. The regression models evaluated include neural networks, K-nearest neighbours, linear and Gaussian process regression.Our experiments demonstrate that accurate crowd counting was achieved across seven benchmark datasets, with optimal performance observed when all features were used and when Gaussian process regression was used. The combination of scene invariance and multi camera crowd counting is evaluated by training the system on footage obtained from the QUT camera network and testing it on three cameras from the PETS 2009 database. Highly accurate crowd counting was observed with a mean relative error of less than 10%.Our approach enables a pre-trained system to be deployed on a new environment without any additional training, bringing the field one step closer toward a ‘plug and play’ system.

Orthogonal projection images for 3D face detection

We describe a real-time 3D face detector based on boosted cascade classifiers that uses a scale-invariant image representation to improve both efficiency and efficacy of the detection process, named orthogonal projection images. In this representation, images are no longer scanned at multiple scales in order to detect faces with different distances in relation to the camera. The proposed detector achieves a high degree of pose invariance by detecting frontal faces not only in the camera viewpoint but also in rotated views of the scene. The robustness of the proposed approach is demonstrated through experimental evaluation of up-to-date and state-of-the-art face databases that have different levels of noise, pose, facial expressions and other artifacts, and also differ in acquisition technology and environment. Our detector was favorably compared against state-of-the-art face detection methods, and achieved 99% detection rate over more than 13,000 test images and a false alarm rate below 1%. Although we only detect faces in our experiments, we believe our method can be extended to detect other objects with small size variation in world units.

The Impact of Image Parameters of Indirect Viewing Systems on Driver Fatigue and Task Performance

In certain driving situations, indirect viewing systems can support drivers when the drivers field of view is restricted. For some special vehicles, by applying indirect vision driving system, drivers could be protected from the dangerous working environment. In four experiments, we measured drivers task performance, workload and fatigue in different viewing systems. The image parameters considered here are camera lens field of view (FOV), image magnification factor and camera viewpoint. The results indicate that larger driver's field of vision causes greater driving fatigue while a smaller image magnification factor could reduce the drivers fatigue. Besides, the differences between eye point and camera viewpoint wont cause driver extra fatigue, but will result errors in the estimation of lateral and longitudinal position. In addition, the participants drove the course faster with natural vision than they did with the indirect vision systems. A smaller image magnification factor can improve the drivers task performance.

“Impossible Narrative Voices”:

“Impossible Narrative Voices”:

Monocular 3-D Gait Tracking in Surveillance Scenes

Gait recognition can potentially provide a noninvasive and effective biometric authentication from a distance. However, the performance of gait recognition systems will suffer in real surveillance scenarios with multiple interacting individuals and where the camera is usually placed at a significant angle and distance from the floor. We present a methodology for view-invariant monocular 3-D human pose tracking in man-made environments in which we assume that observed people move on a known ground plane. First, we model 3-D body poses and camera viewpoints with a low dimensional manifold and learn a generative model of the silhouette from this manifold to a reduced set of training views. During the online stage, 3-D body poses are tracked using recursive Bayesian sampling conducted jointly over the scene's ground plane and the pose-viewpoint manifold. For each sample, the homography that relates the corresponding training plane to the image points is calculated using the dominant 3-D directions of the scene, the sampled location on the ground plane and the sampled camera view. Each regressed silhouette shape is projected using this homographic transformation and is matched in the image to estimate its likelihood. Our framework is able to track 3-D human walking poses in a 3-D environment exploring only a 4-D state space with success. In our experimental evaluation, we demonstrate the significant improvements of the homographic alignment over a commonly used similarity transformation and provide quantitative pose tracking results for the monocular sequences with a high perspective effect from the CAVIAR dataset.

Human action recognition with sparse classification and multiple‐view learning

AbstractEmploying multiple camera viewpoints in the recognition of human actions increases performance. This paper presents a feature fusion approach to efficiently combine 2D observations extracted from different camera viewpoints. Multiple‐view dimensionality reduction is employed to learn a common parameterization of 2D action descriptors computed for each one of the available viewpoints. Canonical correlation analysis and their variants are employed to obtain such parameterizations. A sparse sequence classifier based on L1 regularization is proposed to avoid the problem of having to choose the proper number of dimensions of the common parameterization. The proposed system is employed in the classification of the Inria Xmas Motion Acquisition Sequences (IXMAS) data set with successful results.

Human interaction categorization by using audio-visual cues

Human Interaction Recognition (HIR) in uncontrolled TV video material is a very challenging problem because of the huge intra-class variability of the classes (due to large differences in the way actions are performed, lighting conditions and camera viewpoints, amongst others) as well as the existing small inter-class variability (e.g., the visual difference between hug and kiss is very subtle). Most of previous works have been focused only on visual information (i.e., image signal), thus missing an important source of information present in human interactions: the audio. So far, such approaches have not shown to be discriminative enough. This work proposes the use of Audio-Visual Bag of Words (AVBOW) as a more powerful mechanism to approach the HIR problem than the traditional Visual Bag of Words (VBOW). We show in this paper that the combined use of video and audio information yields to better classification results than video alone. Our approach has been validated in the challenging TVHID dataset showing that the proposed AVBOW provides statistically significant improvements over the VBOW employed in the related literature.