Narrow Results

Two novel methods for fully unsupervised human action retrieval using 3D mesh sequences are presented. The first achieves high accuracy but is suitable for sequences consisting of clean meshes, such as artificial sequences or highly post-processed real sequences, while the second one is robust and suitable for noisy meshes, such as those that often result from unprocessed scanning or 3D surface reconstruction errors. The first method uses a spatio-temporal descriptor based on the trajectories of 6 salient points of the human body (i.e. the centroid, the top of the head and the ends of the two upper and two lower limbs) from which a set of kinematic features are extracted. The resulting features are transformed using the wavelet transformation in different scales and a set of statistics are used to obtain the descriptor. An important characteristic of this descriptor is that its length is constant independent of the number of frames in the sequence. The second descriptor consists of two complementary sub-descriptors, one based on the trajectory of the centroid of the human body across frames and the other based on the Hybrid static shape descriptor adapted for mesh sequences. The robustness of the second descriptor derives from the robustness involved in extracting the centroid and the Hybrid sub-descriptors. Performance figures on publicly available real and artificial datasets demonstrate our accuracy and robustness claims and in most cases the results outperform the state-of-the-art.

Three-dimensional (3D) human mesh sequences obtained by 3D scanning equipment are often used in film and television, games, the internet, and other industries. However, due to the dense point cloud data obtained by 3D scanning equipment, the data of a single frame of a 3D human model is always large. Considering the different topologies of models between different frames, and even the interaction between the human body and other objects, the content of 3D models between different frames is also complex. Therefore, the traditional 3D model compression method always cannot handle the compression of the 3D human mesh sequence. To address this problem, we propose a sequence compression method of 3D human mesh sequence based on the Laplace operator, and test it on the complex interactive behavior of a soccer player bouncing the ball. This method first detects the mesh separation degree of the interactive object and human body, and then divides the sequence into a series of fragments based on the consistency of separation degrees. In each fragment, we employ a deformation algorithm to map keyframe topology to other frames, to improve the compression ratio of the sequence. Our work can be used for the storage of mesh sequences and mobile applications by providing an approach for data compression.