Modeling vehicle–pedestrian interactions outside of crosswalks

Pedestrian and driver behaviors as well as their interactions, are essential in planning, designing and operating highway facilities. Pedestrian crossing outside of a marked or unmarked crosswalk (i.e. jaywalking), is one of those pedestrian behaviors that may highly affect safety and operations. Unlike permissible crossings at crosswalks, jaywalking events are not often anticipated by drivers, which may result in less driver reaction time and different vehicle operation dynamics. It is important to understand pedestrian crossing behavior outside of crosswalks, as well as driver yielding behavior towards them. To date, limited quantitative and behavioral research has been conducted to investigate this interaction or simulate it microscopically. This paper aims to explore both pedestrian jaywalking behavior (gap acceptance and speeds) and the corresponding driver reactions (yielding behavior) for modeling the vehicle–pedestrian interactions (VPI) outside the crosswalks in a micro-simulation environment. The study also quantifies the differences between vehicle–jaywalker and vehicle-permissible crossing. An observational study and an instrumented vehicle study were conducted on the campus of the University of Florida to collect data from pedestrian and driver perspectives, respectively.Crossing speed, yield acceptance and delay of jaywalking crossings and permissible crossings were observed in the study and these attributed can be used for replicating pedestrian operations in simulators. Moreover, behaviors of driver approaching jaywalkers versus pedestrians crossing at designated crosswalks were compared on the basis of yield rates, and vehicle speed profiles. Vehicle yield dynamics were analyzed to model the driver reactions towards jaywalkers. Lastly, it was found that the locations of jaywalking events are highly concentrated and influenced by the crossing environment, such as pedestrian and vehicular volume, bus stops presence and crossing distance.This paper establishes several quantitative relationships describing interactions between pedestrians crossing outside of crosswalks and approaching drivers, which provide the basis and assumptions for modeling such interactions in a micro-simulation environment for traffic operational analyses.     

考虑行人效益的拥挤交叉口多目标配时优化

机动车流量、行人流量均较大的信号控制交叉口,行人路权往往过低,过街闯红灯现象普遍,且难以控制与避免,严重降低了交叉口运行效率。为了有效解决这一问题,采用机动车效益、行人效益、交叉口通行能力为评价指标,以交叉口饱和度为约束条件,在机动车效益指标中考虑了行人闯红灯对机动车造成的延误影响,建立了交叉口多目标信号配时优化模型;设计了求解模型的遗传算法。算例结果表明,所提出的模型能较好改善交叉口综合效益。     

基于行人安全的交通信号灯智能控制算法研究

提出了一种基于深度确定性策略梯度（Deep Deterministic Policy Gradient , DDPG）的行人安全智能交通信号控制算法。通过对交叉口数据的实时观测,综合考虑行人安全与车辆通行效率,智能地调控交通信号周期时长,相位顺序以及相位持续时间,实现交叉路口安全高效的智能控制。同时,采用优先经验回放提高采样效率,加速了算法收敛。由于行人安全与车辆通行效率存在相互矛盾,研究中通过精确地设计强化学习的奖励函数,折中考虑行人违规引起的与车辆的冲突量和车辆通行的速度,引导交通信号灯学习路口行人的行为,学习最佳的配时方案。仿真结果表明在动态环境下,该算法在行人与车辆冲突量,车辆的平均速度、等待时间和队列长度均优于现有的固定配时方案和其他的智能配时方案。     

无信号人行横道交通流特性的研究

考虑到行人穿越人行横道特点和心理因素,构建人行横道处机动车和行人相互干扰行为的元胞自动机模型。模型中重新制定行人和机动车的冲突干扰规则,引入临界安全间隙和临界决策间隙概念计算行人的通过概率,根据行人过街等待时间阈值定义了行人的冒险概率。模拟结果显示,临界决策间隙、行人忍耐时间阈值和行人的临界跟随间隙对机动车与行人流量均有不同程度的影响,其中行人的临界跟随间隙对交通流影响最大。     

A crosswalk pedestrian recognition system by using deep learning and zebra-crossing recognition techniques

Pedestrian detection is essential for improving pedestrian safety in an intelligent traffic system. The efficiency of the system is affected by real-time processing and the error rate of detection. These concerns have not been completely addressed in previous studies. Therefore, this study proposes a real-time pedestrian recognition system that ensures high accuracy by using a deep learning classifier and zebra-crossing recognition techniques. The proposed system was designed to improve pedestrian safety and reduce accidents at intersections. Environmental feature vectors were first used to detect zebra crossings and to determine crossing areas. An adaptive mapping technique was then used to map the pedestrian waiting area based on the crossing area. A dual camera mechanism was used to maintain detection accuracy and improve system fault tolerance. Finally, the you-only-look-once model was used to recognize pedestrians at intersections. A system prototype was implemented to verify the feasibility of the proposed system. The results revealed that the proposed scheme outperforms the conventional histogram of oriented gradients and Haarcascade schemes.     

基于视野域机制的行人轨迹预测

为提高行人在复杂交通场景中交互的安全性,提出一种基于social-GAN（social-generative adversarial network）的行人轨迹预测算法SAN-GAN（social angle norm-GAN）。该算法首先以行人历史位置信息与头部信息为输入,通过轨迹生成器LSTM网络（long short term memory networks）获取行人隐藏特征信息,并基于行人视野域模块捕捉行人视野域动态变化,对所有行人建立扇形视野域并筛选有效信息,从而驱动神经网络模型预测行人未来轨迹变化。将SAN-GAN与LSTM、social-LSTM（social-long short term memory networks）、social-GAN等轨迹预测算法进行对比实验,结果表明SAN-GAN算法相较于其他算法,在预测3.2 s的行人轨迹时,ADE分别平均降低65.8%、51.2%、10.7%,FDE分别平均降低73.6%、60.9%、10.4%。SAN-GAN能够有效地预测行人在复杂交通环境中进行交互的未来轨迹。     

自动驾驶场景的尺度感知实时行人检测

目的 行人检测是目标检测中的一个基准问题,在自动驾驶等场景有着较大的实用价值,在路径规划和智能避障方面发挥着重要作用。受限于现实的算法功耗和运行效率,在自动驾驶场景下行人检测存在检测速度不佳、遮挡行人检测精度不足和小尺度行人漏检率高等问题,在保证实时性的前提下设计一种适合行人检测的算法,是一项挑战性的工作。方法 本文旨在解决自动驾驶场景中耗时长、行人遮挡和小尺度行人检测结果精度低的问题,提出了一种尺度注意力并行检测算法（scale-aware and efficient object detection,Scale-aware EfficientDet）：在特征提取与检测中使用了EfficientDet的主干网络,保证算法效率和功耗的平衡;在行人遮挡方面,为了提高模型对遮挡现象的检测精度,引入了可以增强行人与其他物体之间特征差异的损失函数;在提高小目标行人检测精度方面,采用scale-aware双路网络算法来增加对小目标行人的检测精度。结果 本文选择Caltech行人数据集作为对比数据集,选取YOLO（you only look once）、YOLOv3、SA-FastRCNN（scale-aware fast region-based convolutional neural network）等算法进行对比,在运行效率方面,本文算法在连续输入单帧图像的情况下达到了35帧/s,多图像输入时达到了70帧/s的工作效率;在模型精度测试中,本文算法也略胜一筹。本文算法应用于2020年中国智能汽车大赛中,在安全避障环节皆获得满分。结论 本文设计的尺度感知的行人检测算法,在EfficientDet高性能检测器的基础上,通过结合损失函数、scale-aware双路子网络的改进,进一步提升了本文检测器的鲁棒性。     

Analysis of traffic hazard intensity: A spatial epidemiology case study of urban pedestrians

Traffic safety studies have underscored the hazardous conditions of pedestrians in the United States. This situation calls for increased public awareness of the pedestrian safety issue and better knowledge of the main factors contributing to traffic hazard for urban pedestrians. The purpose of this spatial epidemiology research is to gain greater insights into the geographic dimension exhibited by the intensity of traffic collisions involving urban pedestrians. Pedestrian crashes are studied in Buffalo, NY for years 2003 and 2004. Factors of hazard intensity are determined and compared for three age cohorts as well as for collisions occurring at intersections versus mid-block locations. Physical road characteristics and density of development, as well as socio-economic and demographic variables and potential trip attractors are examined. Spatial regression models are used to account for spatial dependencies. Econometric analysis underscores that all classes of environmental factors tested are significant drivers of pedestrian traffic hazard intensity. Results of the geographic analysis indicate that young and adult pedestrian traffic hazard intensities follow rather distinct logics. In addition, intersection and mid-block crashes differ by their socio-economic correlates, as well as their spatial distribution in the urban fabric.     

An automated signalized junction controller that learns strategies by temporal difference reinforcement learning

This paper shows how temporal difference learning can be used to build a signalized junction controller that will learn its own strategies through experience. Simulation tests detailed here show that the learned strategies can have high performance. This work builds upon previous work where a neural network based junction controller that can learn strategies from a human expert was developed (Box and Waterson, 2012). In the simulations presented, vehicles are assumed to be broadcasting their position over WiFi giving the junction controller rich information. The vehicle's position data are pre-processed to describe a simplified state. The state-space is classified into regions associated with junction control decisions using a neural network. This classification is the strategy and is parametrized by the weights of the neural network. The weights can be learned either through supervised learning with a human trainer or reinforcement learning by temporal difference (TD). Tests on a model of an isolated T junction show an average delay of 14.12 s and 14.36 s respectively for the human trained and TD trained networks. Tests on a model of a pair of closely spaced junctions show 17.44 s and 20.82 s respectively. Both methods of training produced strategies that were approximately equivalent in their equitable treatment of vehicles, defined here as the variance over the journey time distributions.     

Pedestrian detection algorithm based on video sequences and laser point cloud

Pedestrian detection is a critical problem in the field of computer vision. Although most existing algorithms are able to detect pedestrians well in controlled environments, it is often difficult to achieve accurate pedestrian detection from video sequences alone, especially in pedestrian-intensive scenes wherein pedestrians may cause mutual occlusion and thus incomplete detection. To surmount these difficulties, this paper presents pedestrian detection algorithm based on video sequences and laser point cloud. First, laser point cloud is interpreted and classified to separate pedestrian data and vehicle data. Then a fusion of video image data and laser point cloud data is achieved by calibration. The region of interest after fusion is determined using feature information contained in video image and three-dimensional information of laser point cloud to remove false detection of pedestrian and thus to achieve pedestrian detection in intensive scenes. Experimental verification and analysis in video sequences demonstrate that fusion of two data improves the performance of pedestrian detection and has better detection results.     

用于行人轨迹预测的场景限制时空图卷积网络

目的 针对行人轨迹预测问题,已有的几种结合场景信息的方法基于合并操作通过神经网络隐式学习场景与行人运动的关联,无法直观地解释场景对单个行人运动的调节作用。除此之外,基于图注意力机制的时空图神经网络旨在学习全局模式下行人之间的社会交互,在人群拥挤场景下精度不佳。鉴于此,本文提出一种场景限制时空图卷积神经网络（scene-constrained spatial-temporal graph convolutional neural network,Scene-STGCNN）。方法 Scene-STGCNN由运动模块、基于场景的微调模块、时空卷积和时空外推卷积组成。运动模块以时空图卷积提取局部行人时空特征,避免了时空图神经网络在全局模式下学习交互的局限性。基于场景的微调模块将场景信息嵌入为掩模矩阵,用来调节运动模块生成的中间运动特征,具备实际场景下的物理解释性。通过最小化核密度估计下真实轨迹的负对数似然,增强Scene-STGCNN输出的多模态性,减少预测误差。结果 实验在公开数据集ETH （包含ETH和HOTEL）和UCY （包含UNIV、ZARA1和ZARA2）上与其他7种主流方法进行比较,就平均值而言,相对于性能第2的模型,平均位移误差（average displacement error,ADE）值减少了12%,最终位移误差（final displacement error,FDE）值减少了9%。在同样的数据集上进行了消融实验以验证基于场景的微调模块的有效性,结果表明基于场景的微调模块能有效建模场景对行人轨迹的调节作用,从而减小算法的预测误差。结论 本文提出的场景限制时空图卷积网络能有效融合场景和行人运动,在学习局部模式下行人交互的同时基于场景特征对轨迹特征做实时性调节,相比于其他主流方法,具有更优的性能。     

基于特征定位与融合的行人重识别算法

行人外观属性是区分行人差异的重要语义信息。行人属性识别在智能视频监控中有着至关重要的作用,可以帮助我们对目标行人进行快速的筛选和检索。在行人重识别任务中,可以利用属性信息得到精细的特征表达,从而提升行人重识别的效果。文中尝试将行人属性识别与行人重识别相结合,寻找一种提高行人重识别性能的方法,进而提出了一种基于特征定位与融合的行人重识别框架。首先,利用多任务学习的方法将行人重识别与属性识别结合,通过修改卷积步长和使用双池化来提升网络模型的性能。其次,为了提高属性特征的表达能力,设计了基于注意力机制的平行空间通道注意力模块,它不仅可以在特征图上定位属性的空间位置,而且还可以有效地挖掘与属性关联度较高的通道特征,同时采用多组平行分支结构减小误差,进一步提高网络模型的性能。最后,利用卷积神经网络设计特征融合模块,将属性特征与行人身份特征进行有效融合,以获得更具鲁棒性和表达力的行人特征。实验在两个常用的行人重识别数据集DukeMTMC-reID和Market-1501上进行,结果表明,所提方法在现有的行人重识别方法中处于领先水平。     

High-performance and deep pedestrian detection based on estimation of different parts

Pedestrian detection, despite the recent advances, still is of a great challenge to computer vision in wide range of diversified applications such as urban autonomous driving and intelligent transportation. Deep convolutional neural network has greatly contributed to the recent advances in pedestrian detection algorithms. The aim of this paper is to use modified single-shot detector (SSD) approach in pedestrian detection and then improve it by a novel deep architecture. The proposed deep architecture extracts initial Region of Interests (RoIs) using SSD approach, while it employs nine parallel fast RCNNs based on inception modules to estimate nine different parts of body. The proposed method takes the advantage of a secure border in each initial RoI to both create an Extended Region of Candidate Pedestrian (ERCP) and also to extract multi-RoIs. It then selects a number of RoIs within the ERCP as detected pedestrians which satisfy few reasonable criteria. We also propose a new training approach based on different body parts estimation which searches the best RoIs. Comprehensive experimental results demonstrate that the proposed method, deep model based on parts in pedestrian proposals, is a highly effective method that achieves very competitive performance on two most popular pedestrian detection datasets: Caltech-USA and INRIA. We have improved the log-average miss rate on the Caltech-USA and INRIA pedestrian datasets to 7.28% and 4.96%, respectively.

     

夜间多场景的邻近感知实时行人检测算法

目的 行人检测是自动驾驶、监控安防等领域的关键技术，为了解决目标检测算法在夜间复杂场景以及遮挡情况下造成的行人检测精度降低的问题，本文提出将低光增强算法（low-light image enhancement）添加到夜间行人检测任务中进行联合训练，并引入邻近感知模块（nearby objects hallucinator,NOH），提出了一种改进的夜间监控场景下的邻近感知行人检测算法（nearby-aware surveillance pedestrian detection algorithm,NSPDet）。方法 为了提升夜间检测行人的准确率，在基线模型中加入低光增强模块（zero-reference deep curve estimation,Zero-DCE）。为了降低密集人群、遮挡造成的漏检、误检，利用NOH建模周围行人分布信息，提出了行人检测头（PedestrianHead）。为了减少模型参数，提升推理速度，本文利用深度可分离卷积将模型进行轻量化。结果 在NightSurveillance数据集上进行3组消融实验，相比基线模型YOLOX(exceeding YOLO (yo...     

联合语义分割与注意力机制的行人再识别模型

受行人姿态变化、光照视角、背景变换等因素的影响,现有行人再识别模型通常对数据集中的行人分成若干块提取图像的局部特征进行辨识以提高识别精度,但存在人体局部特征不匹配、容易丢失非人体部件的上下文线索等问题。构建一种改进的行人再识别模型,通过将人体语义解析网络的局部特征进行对齐,增强行人语义分割模型对图像中行人任意轮廓的建模能力,利用局部注意力网络捕捉非人体部分丢失的语境线索。实验结果表明,该模型在Market-1501、DukeMTMC和CUHK03数据集上的平均精度均值分别达到83.5%、80.8%和92.4%,在DukeMTMC数据集上的Rank-1为90.2%,相比基于注意力机制、行人语义解析和局部对齐网络的行人再识别模型具有更强的鲁棒性和迁移性。     

MARL-Ped: A multi-agent reinforcement learning based framework to simulate pedestrian groups

Pedestrian simulation is complex because there are different levels of behavior modeling. At the lowest level, local interactions between agents occur; at the middle level, strategic and tactical behaviors appear like overtakings or route choices; and at the highest level path-planning is necessary. The agent-based pedestrian simulators either focus on a specific level (mainly in the lower one) or define strategies like the layered architectures to independently manage the different behavioral levels. In our Multi-Agent Reinforcement-Learning-based Pedestrian simulation framework (MARL-Ped) the situation is addressed as a whole. Each embodied agent uses a model-free Reinforcement Learning (RL) algorithm to learn autonomously to navigate in the virtual environment. The main goal of this work is to demonstrate empirically that MARL-Ped generates learned behaviors adapted to the level required by the pedestrian scenario. Three different experiments, described in the pedestrian modeling literature, are presented to test our approach: (i) election of the shortest path vs. quickest path; (ii) a crossing between two groups of pedestrians walking in opposite directions inside a narrow corridor; (iii) two agents that move in opposite directions inside a maze. The results show that MARL-Ped solves the different problems, learning individual behaviors with characteristics of pedestrians (local control that produces adequate fundamental diagrams, route-choice capability, emergence of collective behaviors and path-planning). Besides, we compared our model with that of Helbing’s social forces, a well-known model of pedestrians, showing similarities between the pedestrian dynamics generated by both approaches. These results demonstrate empirically that MARL-Ped generates variate plausible behaviors, producing human-like macroscopic pedestrian flow.     

A novel small-scale pedestrian detection method base on residual block group of CenterNet

Pedestrian detection is widely used in intelligent supervision and assisted driving. With the development of deep learning, the accuracy of pedestrian detection has been greatly improved. In actual scenes, there are often pedestrians who are far away from the camera. Such pedestrians usually have small image sizes, while existing algorithms still have defects such as missed detection for similar small-scale pedestrian detection, which will reduce the accuracy of operation. Therefore, this paper designs a Three ResNet Blocks based on CenterNet detection model. Aiming at the limited ability of a single feature extraction block to extract semantic information at different levels in the network, this paper proposes Three ResNet Blocks, which is a simple and effective multi-block group. This block group integrates three different basic blocks, each of which extracts pedestrian information separately to enhance information flow in the network structure and make detection results more accurate. In addition, combined with the advantages of activation function in the model expression, the relu6 activation function is introduced to improve the performance of the detector by preventing numerical explosion being sensitive to decimal. Comprehensive experiments on pedestrian detection datasets (Caltech and ETH) show that the proposed method exhibits excellent accuracy and detection speed, especially for small-scale pedestrian detection.     

基于车内外视觉信息的行人碰撞预警方法

行人碰撞预警系统通常依据行人检测与碰撞时间判断的方式为驾驶员提供预警信息。为了提供更加可靠的危险判断依据，本文提出一种同时分析道路状况与驾驶员头部姿态的行人碰撞预警方法，用两个单目相机分别获取车辆内外环境图像。通道特征检测器用于定位行人，根据单目视觉距离测量方法估计出行人与自车间的纵向与横向距离。多任务级联卷积网络用于定位驾驶员面部特征点，通过求解多点透视问题获取头部方向角以反映驾驶员注意状态。结合行人位置信息与驾驶员状态信息，本文构建模糊推理系统判断碰撞风险等级。在实际路况下的实验结果表明，根据模糊系统输出的风险等级可以为预防碰撞提供有效的指导。     

基于动态进化图的行人轨迹预测

行人轨迹预测在自动驾驶和社交机器人等领域有着广泛的应用.对行人间复杂的交互关系进行有效建模是提高轨迹预测准确性的关键问题.然而,基于图神经网络的方法建模行人间的复杂交互时,存在行人间交互关系不会随着时间推移而改变,并且图模型无法自适应地调整网络参数,导致预测轨迹与真实轨迹偏差较大.为此,提出基于动态进化图的行人轨迹预测方法,设计动态特征更新(DFU)以定义行人间的动态特性,对行人间动态交互进行建模以构建时间域的网络动态性,提升对行人间复杂交互关系建模的能力.采用进化图卷积单元优化编码器,灵活进化图模型网络参数,增强图模型的自适应能力.研究结果表明,在预测8个时间步长下,与STGAT模型相比,所提出模型在两个公开数据集(ETH和UCY)上取得了更好的性能,平均位移误差降低12.26%,最终位移误差降低14.10%.     

融合密度和精细分数的行人检测

目的 行人检测是指使用矩形框和置信度找出图像或者视频中的所有行人。传统的图像行人检测方法对于姿态各异或者相互遮挡的行人无能为力。深度神经网络（deep neural networks,DNN）在目标检测领域表现出色,然而依然难以解决行人检测中一些问题。本文提出一种融合密度和精细分数的行人检测方法DC-CSP（density map and classifier modules with center and scale prediction）。方法 首先,在CSP（center and scale prediction）网络的基础上添加密度图模块（density map module,DMM）和分类器模块（classifier module,CM）,得到DC-CSP网络;然后,针对置信度不精确问题,利用不同模块对分数预测结果的互补性质,设计阶段分数融合（stage score fusion,SSF）规则对检测分数进行更新,使得行人置信度上升、背景置信度下降;最后,基于NMS（non-maximum suppression）,利用估计的行人密度图,设计改进的自适应NMS（improved adaptive NMS,IAN）后处理方法,能够进一步改善检测结果,对相互遮挡行人提高交并比（intersection over union,IOU）阈值从而减少漏检,对单个行人降低IOU阈值从而减少错检。结果 在公开数据集Citypersons和Caltech上进行定量和定性分析。定量分析中,与其他方法相比,本文方法在Citypersons数据集的Reasonable、Heavy、Partial以及Bare子集上,对数平均漏检率分别下降了0.8%、1.3%、1.0%和0.8%,在Caltech数据集的Reasonable和All子集上分别下降了0.3%和0.7%;在定性分析中,可视化结果表明,本文方法在一定程度上解决了各种不同场景下存在的相互遮挡行人漏检、单个行人错检以及置信度不精确等一系列问题。此外,消融实验证明了所设计模块及其对应规则的有效性。结论 本文方法使用联合多个模块的卷积神经网络（convolutional neural network,CNN）,针对密度特征、分类特征分别设计IAN方法和SSF规则,在一定程度上解决了相互遮挡行人漏检、单个行人错检以及置信度不精确的问题,在多个数据集上证明了方法的有效性和鲁棒性。