Pedestrian Crash Prediction Models and Validation of Effective Factors on Their Safety (Case Study: Tehran Signalized Intersections)

The quantity and severity of traffic accidents have increased with the development of machinery life and traffic growth in cities and roads in the past 50 years. Among the road users, pedestrians are the most vulnerable groups to be exposed to high risks. Vehicle crashes with pedestrian are almost inevitable and cause injury or death to pedestrian. Crash investigation and statistical studies indicate that percentage of pedestrian deaths caused by vehicle accidents are much more than all deaths. A considerable amount of accidents occur at signalized and urban intersections which are the intensive crash places. Therefore in this paper appropriate models that could specify safety indicators have been indicated with existing information by characterized parametric and nonparametric variables for twenty signalized intersections. Categories and correlations of variables also have been investigated. Three models including Regression, Poisson, and Negative binomial with defined variables have been determined. T and chi square tests, calibration and comparison of variables have been done by curve fitting. The role of each parameter was specified in pedestrian crashes. Validating models had the following outcomes: Pedestrian crash prediction models were based on none linear relations at intersections. Predictable variables, developing extended linear models and also pedestrian crash prediction are on the basis of Negative binomial distribution which is used due to more data dispersion. As observed, the Negative binomial regression because of its more R2 correlation factor has more validity among other regression models such as linear regression and Poisson. Calibrated models are put into sensitivity analysis to study the effect of each previously mentioned parameter in overall performance. Hence much better perception of future transportation plans can be achieved by development of safety models at planning levels.

Pedestrian wind flow prediction using spatial-frequency generative adversarial network

Pedestrian wind flow is a critical factor in designing livable residential environments under growing complex urban conditions.Predicting pedestrian wind flow during the early design stages is essential but currently suffers from inefficiencies in numerical simulations.Deep learning,particularly generative adversarial networks(GAN),has been increasingly adopted as an alternative method to provide efficient prediction of pedestrian wind flow.However,existing GAN-based wind flow prediction schemes have limitations due to the lack of considering the spatial and frequency characteristics of wind flow images.This study proposes a novel approach termed SFGAN,which embeds spatial and frequency characteristics to enhance pedestrian wind flow prediction.In the spatial domain,Gaussian blur is employed to decompose wind flow into components containing wind speed and distinguished flow edges,which are used as the embedded spatial characteristics.Detailed information of wind flow is obtained through discrete wavelet transformation and used as the embedded frequency characteristics.These spatial and frequency characteristics of wind flow are jointly utilized to enforce consistency between the predicted wind flow and ground truth during the training phase,thereby leading to enhanced predictions.Experimental results demonstrate that SFGAN clearly improves wind flow prediction,reducing Wind_MAE,Wind_RMSE and the Fréchet Inception Distance(FID)score by 5.35%,6.52%and 12.30%,compared to the previous best method,respectively.We also analyze the effectiveness of incorporating the spatial and frequency characteristics of wind flow in predicting pedestrian wind flow.SFGAN reduces errors in predicting wind flow at large error intervals and performs well in wake regions and regions surrounding buildings.The enhanced predictions provide a better understanding of performance variability,bringing insights at the early design stage to improve pedestrian wind comfort.The proposed spatial-frequency loss term is general and can be flexibly integrated with other generative models to enhance performance with only a slight computational cost.

Pedestrian trajectory prediction method based on the Social-LSTM model for vehicle collision

Techniques for predicting the trajectory of vulnerable road users are important to the development of perception systems for autonomous vehicles to avoid accidents.The most effective trajectory prediction methods,such as Social-LSTM,are often used to predict pedestrian trajectories in normal passage scenarios.However,they can produce unsatisfactory prediction results and data redundancy,as well as difficulties in predicting trajectories using pixel-based coordinate systems in collision avoidance systems.There is also a lack of validations using real vehicle-to-pedestrian collisions.To address these issues,some insightful approaches to improve the trajectory prediction scheme of Social-LSTM were proposed,such methods included transforming pedestrian trajectory coordinates and converting image coordinates to world coordinates.The YOLOv5 detection model was introduced to reduce target loss and improve prediction accuracy.The DeepSORT algorithm was employed to reduce the number of target transformations in the tracking model.Image Perspective Transformation(IPT)and Direct Linear Transformation(DLT)theories were combined to transform the coordinates to world coordinates,identifying the collision location where the accident could occur.The performance of the proposed method was validated by training tests using MS COCO(Microsoft Common Objects in Context)and ETH/UCY datasets.The results showed that the target detection accuracy was more than 90%and the prediction loss tends to decrease with increasing training steps,with the final loss value less than 1%.The reliability and effectiveness of the improved method were demonstrated by benchmarking system performance to two video recordings of real pedestrian accidents with different lighting conditions.

Tracking Pedestrians Under Occlusion in Parking Space

Many traffic accidents occur in parking lots.One of the serious safety risks is vehicle-pedestrian conflict.Moreover,with the increasing development of automatic driving and parking technology,parking safety has received significant attention from vehicle safety analysts.However,pedestrian protection in parking lots still faces many challenges.For example,the physical structure of a parking lot may be complex,and dead corners would occur when the vehicle density is high.These lead to pedestrians’sudden appearance in the vehicle’s path from an unexpected position,resulting in collision accidents in the parking lot.We advocate that besides vehicular sensing data,high-precision digital map of the parking lot,pedestrians’smart device’s sensing data,and attribute information of pedestrians can be used to detect the position of pedestrians in the parking lot.However,this subject has not been studied and explored in existing studies.Tofill this void,this paper proposes a pedestrian tracking framework integrating multiple information sources to provide pedestrian position and status information for vehicles and protect pedestrians in parking spaces.We also evaluate the proposed method through real-world experiments.The experimental results show that the proposed framework has its advantage in pedestrian attribute information extraction and positioning accuracy.It can also be used for pedestrian tracking in parking spaces.

基于扩展卡尔曼滤波的交互式多模型跟踪算法研究

在辅助驾驶系统中,行人轨迹跟踪一直是一项有挑战性的任务,因为行人的回波信号中往往存在着许多干扰噪声。此外,行人在运动过程中可能会做出突然转身或其他改变方向的行为,这将直接导致行人运动轨迹呈现出非线性特征。针对上述问题,文中提出一种基于扩展卡尔曼滤波的交互式多模型跟踪(IMM-EKF)方法,适用于毫米波雷达对行人进行轨迹跟踪。首先,在扩展卡尔曼滤波算法(EKF)的基础上重构状态预测协方差矩阵,来补偿EKF非线性化过程中引入的误差;然后将改进的EKF作为交互式多模型算法(IMM)中的滤波器,根据行人运动特性选择匀速模型和协调转弯模型作为跟踪模型,利用所提出的IMM-EKF算法进行轨迹跟踪。实验结果表明,所提出的滤波算法较典型的EKF和改进的EKF算法,在跟踪滤波精度方面均有所提升,同时具备更优的跟踪鲁棒性。

基于信息分形的行人轨迹预测方法

行人轨迹预测应用十分广泛,比如自动驾驶、机器人导航等。在轨迹预测中,一些不确定信息给轨迹预测任务带来了挑战,比如判别器中对轨迹信息判别的不确定,复杂的交互信息。在不确定信息处理科学领域,信息分形能有效处理不确定信息的不确定性和复杂性。受此启发,为了充分处理判别器中轨迹信息判别的不确定性,提升预测精度,该文提出了基于信息分形的轨迹预测方法。首先,场景信息和历史轨迹信息被特征提取模块提取。然后,通过注意力模块获取到场景-行人之间的交互信息与行人-行人之间的交互信息。最后基于生成对抗网络和信息分形生成合理的轨迹。在两个公共数据集ETH/UCY上实验表明,该方法能有效处理轨迹信息的不确定性,提高轨迹预测的精度。比如突然转弯、从后方超越前人、避让等行为的轨迹都能有效预测。在平均位移误差(ADE)和终点位移误差(FDE)上相比基准模型误差平均降低了11.11%和23.48%。

用于学习和解释行人预期行为的群体交互场

Anticipating others’actions is innate and essential in order for humans to navigate and interact well with others in dense crowds.This ability is urgently required for unmanned systems such as service robots and self-driving cars.However,existing solutions struggle to predict pedestrian anticipation accurately,because the influence of group-related social behaviors has not been well considered.While group relationships and group interactions are ubiquitous and significantly influence pedestrian anticipation,their influence is diverse and subtle,making it difficult to explicitly quantify.Here,we propose the group interaction field(GIF),a novel group-aware representation that quantifies pedestrian anticipation into a probability field of pedestrians’future locations and attention orientations.An end-to-end neural network,GIFNet,is tailored to estimate the GIF from explicit multidimensional observations.GIFNet quantifies the influence of group behaviors by formulating a group interaction graph with propagation and graph attention that is adaptive to the group size and dynamic interaction states.The experimental results show that the GIF effectively represents the change in pedestrians’anticipation under the prominent impact of group behaviors and accurately predicts pedestrians’future states.Moreover,the GIF contributes to explaining various predictions of pedestrians’behavior in different social states.The proposed GIF will eventually be able to allow unmanned systems to work in a human-like manner and comply with social norms,thereby promoting harmonious human-machine relationships.

基于知识学习的行人轨迹预测方法综述

随着深度学习的快速发展,行人轨迹预测任务已经成为计算机视觉领域的研究热点之一,在自动驾驶、视频监控、人机交互等领域基于深度学习的行人轨迹预测方法得到了广泛应用。首先,介绍过去几年该领域的概况(特别关注基于知识学习的方法),将这些算法分成基于统计学模型的轨迹预测方法和基于知识学习的轨迹预测方法两大类,并分析每类方法的主要算法;然后,讨论行人轨迹预测任务中使用的数据集和常见的评估指标,对比基于知识学习分类的各个方法在主流数据集下的预测性能。最后,对行人轨迹预测的发展进行展望。

基于GAN的社会和场景感知行人轨迹预测

针对状态精细化长短期记忆网络(SR-LSTM)未考虑周围物理场景对行人轨迹预测的影响,且无法生成多种可能性样本的问题,提出一种基于生成对抗网络(Generative Adversarial Networks,GAN)的社会和场景感知行人轨迹预测模型。此模型引入社会注意力及语义池机制,社会注意力建模邻人当前重要意图,以从相邻行人中选择重要的信息,语义池定义物理场景语义并学习与行人轨迹相关性。由于GAN易发生模式崩溃和下降,采用Info-GAN进行训练生成更真实的样本。在ETH和UYC两个数据集上进行实验,结果表明该方法较于SR-LSTM,ADE降低8.9百分点,FDE降低12.8百分点,且可生成更多合理的样本。

Action-Aware Encoder-Decoder Network for Pedestrian Trajectory Prediction

Accurate pedestrian trajectory predictions are critical in self-driving systems,as they are fundamental to the response-and decision-making of ego vehicles.In this study,we focus on the problem of predicting the future trajectory of pedestrians from a first-person perspective.Most existing trajectory prediction methods from the first-person view copy the bird’s-eye view,neglecting the differences between the two.To this end,we clarify the differences between the two views and highlight the importance of action-aware trajectory prediction in the first-person view.We propose a new action-aware network based on an encoder-decoder framework with an action prediction and a goal estimation branch at the end of the encoder.In the decoder part,bidirectional long short-term memory(Bi-LSTM)blocks are adopted to generate the ultimate prediction of pedestrians’future trajectories.Our method was evaluated on a public dataset and achieved a competitive performance,compared with other approaches.An ablation study demonstrates the effectiveness of the action prediction branch.

基于双重注意力机制、轨迹预测与无人机遥感技术的人车流量检测方法

人、车流量的精确检测统计对于公共安全和资源管理有重要意义。针对现有检测统计方法设备投入大、维护成本高、检测区域受限、易受环境因素影响等问题,提出一种基于双重注意力机制、轨迹预测与无人机遥感技术的人车流量检测方法。该方法在RefineDet网络的基础上加以改进,通过引入双重注意力机制、使用特征融合RNN替换TCB模块等方式增强对小目标的识别能力。同时通过自适应卡尔曼滤波预测轨迹,实现了对运动目标的实时跟踪,避免了因目标距离过远、行动方式特殊导致的错误统计。自建数据集训练网络模型,并在不同场景下进行测试。实验结果表明,所提改进模型相较对照模型的MOTA提高了2.3%~3.3%,在精度接近最新模型的同时检测速度更快,基本达到实时性需求;在多种场景下的综合评价指标F值均能达到95%以上,在多种实际场景检测中的误检率均低于0.3%。

基于图神经网络的行人轨迹预测研究综述

随着计算机视觉和自动驾驶技术的快速发展,自动感知、理解和预测人类行为的能力变得越来越重要。各类传感器的普及使得社会中产生了大量运动物体的位置数据。基于这些数据预测行人的运动轨迹在社交预测等多个领域都有着极大的价值。为了深入了解这方面的发展,对基于图神经网络的行人轨迹预测方法进行了综述,从多个角度比较、分析和总结了行人轨迹预测的图神经网络算法,讨论了不同算法在该领域的研究与发展;在目前的公共数据集上进行了对比和分析,介绍了相应性能指标,给出了不同算法的性能比较结果,提出了目前研究仍存在的问题,拓展研究思路和方法;展望了未来可能出现的研究方向。

基于状态精细化长短期记忆和注意力机制的社交生成对抗网络用于行人轨迹预测

针对当前行人轨迹预测研究仅考虑影响行人交互因素的问题,基于状态精细化长短期记忆(SR-LSTM)和注意力机制提出一种用于行人轨迹预测的社交生成对抗网络(SRA-SIGAN)模型,利用生成对抗网络(GAN)学习获得目标行人的运动规律。首先,使用SR-LSTM作为位置编码器提取运动意图信息;其次,通过设置速度注意力机制合理地为同一场景中的行人分配影响力,以更好地处理行人的交互;最后,由解码器生成预测的未来轨迹。在多个公开数据集上的测试实验结果表明,SRA-SIGAN模型的总体表现良好。特别是在Zara1数据集上,与SR-LSTM模型相比,SRA-SIGAN模型的平均位移误差(ADE)和最终位移误差(FDE)分别减小了20.0%和10.5%;与社交生成对抗网络(SIGAN)模型相比,SRA-SIGAN的ADE和FDE分别下降了31.7%和24.4%。

基于多模态特征融合的行人穿越意图预测方法

行人行为预测是城市环境智能汽车决策规划系统面临的主要挑战之一,提升行人穿越意图的预测准确率对于行车安全意义重大。针对现有方法过度依赖行人的边界框位置信息,且很少考虑交通场景中环境信息及交通对象间的交互关系等问题,本文提出一种基于多模态特征融合的行人过街意图预测方法。首先结合多种注意力机制构建了一种新型全局场景上下文信息提取模块和局部场景时空特征提取模块来增强其提取车辆周边场景时空特征的能力,并依赖场景的语义解析结果来捕获行人与其周围环境之间的交互关系,解决了交通环境上下文信息与交通对象之间的交互信息应用不充分的问题。此外,本文设计了一种基于混合融合策略的多模态特征融合模块,根据不同信息源的复杂程度实现了对视觉特征和运动特征的联合推理,为行人穿越意图预测模块提供可靠信息。基于JAAD数据集的测试表明,所提出方法的预测Accuracy为0.84,较基线方法提升了10.5%,相比于现有的同类型模型,所提出方法的综合性能最佳,且具有更广泛的应用场景。

基于多头软注意力图卷积网络的行人轨迹预测

行人间交互作用的复杂性给行人轨迹预测带来了挑战,且现有算法难以捕获行人间有意义的交互信息,不能直观地建模行人间的交互作用。针对以上问题,提出多头软注意力图卷积网络。首先利用多头软注意力(MS ATT)结合内卷网络Involution分别从空间图和时间图输入中提取稀疏空间和稀疏时间邻接矩阵,生成稀疏空间和稀疏时间有向图;然后,利用图卷积网络(GCN)从稀疏空间和稀疏时间有向图中学习交互作用与运动趋势特征;最后,将学习到的轨迹特征输入时间卷积网络(TCN)以预测双高斯分布参数,生成行人预测轨迹。在ETH和UCY数据集上的实验结果表明:相较于空时社交关系池化行人轨迹预测模型(SOPM),所提算法的平均位移误差(ADE)降低了2.78%;相较于稀疏图卷积网络(SGCN),所提算法的最终位移误差(FDE)降低了16.92%。

基于第一视角的非自回归行人轨迹预测模型

行人轨迹预测在自动驾驶和监控系统等多个应用中具有重要意义.目前大多数行人轨迹预测模型采用基于循环神经网络的编码器-解码器结构,其自回归的解码结构存在一定的累积误差,而且循环神经网络对序列的长期依赖问题仍然无法很好地解决.本文提出一种基于Transformer网络的非自回归行人轨迹预测模型,非自回归的解码结构能够同时生成所有预测值来减少累积误差,Transformer网络中的自注意力机制能够改善长期依赖问题.本文还设计一个局部信息加强模块来捕获行人运动趋势发生变化的局部特征,同时结合边界框的位置信息和大小信息来编码第一视角下透视投影产生的影响,使得模型提取到的轨迹特征更加有效.实验结果表明,在基于第一视角的公开数据集PIE(Pedestrian Intention Estimation)上,本文提出的模型比PIE预测模型在15、30、45帧的平均位移误差和终点位移误差上分别降低了24%,14.5%,11%和6%.

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。

基于边特征融合的行人轨迹预测算法研究

行人轨迹预测为自动驾驶提供了重要参考,准确预测行人的移动轨迹对减少交通事故发生有重要的意义。针对现有轨迹模型在社交关系提取不足和空间上交互不充分的问题,本文提出一种结合行人距离和图注意力机制的时空网络预测模型(STGEF-GATv2)。该模型通过行人欧氏距离的邻接矩阵表示行人边特征,对边特征进行对称归一化从而加快模型收敛速度;利用长短记忆网络LSTM提取观测时段行人运动隐藏信息,并采用加法操作融合边特征,从而降低模型复杂度;通过GAT变体GATv2进行空间交互。最终将时空信息融合,结合高斯噪声输入到解码器,解码器通过已知信息生成预测轨迹。在公开数据集ETH和UCY上进行实验,实验结果表明,与现有部分轨迹预测模型相比,该模型有更高的精确度。

基于Transformer-LSTM模型的毫米波雷达行人轨迹预测方法

在智能交通领域,对行人轨迹进行预测是一个非常有意义的研究课题。毫米波雷达凭借其强大的抗干扰能力和高隐私保密性,在行人轨迹预测方面具有良好的应用前景。为了解决传统预测算法准确度不高、长序列预测效果不佳的问题,文中提出一种基于Transformer-LSTM毫米波雷达的行人轨迹预测方法。该方法利用毫米波雷达采集行人轨迹数据,通过DBSCAN方法合成轨迹点,建立行人轨迹数据集,进行模型训练。该模型利用Transformer的动态权重分配和长短期记忆网络(Long Short-Term Memory,LSTM)的轨迹特征提取,深度挖掘不同层次的关键信息。实验结果表明,与经典运动学模型和传统神经网络模型相比,Transformer-LSTM在行人轨迹预测任务中预测精度更高、稳定性更好。

基于行人姿态的轨迹预测方法

在自动驾驶领域,行人轨迹预测一直是研究热点之一,行人行为的不确定性给轨迹预测带来很大的挑战。目前大部分轨迹预测方法只专注于行人之间的信息交互,忽略了行人意图和场景中其他语义信息对行人轨迹的影响。为此,提出一种基于行人姿态的卷积编码器-解码器网络(PKCEDN)来预测目标行人轨迹的方法,所提方法包含基于卷积、长短时记忆(LSTM)网络的编码器-解码器模型和能够学习当前时刻与过去时刻轨迹相关性的注意力机制。所提方法在MOT16、MOT17和MOT20公开数据集上进行了相关测试,与Linear、LSTM、Social-LSTM、Social-生成对抗网络(GAN)、SR-LSTM和Msgtv等主流方法相比,在保证预测速度不降低的前提下,平均误差降低约36%。