Recommended by Fredrik GustafssonSelf-localization and object tracking are key technologies for human-robot interactions. Most previous tracking algorithms focus on how to correctly estimate the position, velocity, and acceleration of a moving object based on the prior state and sensor information. What has been rarely studied so far is how a robot can successfully track the partially observable moving object with laser range finders if there is no preanalysis of object trajectories. In this case, traditional tracking algorithms may lead to the divergent estimation. Therefore, this paper presents a novel laser range finder based partially observable moving object tracking and self-localization algorithm for interactive robot applications. Dissimilar to the previous work, we adopt a stream field-based motion model and combine it with the Rao-Blackwellised particle filter (RBPF) to predict the object goal directly. This algorithm can keep predicting the object position by inferring the interactive force between the object goal and environmental features when the moving object is unobservable. Our experimental results show that the robot with the proposed algorithm can localize itself and track the frequently occluded object. Compared with the traditional Kalman filter and particle filter-based algorithms, the proposed one significantly improves the tracking accuracy.

Multi-view applications provide viewers a whole new viewing experience, and multi-view video coding (MVC) plays a key role in distributing multi-view video contents through networks with limited bandwidth. However, the computational load of a MVC encoder is pretty heavy so that it is hard to be realized in real-time applications. One reason behind this is that a MVC encoder has to make a decision of prediction direction based on rate-distortion optimization from both motion compensation prediction (MCP) and disparity compensation prediction (DCP) for multiple views. Motivated by this, this paper presents a novel fast MVC algorithm where a fast decision strategy of prediction direction of MCP and DCP is designed. Blocks with slow motion (SMBs) of all pictures in the base view and anchor pictures in enhancement views are identified based on MVs from MCP without additionalcomputations. Then, the identification of SMBs in nonanchor frames of an enhancement view will be inferred from the SMBs of base view or the other coded enhancement views. Finally, the fast algorithm is achieved by applying MCP to SMBs of non-anchor pictures in enhancement views within the same GGOP. Experimental results conducted by JMVM 6.0 show that the average time reduction is 20% while the bitrate increase and PSNR loss are less than 0.25% and 0.0045 dB, respectively.

No abstract

Estimation of people tracking may become divergent in the presence of occlusion. Since the interactions between people and environments can be mathematically modeled and probabilistically estimated, stream field based tracking provides the solution where the state of the occluded people is estimated by inferring the interactive force between the virtual goal of a person and environmental features. Such tracker suffers from high computation complexity because of the multi-hypotheses of the person's goal and feature-based map. Therefore, this paper proposes a novel virtual force field (VFF) based tracking algorithm that can be realized with a single hypothesis for the person's goal and grid-based map. The occupied grids generate repulsive forces while the person's goal generates attractive force in the virtual force field. Since the virtual force field based tracking integrates map, person, and the person's goal, the position of the person sheltered by the environment can be accurately estimated in unknown environments. Compared with the Kalman filter with constant acceleration (CA) model and stream field based algorithms, our proposed scheme significantly improves the tracking accuracy in case of occlusion.

The high computational complexity of multi-view video codecs makes it necessary to speed up for their realization in consumer electronics. Since fast encoding algorithms are expected to adapt to different video sequences, this paper proposes a fast algorithm that consists of fast mode decision and fast disparity estimation for multi-view video coding. The fast mode decision algorithm applies to both temporal and inter-view predictions. The candidates for mode decision are reduced based on a set of thresholds. Differ from the previous fast mode decision algorithms for MVC, this scheme determines the thresholds according to the online statistical analysis of motion and disparity costs of the first GOP in each view. Since the inter-view prediction is time consuming, we propose a fast disparity estimation algorithm to save encoding time. Experimental results show that our proposed scheme reduces the computational complexity significantly with negligible degradation of coding efficiency.Index Terms-Multi-view video coding, fast mode decision, statistical analysis, RD cost, motion and disparity estimation.