There are two key issues that limit further improvements in the performance of existing rotational detectors: 1) Periodic sudden change of the parameters in the rotating bounding box (RBBox) definition causes a numerical discontinuity in the loss (such as smoothL1 loss). 2) There is a gap of optimization asynchrony between the loss in the RBBox regression and evaluation metrics. In this paper, we define a new distance formulation between two convex polygons describing the overlapping degree and non-overlapping degree. Based on this smooth distance, we propose a loss called Polygon-to-Polygon distance loss (P2P Loss). The distance is derived from the area sum of triangles specified by the vertexes of one polygon and the edges of the other. Therefore, the P2P Loss is continuous, differentiable, and inherently free from any RBBox definition. Our P2P Loss is not only consistent with the detection metrics but also able to measure how far, as well as how similar, a RBBox is from another one even when they are completely non-overlapping. These features allow the RetinaNet using the P2P Loss to achieve 79.15% mAP on the DOTA dataset, which is quite competitive compared with many state-of-the-art rotated object detectors.

This paper investigates the stability analysis problem for a class of discrete-time networked control systems (NCSs) with random time delays and packet dropouts based on unified Markov jump model. The random time delays and packet dropouts existed in feedback communication link are modeled by two independent Markov chains; the resulting closed-loop system is described by a new Markovian jump linear system (MJLS) with Markov delays. Sufficient conditions of the stochastic stability for NCSs is obtained by constructing a novel Lyapunov functional, and the mode-dependent output feedback controller design method is presented based on linear matrix inequality (LMI) technique. A numerical example is given to illustrate the effectiveness of the proposed method.

Multiple-target tracking in video (MTTV) presents a technical challenge in video surveillance applications. In this paper, we formulate the MTTV problem using dynamic Markov network (DMN) techniques. Our model consists of three coupled Markov random fields: 1) a field for the joint state of the multitarget; 2) a binary random process for the existence of each individual target; and 3) a binary random process for the occlusion of each dual adjacent target. To make the inference tractable, we introduce two robust functions that eliminate the two binary processes. We then propose a novel belief propagation (BP) algorithm called particle-based BP and embed it into a Markov chain Monte Carlo approach to obtain the maximum a posteriori estimation in the DMN. With a stratified sampler, we incorporate the information obtained from a learned bottom-up detector (e.g., support-vector-machine-based classifier) and the motion model of the target into the message propagation. Other low-level visual cues such as motion and shape can be easily incorporated into our framework to obtain better tracking results. We have performed extensive experimental verification, and the results suggest that our method is comparable to the state-of-art multitarget tracking methods in all the cases we tested.

Various parametric skewed distributions are widely used to model the time-to-failure (TTF) in the reliability analysis of mechatronic systems, where many items are unobservable due to the high cost of testing. Estimating the parameters of those distributions becomes a challenge. Previous research has failed to consider this problem due to the difficulty of dependency modeling. Recently the methodology of Bayesian networks (BNs) has greatly contributed to the reliability analysis of complex systems. In this paper, the problem of system reliability assessment (SRA) is formulated as a BN considering the parameter uncertainty. As the quantitative specification of BN, a normal distribution representing the stochastic nature of TTF distribution is learned to capture the interactions between the basic items and their output items. The approximation inference of our continuous BN model is performed by a modified version of nonparametric belief propagation (NBP) which can avoid using a junction tree that is inefficient for the mechatronic case because of the large treewidth. After reasoning, we obtain the marginal posterior density of each TTF model parameter. Other information from diverse sources and expert priors can be easily incorporated in this SRA model to achieve more accurate results. Simulation in simple and complex cases of mechatronic systems demonstrates that the posterior of the parameter network fits the data well and the uncertainty passes effectively through our BN based SRA model by using the modified NBP.

No abstract

To achieve robust system, more and more vision researchers take into account fusing multiple visual cues. In this paper, we propose a novel strategy to integrate multiple naive cues for head tracking. Firstly, a cue dependency model is constructed via graphical model. Secondly, a new inference procedure based on non-parametric belief propagation is built for cue integration. The work presented is thus a general framework easy to extend for other computer vision research problems. Experimental results imply that the strategy we propose is effective, and it is robust without estimation of cue reliability.

No abstract

The inherent uncertainties of vehicle suspension systems challenge not only the capability of ride comfort and handling performance, but also the reliability requirement. In this research, a dynamic-reliable multiple model adaptive (MMA) controller is developed to overcome the difficulty of suspension uncertainties while considering performance and reliability at the same time. The MMA system consists of a finite number of optimal sub-controllers and employs a continuous-time based Markov chain to guide the jumping among the sub-controllers. The failure mode considered is the bottoming and topping of suspension components. A limitation on the failure probability is imposed to penalize the performance of the sub-controllers and a gradient-based genetic algorithm yields their optimal feedback gains. Finally, the dynamic reliability of the MMA controller is approximated by using the integration of state covariances and a judging condition is induced to assert that the MMA system is dynamic-reliable. In numerical simulation, a long scheme with piecewise time-invariant parameters is employed to examine the performance and reliability under the uncertainties of sprung mass, road condition and driving velocity. It is shown that the dynamic-reliable MMA controller is able to trade a small amount of model performance for extra reliability.