3D flow estimation for human action recognition from colored point clouds

Munaro, Matteo; Ballin, Gioia; Michieletto, Stefano; Menegatti, Emanuele

doi:10.1016/j.bica.2013.05.008

Cited by 26 publications

(20 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The latter verifies the capabilities of the proposed descriptor in efficiently defining an appropriate local-coordinate system and also encoding spatial distribution/surface-related information, as detailed in Section 3. It must be noted that the global 3D flow descriptor of [17] (mentioned in Section 2.2) was not included in the conducted comparative evaluation. This is due to the descriptor of [17] being view-dependant, since it employs a static 3D space grid division that is defined according to the single Kinect sensor that is assumed to be present.…”

Section: Resultsmentioning

confidence: 99%

“…It must be noted that the global 3D flow descriptor of [17] (mentioned in Section 2.2) was not included in the conducted comparative evaluation. This is due to the descriptor of [17] being view-dependant, since it employs a static 3D space grid division that is defined according to the single Kinect sensor that is assumed to be present. Hence, the comparison with the view-invariant HOF3D and the proposed descriptor would not be fair.…”

Section: Resultsmentioning

confidence: 99%

“…in [13,17], has not been extensively investigated, mainly due to the multiple challenges and the increased computational complexity that need to be tackled for achieving good flow estimation results.…”

Section: Flow Estimationmentioning

confidence: 99%

See 2 more Smart Citations

Local descriptions for human action recognition from 3D reconstruction data

Papadopoulos

Daras

2014

2014 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

In this paper, a view-invariant approach to human action recognition using 3D reconstruction data is proposed. Initially, a set of calibrated Kinect sensors are employed for producing a 3D reconstruction of the performing subjects. Subsequently, a 3D flow field is estimated for every captured frame. For performing action recognition, the 'Bag-of-Words' methodology is followed, where SpatioTemporal Interest Points (STIPs) are detected in the 4D space (xyzcoordinates plus time). A novel local-level 3D flow descriptor is introduced, which among others incorporates spatial and surface information in the flow representation and efficiently handles the problem of defining 3D orientation at every STIP location. Additionally, typical 3D shape descriptors of the literature are used for producing a more complete representation. Experimental results as well as comparative evaluation using datasets from the Huawei/3DLife 3D human reconstruction and action recognition Grand Challenge demonstrate the efficiency of the proposed approach.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Local descriptions for human action recognition from 3D reconstruction data

Papadopoulos

Daras

2014

2014 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

show abstract

“…Wang et al [33] extract Random Occupancy Patterns from depth sequences, use sparse coding to encode these features, and classify actions. Munaro et al [22] developed a 3D grid-based descriptor from point cloud data and recognize different actions using nearest neighbors.…”

Section: A Input Featuresmentioning

confidence: 99%

3D human action segmentation and recognition using pose kinetic energy

Shan

Akella

2014

2014 IEEE International Workshop on Advanced Robotics and Its Social Impacts

View full text Add to dashboard Cite

Human action recognition is challenging, due to large temporal and spatial variations in actions performed by humans. These variations include significant nonlinear temporal stretching. In this paper, we propose an intuitively simple method to extract action templates from 3D human joint data that is insensitive to nonlinear stretching. The extracted action templates are used as the training instances of the actions to train multiple classifiers including a multi-class SVM classifier. Given an unknown action, we first extract and classify all its constituent atomic actions and then assign the action label via an equal voting scheme. We have tested the method on two public datasets that contain 3D human skeleton data. The experimental results show the proposed method can obtain a comparable or better performance than published state-ofthe-art methods. Additional experiments also demonstrate the method works robustly on randomly stretched actions.

show abstract

“…In the literature, one can find a number of activity recognition approaches based on image sequences, point clouds or depth maps, where occupancy patterns are calculated [47] or different features are extracted such as spatio-temporal context distribution of interest points [48], histogram of oriented principal components [49] or oriented 4D normals [50], and 3D flow estimation [51]. However, the sparsity of Lidar point clouds (versus Kinect) becomes a bottleneck for extracting the above features.…”

Section: Action Recognitionmentioning

confidence: 99%