Abstract-Pedestrian detection is essential to avoid dangerous traffic situations. In this paper, we present a fast and robust algorithm for detecting pedestrians in a cluttered scene from a pair of moving cameras. This is achieved through stereo-based segmentation and neural network-based recognition. The algorithm includes three steps. First, we segment the image into sub-image object candidates using disparities discontinuity. Second, we merge and split the sub-image object candidates into sub-images that satisfy pedestrian size and shape constrains. Third, we use intensity gradients of the candidate sub-images as input to a trained neural network for pedestrian recognition. The experiments on a large number of urban street scenes demonstrate that the proposed algorithm: 1) can detect pedestrians in various poses, shapes, sizes, clothing, and occlusion status; 2) runs in real-time; and 3) is robust to illumination and background changes.Index Terms-Driver assistance system, neural networks, object detection, pedestrian detection, range image segmentation, stereo vision.

In a wireless packet (ATM) network that supports an integrated mix of multimedia traffic, the channel access protocol needs to be designed such that mobiles share the limited communications bandwidth in an eficient manner: maximizing the utilization of the frequency spectrum and minimizing the delay experienced by mobiles. In this paper, we propose and study an eficient demandassignment channel access protocol, which we call DistributedQueueing Request Update Multiple Access (DQRUMA). The protocol can be used for a wide range of applications and geographic distances. Mobiles need to send requests to the base station only for packets that arrive to an empty buffer. For packets that arrive to a non-empty buffer, transmission requests are placed collision-free by piggybacking the requests with packet transmissions. The simulation results show that even with the "worst possible" traffic characteristics, the delay-throughput performance of DQRUMA is close to the best possible with any access protocol. In addition, explicit slot-by-slot announcement of the "transmit permissions" gives the base station a lot of control over the order in which mobiles transmit their packets.

Sun-induced Fluorescence (SIF) and Photochemical Reflectance Index (PRI) data were collected in the field over maize to study their diurnal responses to different water stresses at the canopy scale. An automated field spectroscopy system was used to obtain continuous and long-term measurements of maize canopy in four field plots with different irrigation treatments. This system collects visible to near-infrared spectra with a spectrometer, which provides a sub-nanometer spectral resolution in the spectral range of 480~850 nm. The red SIF (FR) and far red SIF (FFR) data were retrieved by Spectral Fitting Methods (SFM) in the O 2 -A band and O 2 -B band, respectively. In addition to PRI, Δ PRI values were derived from PRI by subtracting an early morning PRI value. Photosynthetic active radiation (PAR) data, the canopy fraction of absorbed PAR (fPAR), and the air/canopy temperature and photosystem II operating efficiency (YII) at the leaf scale were collected concurrently. In this paper, the diurnal dynamics of each parameter before and after watering at the jointing stage were compared. The results showed that (i) both FR and FFR decreased under water stress, but FR always peaked at noon, and the peak of FFR advanced with the increase in stress. Leaf folding and the increase in Non-photochemical Quenching (NPQ) are the main reasons for this trend. Leaf YII gradually decreased from 8:00 to 14:00 and then recovered. In drought, leaf YII was smaller and decreased more rapidly. Therefore, the fluorescence yield at both the leaf and canopy scale responded to water stress. (ii) As good indicators of changes in NPQ, diurnal PRI and Δ PRI data also showed specific decreases due to water stress. Δ PRI can eliminate the impact of canopy structure. Under water stress, Δ PRI decreased rapidly from 8:00 to 13:00, and the maximum range of this decrease was approximately 0.05. After 13:00, their values started to increase but could not recover to their morning level. (iii) Higher canopy-air temperature differences ( Δ T ) indicate that stomatal closure leads to an increase in leaf temperature, which maintains a higher state in the afternoon. In summary, to cope with water stress, both leaf folding and changes in physiology are activated. To monitor drought, SIF performs best around midday, and PRI is better after noon.

No abstract

Description logics provide powerful languages for representing and reasoning about knowledge of static application domains. The main strength of description logics is that they offer considerable expressive power going far beyond propositional logic, while reasoning is still decidable. There is a demand to bring the power and character of description logics into the description and reasoning of dynamic application domains which are characterized by actions. In this paper, based on a combination of the propositional dynamic logic PDL, a family of description logics and an action formalism constructed over description logics, we propose a family of dynamic description logics DDL(X @ ) for representing and reasoning about actions, where X represents well-studied description logics ranging from the 2 L. Chang et al.ALCO to the ALCHOIQ , and X @ denotes the extension of X with the @ constructor. The representation power of DDL(X @ ) is reflected in four aspects. Firstly, the static knowledge of application domains is represented as RBoxes and acyclic TBoxes of the description logic X. Secondly, the states of the world and the pre-conditions of atomic actions are described by ABox assertions of the description logic X @ , and the post-conditions of atomic actions are described by primitive literals of X @ . Thirdly, starting with atomic actions and ABox assertions of X @ , complex actions are constructed with regular program constructors of PDL, so that various control structures on actions such as the "Sequence", "Choice", "Any-Order", "Iterate", "IfThen-Else", "Repeat-While" and "Repeat-Until" can be represented. Finally, both atomic actions and complex actions are used as modal operators for the construction of formulas, so that many properties on actions can be explicitly stated by formulas. A tableau-algorithm is provided for deciding the satisfiability of DDL(X @ )-formulas; based on this algorithm, reasoning tasks such as the realizability, executability and projection of actions can be effectively carried out. As a result, DDL(X @ ) not only offers considerable expressive power going beyond many action formalisms which are propositional, but also provides decidable reasoning services for actions described by it.

We describe a system for human body pose estimation from multiple views that is fast and completely automatic. The algorithm works in the presence of multiple people by decoupling the problems of pose estimation of different people.The pose is estimated based on a likelihood function that integrates information from multiple views and thus obtains a globally optimal solution. Other characteristics that make our method more general than previous work include: (1) no manual initialization, (2) no specification of the dimensions of the 3D structure, (3) no reliance on some learned poses or patterns of activity, and (4) insensitivity to edges and clutter in the background and within the foreground. The algorithm has applications in surveillance and promising results have been obtained.

Digital data trails from disparate sources covering different aspects of student life are stored daily in most modern university campuses. However, it remains challenging to (i) combine these data to obtain a holistic view of a student, (ii) use these data to accurately predict academic performance, and (iii) use such predictions to promote positive student engagement with the university. To initially alleviate this problem, in this paper, a model named Augmented Education (AugmentED) is proposed. In our study, (1) first, an experiment is conducted based on a real-world campus dataset of college students (N = 156) that aggregates multisource behavioral data covering not only online and offline learning but also behaviors inside and outside of the classroom. Specifically, to gain in-depth insight into the features leading to excellent or poor performance, metrics measuring the linear and nonlinear behavioral changes (e.g., regularity and stability) of campus lifestyles are estimated; furthermore, features representing dynamic changes in temporal lifestyle patterns are extracted by the means of long short-term memory (LSTM). (2) Second, machine learning-based classification algorithms are developed to predict academic performance. (3) Finally, visualized feedback enabling students (especially at-risk students) to potentially optimize their interactions with the university and achieve a study-life balance is designed. The experiments show that the AugmentED model can predict students' academic performance with high accuracy. INDEX TERMS academic performance prediction, behavioral pattern, digital campus, machine learning (ML), long short-term memory (LSTM)

No abstract