Search

This article investigates the multirobot efficient search (MuRES) for a nonadversarial moving target problem from the multiagent reinforcement learning (MARL) perspective. MARL is deemed as a promising research field for cooperative multiagent applications. However, one of the main bottlenecks of applying MARL to the MuRES problem is the nonstationarity introduced by multiple learning agents. With learning agents simultaneously updating their policies, the environment cannot be modeled as a stationary Markov decision process, which results in the inapplicability of fundamental reinforcement learning techniques such as deep $Q$-network and policy gradient (PG). In view of that, we adopt the centralized training and decentralized execution scheme and thereby propose a cross-entropy regularized policy gradient (CE-PG) method to train the learning agents/robots. We let the robots commit to a predetermined policy during execution, collect the trajectories, and then perform centralized training for the corresponding policy improvement. In this way, the nonstationarity problem is overcome, in that the robots do not update their policies during execution. During the centralized training stage, we improve the canonical PG method to consider the interactions among robots by adding a cross-entropy regularization term, which essentially functions to “disperse” the robots in the environment. Extensive simulation results and comparisons with state of the art show CE-PG's superior performance, and we also validate the algorithm with a real multirobot system in an indoor moving target search scenario.

Action recognition from videos has many potential applications. However, there are many unresolved challenges, such as pose-invariant recognition, robustness to occlusion and others. In this paper, we propose to combine motion of body parts and pose hypothesis generation validated with specific canonical poses observed in a novel mutually reinforcing framework to achieve pose-invariant action recognition. To capture the temporal dynamics of an action, we introduce temporal stick features computed using the stick poses obtained. The combination of pose-invariant kinematic features from motion, pose hypothesis and temporal stick features are used for action recognition, thus forming a mutually reinforcing framework that repeats until the action recognition result converges. The proposed mutual reinforcement framework is capable of handling changes in posture of the person, occlusion and partial view-invariance. We perform experiments on several benchmark datasets which showed the performance of the proposed algorithm and its ability to handle pose variation and occlusion.

Fingerprint match-on-card is receiving more attention from society because of higher level of security and less privacy concern. The ISO/IEC19794-2 for match-on-card defines the minutiae data format that contains the basic information. Therefore, match-on-card is a challenge, especially when dealing with deformation and common correspondence for alignment. In this paper, a novel in-matcher clustering method is proposed to search for the matched minutia clusters to cater for deformation. Moreover, a further matching step using Mahalanobis distance to measure the inter-cluster similarity is proposed to remove the wrongly matched clusters. Finally, the overall match score is generated by combining the scores from matched clusters and the geometrical structure of clusters. The proposed algorithm achieved an average EER ⇐ 5.1979% using all FVC databases. In the NIST evaluation, the achieved false match rate (FMR) = 0.001 and false non-match rate (FNMR) = 0.08 and the average on-card verification time is 1.01s using 8-bit smartcard.

Given a video sequence, the task of action recognition is to identify the most similar action among the action sequences learned by the system. Such human action recognition is based on evidence gathered from videos. It has wide application including surveillance, video indexing, biometrics, telehealth, and human–computer interaction. Vision-based human action recognition is affected by several challenges due to view changes, occlusion, variation in execution rate, anthropometry, camera motion, and background clutter. In this survey, we provide an overview of the existing methods based on their ability to handle these challenges as well as how these methods can be generalized and their ability to detect abnormal actions. Such systematic classification will help researchers to identify the suitable methods available to address each of the challenges faced and their limitations. In addition, we also identify the publicly available datasets and the challenges posed by them. From this survey, we draw conclusions regarding how well a challenge has been solved, and we identify potential research areas that require further work.

Facial expression is crucial for proper analysis of a person’s face. It is an indicator of the emotion of a person and thus has attracted the attention of many researchers. In this work, a novel local spatio-temporal descriptor is proposed for motion pattern detection. The proposed feature comprises histogram of 3D gradients and the gradients’ variation over time to robustly describe the spatial and temporal information. It also incorporates spatio-temporal pyramid structure to handle different resolution and frame rate. To reduce the dimension of the feature, we applied genetic algorithm for region-based feature selection. We evaluated the performance of our proposed descriptors on facial expression recognition using the Cohn-Kanade (CK + ) database. The experimental results achieved 96.10% accuracy in detecting six basic emotions. The key advantages of our proposed method are: local and dynamic processing, simple implementation, high performance, and robustness to variation of video resolution or temporal sampling rate.

Family members have certain facial resemblances due to genetic similarities. Such resemblance allows us to guess the family relationship which has received little attention to date. In this paper, we propose to recognise immediate members of the same family known as ‘family verification’ through identification of members’ facial resemblance features. At first, the proposed operator extracts detailed facial information comparing to the state-of-the-art face descriptors. A novel redundant feature set removal is incorporated to reduce feature dimension. Experiments are conducted to compare performance of various features employing the proposed method and the state-of-the-art face recognition as well as the proposed redundant feature removal. The experiments also cover various scenarios where different members of the family are absent from training but present in the testing applied for missing family member verification. Results obtained show that the proposed approach is feasible even in case of missing member verification.

On-card biometric comparison is getting more attention from government and the IT industry because of the higher level of security and its ability to prevent the enrolment template from leakage that can cause privacy concern. To address the interoperability needs, the ISO/IEC SC17 WG11 published a standard entitled ISO/IEC 24787 – ‘On-card biometric comparison’. This paper will give an introduction to this new standard and analyse the limitation of implementing minutiae matching algorithm using this standard on low-cost smartcard. We will present a novel algorithm to perform off-card template alignment using the work-sharing mechanism. The final matching process is executed on-card to ensure that the security is not compromised. The average verification time of our proposed algorithm is about 2.5 seconds with an 8-bit Java card with an average EER <= 4.3% using FVC2000 and FVC2002 databases. Hence, the performance is comparable to those minutiae matching algorithms running on PC.

There are several shortcomings with the traditional token based authentication using magnetic strip card with PIN/password protection such as easy to duplicate and from forgotten to stolen password etc. Fingerprint authentication is more secure and is a convenient replacement of PIN/password. With the advancement in the microprocessor technology for token such as smartcard and fingerprint authentication technology, token based fingerprint authentication becomes a viable alternative to replace the existing magnetic strip card with PIN/password protection. In this paper, an overview of the token-based fingerprint authentication will be given and a review of the recent publications and patents will be presented.

Real-time computation of exact depth is not feasible in an active vision setup. Instead, reliable relative depth information which can be rapidly computed is preferred. In this paper, a stereo cue for computing relative depth obtained from an active stereo vision system is proposed. The proposed stereo cue can be computed purely from the coordinates of points in the stereo pair. The computational cost required is very low. No camera calibration or prior knowledge of the parameters of the stereo vision system is required. We show that the relationship between the relative depth cue and the actual depth in the three-dimensional (3D) space is monotonic. Such a relation is maintained even when the focal length and the vergence angle are changed, so long as the focal lengths of the two cameras are similar. Therefore, real-time implementation in an active vision setup can be realized. Stability analysis shows that the proposed method will be stable in practical situations, unless the stereo camera diverges. Experimental results are presented to highlight the properties and advantages of the proposed method.

Industrial cameras are coupled into robotic systems to increase the flexibility of the robots. Such hand-eye coordination usually requires calibration of the cameras to compute the three-dimensional (3D) coordinates of positions in the robot space. This paper describes a new hand-eye coordination approach which does not require camera calibration. Instead, we propose the use of relative stereo disparity to compute the relative depth between the perceived objects. Incorporating the relative depth measure into the image space converts the hand-eye coordination problem into a linear transformation between the pseudo-3D image space and the 3D robot space. Moreover, the transformation matrix involved is square and can be easily estimated and updated by using visual feedback. The proposed method is fast and simple, making it feasible for real-time visual feedback implementation. Furthermore, since no calibration is required, the method proposed is robust to substantial changes in the hand-eye system configuration. Experiments are conducted to verify the accuracy and robustness of the proposed method. The main contributions of this paper are: (i) introducing a stereo attribute that measures relative depth, (ii) formulating a pseudo-3D image space, (iii) relating the pseudo-3D image space to the robot space for computing the hand-eye transformation and (iv) realizing a robust hand-eye coordination system which incorporates visual feedback control.