The relation between driving experience and recognition of road signs relative to their locations

OBJECTIVES: Examine how driving experience and expectations affect the ability of experienced drivers to identify traffic signs--specifically, no right turn (NRT) and no left turn (NLT) at intersections. BACKGROUND: Failure to heed signs is a frequent cause of accidents, and the authors focused on the contributions of experience and expectancy to sign identification. METHOD: Inexperienced and experienced drivers were connected to an eye tracker system and briefly exposed to various traffic scenes. Some of the pictures included an NRT sign at the expected location (on the right), and some included the same sign at an unexpected location (on the left). The same procedure was used with an NLT traffic sign. RESULTS: Experienced drivers identified traffic signs better than inexperienced drivers did when the signs were posted at the expected location but identified them worse than did inexperienced drivers when they were at unexpected locations. CONCLUSIONS: With experience, drivers' expectations regarding the expected location of traffic signs become so strong that violating these expectancies results in more identification errors among experienced drivers than among inexperienced drivers. To optimize experienced drivers' traffic sign identification, signs must be located in accordance with drivers' expectations--specifically, on the right side of the road. APPLICATIONS: When signs are misplaced, crashes can be caused by inappropriate placement rather than inappropriate driving. Highway designers should ensure that their design conforms to standards that shape experienced drivers' expectations.     

The effect of the size of one and two traffic signs on two measures of perceptual performance.

A large object can delay the perception of a small object, which hints that the use of larger objects in general and larger traffic signs in particular may be counterproductive. The present study employed either one traffic sign or two contiguous traffic signs on a post in a photograph of a road scene. Identification of a small target sign was not slower when a second sign was contiguous than when there was only the one small target sign, regardless of whether the contiguous sign was large or small. Furthermore, when a counting task was used with the same stimuli, a large sign accelerated the response to two-sign stimuli, again regardless of the contiguous sign's size. In addition, the response to two-sign stimuli was not slower than the response to one-sign stimuli. The large small comparisons of both experiments reduce the concern that employing a large traffic sign will delay the perception of a second contiguous sign. Moreover, the one-two comparisons of both experiments imply that employing two contiguous signs will not delay the perception of a sign, despite widespread evidence that a nearby object will delay the perception of a target object.     

Rectification of legibility distance in a driving simulator

Visual differences lead to differences in the legibility distances of traffic signs between driving simulators and real road environments. To ensure that the legibility distance in a simulator is similar to that in the real world, this study proposes a theoretical equation for predicting legibility distance and a simple algorithm for determining the magnifying power of a traffic sign for a display system in a simulator. Experiments of traffic sign recognition using a simulator were conducted under quasi-static and dynamic driving conditions. On-road tests were also carried out under quasi-static and dynamic driving conditions. Thirty healthy and non-disabled volunteers were recruited. The experimental results showed that the proposed theoretical equation for predicting legibility distance and the simple algorithm for determining the magnifying power of traffic signs reduced the difference in legibility distances between the simulator and real road environment under quasi-static and dynamic driving conditions.     

The role of peripheral vision in vertical road sign identification and discrimination

Abstract

The role of peripheral vision in road sign identification and discrimination was investigated in two studies. Peripheral vision plays an important role in road signs perception due to their lateral positioning. In the first study 20 participants identified road signs presented at five levels of horizontal eccentricity (1.1°–12.4°), and two levels of vertical eccentricity (0°–2.5°). In the second study road sign discrimination was tested in a same-different discrimination task. The first study showed that a vertical offset of 2.5° degraded proportion correct rate by 9%. Proportion correct rate decreased from 79% to 41% in the transition from 1.1° to 12.4° of horizontal offset. The second study showed an accurate discrimination for road signs presented within a horizontal offset of 6.4°. Road signs with angular shapes and prominent vertexes as triangular or cross signs were better identified in peripheral vision than signs with more compact shapes (circular signs).

Practitioner summary: Vertical road signs, due to their lateral positioning, are often perceived in peripheral vision. Horizontal and vertical eccentricity negatively impacts the driver’s ability to correctly identify and discriminate traffic signs. The use of singular shapes and a design with simple pictograms and large contrasting areas strongly facilitate road sign perception in peripheral vision.     

基于多尺度特征提取与特征融合的交通标志检测

基于卷积神经网络的交通标志检测算法在对现实中复杂的交通场景图像进行交通标志检测时,难以同时解决定位和分类两项任务,并且目标检测领域相关算法所使用的公开数据集提供的图像和交通标志的种类不能满足现实交通场景中复杂的情况。建立一个新的道路交通标志数据集,在YOLOv4算法的基础上针对现实交通场景图像的复杂性和图像中交通标志尺寸差异较大的特点,设计多尺寸特征提取模块和增强特征融合模块,提高算法同时定位和分类交通标志的能力。在此基础上,对算法中不同的模块设置不同的参数进行对照实验,得到一组表现最优的参数,用于检测现实交通场景图片中的交通标志。在道路交通标志数据集上的实验结果表明,该算法相比基于卷积神经网络的同类型任务目标检测算法具有更高的检测精度,平均精度均值达到83.63%。     

复杂环境下的交通标志检测与识别方法综述

交通标志检测与识别是无人驾驶三大模块中环境感知的研究热点之一,检测和识别交通标志可以向无人车传递道路交通信息,优化行车决策。在暴雨、大雾以及光线昏暗等复杂环境下,拍摄到的图像往往会被遮挡,变得模糊。这不仅影响图像的质量,还会对后期标志的检测与识别带来巨大的困难。简述了交通标志检测与识别方法,对近年来国内外学者解决各类复杂环境下交通标志检测与识别的方法、原理和步骤进行了总结归纳,有利于人们更好地解决此类问题。同时,对常用的交通标志数据集进行了总结,并对数据集里在复杂环境下拍摄的图像比例给予了说明。     

A simulated study on the effects of information volume on traffic signs, viewing strategies and sign familiarity upon driver''s visual search performance

A two-stage simulation experiment was conducted to investigate the effect of information volume on traffic regulatory/road direction signs, drivers’ viewing strategies and sign familiarity on performance in visual search. In Stage I experiment, the amount of information on a total of 187 traffic regulatory and 1272 road direction signs was calculated and divided into five information levels using cluster analysis. In Stage II experiment, 24 subjects participated in a 2 (familiarity)×5 (information volume level)×2 (viewing strategy) mixed factorial experiment. Each subject was required to perform a visual search task and a question-and-answer (Q&A) task. Visual search time and number of correct responses collected serve as the objective dependent variables. Subjective workload related to time stress and visual effort was gathered through a modified three-point rating. Results show that information volume on traffic signs had significant impact on drivers’ visual search performance. Generally, the greater the amount of information, the slower the drivers in visual search are. However, while drivers had the highest accuracy rate in remembering purely pictorial traffic signs, these signs within the smallest information volume level required a relatively longer search time. Different viewing strategies also led to different performances. The back-and-forth strategy yielded better search performance than the fixed strategy. Subjective workload evaluation indicates that drivers with less sign familiarity will be under greater time/visual pressures. Guidelines for designing for traffic signs or in-vehicle signing systems are provided.

Relevance to industry

This study calculated the amount of information on traffic signs and compared the effects of different information volume on drivers’ visual search performance. The results can provide guidelines for traffic sign designers to help decide on the presentation format for in-vehicle signing information systems.     

Estimation of Vehicle Pose and Position with Monocular Camera at Urban Road Intersections

With the rapid development of urban, the scale of the city is expanding day by day. The road environment is becoming more and more complicated. The vehicle ego-localization in complex road environment puts forward imperative requirements for intelligent driving technology. The reliable vehicle ego-localization, including the lane recognition and the vehicle position and attitude estimation, at the complex traffic intersection is significant for the intelligent driving of the vehicle. In this article, we focus on the complex road environment of the city, and propose a pose and position estimation method based on the road sign using only a monocular camera and a common GPS (global positioning system). Associated with the multi-sensor cascade system, this method can be a stable and reliable alternative when the precision of multi-sensor cascade system decreases. The experimental results show that, within 100 meters distance to the road signs, the pose error is less than 2 degrees, and the position error is less than one meter, which can reach the lane-level positioning accuracy. Through the comparison with the Beidou high-precision positioning system L202, our method is more accurate for detecting which lane the vehicle is driving on.     

Traffic flow harmonization in expressway merging

Steering a vehicle is a task increasingly challenging the driver in terms of mental resources. Reasons for this include the increasing volume of road traffic and a rising quantity of road signs, traffic lights, and other distractions at the roadside (such as billboards), to name a few. The application of Advanced Driver Assistance Systems, in particular if taking advantage of Ambient Intelligence (AmI) technology, can help to increase the perceptivity of a driver, leading as a direct consequence to more relaxed mental stress of the same. One situation where we see potential in the application of such a system are merging areas on the expressway where two or more varying traffic streams converge into a single one. In order to reduce cognitive liabilities (in this work expressed as panic or anger), drivers are exposed to while merging, we have developed two behavioral rules. The first (“increased range of perception”) enables drivers to change early upstream into a spare lane, allowing the merging traffic to join into mainline traffic at reduced conflicts, the second (“inter-car distance management” in the broader area of merging) provide drivers with recommendations of when and how to change lanes at the best. From a technical point of view, the “VibraSeat” a in-house developed car seat with integrated tactile actuators, is used for delivering information about perception range and inter-car distances to the driver in a way that does not stress his/her mental capabilities. To figure out possible improvements in its application in real traffic and at a meaningful scale, cellular automaton–based simulation of a specific section of Madrid expressway M30 was performed. Results from the data-driven simulation experiments on the true to scale model indicate that AmI technology has the potential to increase road throughput or average driving speed and furthermore to decrease the panic of drivers while merging into an upper (the main) lane.     

Design of a visual perception model with edge-adaptive Gabor filter and support vector machine for traffic sign detection

Traffic sign detection is a useful application for driving assistance systems, and it is necessary to accurately detect traffic signs before they can be identified. Sometimes, however, it is difficult to detect traffic sign, which may be obscured by other objects or affected by illumination or lightning reflections. Most previous work on this topic has been based on region of interest analysis using the color information of traffic signs. Although this provides a simple way to segment signs, this approach is weak when a sign is affected by illumination or its own color information is distorted. To overcome this, this paper introduces a robust traffic detection framework for cluttered scenes or complex city views that does not use color information. Moreover, the proposed method can detect traffic sign in the night. We establish an edge-adaptive Gabor function, which is derived from human visual perception. It is an enhanced version of the original Gabor filter, and filters out unnecessary information to provide robust recognition. It decomposes the directional information of objects and reflects specific shapes of traffic signs. Once the extracted feature is obtained, a support vector machine detects the traffic sign. Applying scale-space theory, it is possible to resolve the scaling problem of the objects that we want to find. Our system shows robust performance in traffic sign detection, and experiments on real-world scenes confirmed its properties.     

Visual search for traffic signs: the effects of clutter, luminance, and aging

Latency and eye movement measures were used to examine the effects of aging, clutter, and luminance on visual search for traffic signs embedded in digitized images of driving scenes. Initially 14 older and 14 younger observers classified daytime and nighttime traffic scenes as containing low or high amounts of clutter. Next, an independent sample of 14 younger and 14 older participants searched for traffic signs contained within these scenes. Errors were more common among the elderly. Search efficiency declined with increased clutter and with aging. However, relative to the young, older adults did not suffer disproportionately as a result of increased clutter. The methods developed might be profitably employed to assess sign conspicuity and sign acquisition during driving.     

Effects of text space of Chinese-English bilingual traffic sign on driver reading performance

Chinese-English bilingual traffic signs (CEBTS) are widely applied in public spaces within China, but few studies have addressed the spatial arrangement of the two languages on CEBTS and road standards provide little guidance for it. This study investigates whether changes in ‘separating space’, the vertical space between bilingual place names on CEBTS, affects driver reading performance. Ten sign combinations were developed using variations in the sign’s graphic complexity, total number of place names, and the spatial distribution of the place names. Each sign was displayed in a 3D graphics rendered clip of someone driving towards the road sign in random order. Participants identified destinations by reading the signs shown in the clip. The speed and the accuracy of responses was recorded. 39 English-reader participants were recruited and equally split into three groups. Three levels of separating space 0.5H, 0.75H and 1H; H refers to one Chinese character height, were prepared according to a between-subject factorial design. The results show that the separating space affected the response time regardless of sign combinations. Both 0.5H and 0.75H separations caused significant faster reading time and higher accuracy than 1H, but 0.75H appears to achieve better result. The findings contribute to the legibility research concerning bilingual signs with two very different character sets. This study considers improving sign legibility through spatial arrangement of the bilingual text, which is currently limited in the field of Chinese road sign legibility. The findings could inform future standards for Chinese-English road signs.     

Exploiting street-level panoramic images for large-scale automated surveying of traffic signs

Accurate and up-to-date inventories of traffic signs contribute to efficient road maintenance and a high road safety. This paper describes a system for the automated surveying of road signs from street-level images. This is an extremely challenging task, as the involved capturings are non-densely sampled, captured under a wide range of weather conditions and signs may be distorted. The described system is designed in a generic and learning-based fashion, which enables the recognition of different sign appearance classes with the same algorithms, based on class-specific training data. The system starts with detection of the signs visible within each image, using a detection cascade. Next, the 3D position of the signs that are detected consequently within consecutive capturings is calculated. Afterwards, each positioned road sign is classified to retrieve its sign type, thereby exploiting all detections used during positioning of the respective sign. The presented system is intended for large-scale application and currently supports 11 sign appearance classes, containing 176 different sign types. Performance evaluations conducted on a large, real-world dataset (68,010 images) show that our approach accurately positions 95.5 % of the 3,385 present signs, where 96.3 % of them are also correctly classified. Furthermore, our system localized 98.5 % of the signs in at least a single image. Our system design allows for appending a limited manual correction stage to attain a very high performance, so that sign inventories can be created cost effectively.     

Effects of using head-up display in automobile context on attention demand and driving performance

This study aimed to investigate the difference in driving performance between drivers’ attention on the head-up display (HUD)/road under low/high road conditions via a driving simulator experiment. Experimental driving included four driving scenarios with attention-on-the-HUD followed by attention-on-the-road or vice versa under high or low driving load conditions. Each scenario took about a 30-min driving consisting of two 15-min sections for each attention location. Forty-eight participants, divided into four groups, drove one of the four scenarios once. Besides driving safely within speed limit, participants were also required to perform detection task and speed limit sign response task. Results revealed that drivers paying attention to the HUD, under both low and high driving load conditions, reacted faster to speed limit sign changes than when paying attention to the road. In addition, attention-to-the-HUD under low driving load condition caused the smallest variation in steering wheel angle and lateral acceleration. These differences can be attributed to the driver's enhanced awareness and the cognitive capture effect, and tended to diminish with increasing driving workload. Finally, attention shift of drivers and the so-called novelty effect for using new technology product were also found.     

Hole-based traffic sign detection method for traffic signs with red rim

A traffic sign detection method is proposed in this paper based on detecting the inner nonred region of prohibitory and danger signs. It consists of five steps: red pixel extraction, red hole extraction, hole region filtering, semicircle combination, and final decision. In the step of red pixel extraction, a new red bitmap extraction method considering relative color of neighboring pixels is proposed to improve the accuracy of the extracted red bitmap. A series of weak classifiers are cascaded in the step of hole region filtering to quickly filter out the false alarms. A support vector machine is adopted in the final decision step to further reduce the false alarm ratio. Experimental results indicate that the proposed method is robust to many kinds of adverse situations including bad lighting condition, small rotation, out-of-plane rotation, similar background color, multiple signs clustered, and partial occlusion. Experiments on the GTSDB traffic sign dataset show that the proposed method achieves the recall of 99 % and 97 % for prohibitory and danger signs, respectively, while keeps the precision above 99 %. In addition, the mean processing time of the proposed method is about 100 ms for each 1366×768 image from GTSDB on a Core I3 CPU, which confirms its suitability for real-time applications such as driver assistance systems.     

基于改进卷积神经网络的交通标志识别方法

针对传统卷积神经网络时间成本高的不足,对卷积神经网络进行了改进,减少其卷积核的数量,增加池化方式.为解决真实场景中自动驾驶系统和辅助驾驶系统中的道路交通标志识别问题,将改进的卷积神经网络运用到道路交通标志识别当中,以达到在较短时间内识别出交通标志的目的.以图形数据集GTRSB实景交通标志图像数据作为样本,用改进的卷积神经网络对实景交通标志进行识别,其识别总体准确率达到98.38%.实验结果表明,本方法可以在保持较高识别准确率的同时减少其识别的时间.     

基于多尺度卷积神经网络的交通标志识别

基于多纵卷积神经网络的交通标志识别算法识别率较高,但识别和训练时间较长,实用性较差。为此,构造一种基于多尺度卷积神经网络的道路交通标志识别模型。通过改进单尺度卷积神经网络中特征提取的基网络,将网络不同层级所产生的特征融合为多尺度特征并提供给分类器,以提高低层特征的利用率。在GTSRB数据集上的实验结果表明,该模型准确识别率达到99.25%,与多纵卷积神经网络模型相比,其在保证高精度的同时,识别和训练时间的降幅均超过90%,更适用于真实路况下交通标志的精准检测。     

交通标志识别方法综述

交通标志识别对于车辆安全行驶具有重要作用，特别是针对存在光照变化和遮挡的情况，准确性高、实时性好的交通标志自动识别亟需解决。对交通标志的检测和交通标志的识别分别进行了综述，给出了它们的原理、步骤、特点和性能，以及进行算法研究常用的交通标志数据库。相比于传统的特征检测和识别方法，采用深度学习有助于解决光照变化、部分遮挡等情况下的交通标志识别难题，是今后自动驾驶和无人驾驶性能提升的主要途径。     

Evaluating the effects of bilingual traffic signs on driver performance and safety

Variable message signs (VMS) can provide immediate and relevant information to road users and bilingual VMS can provide great flexibility in countries where a significant proportion of the population speak an alternative language to the majority. The study reported here evaluates the effect of various bilingual VMS configurations on driver behaviour and safety. The aim of the study was to determine whether or not the visual distraction associated with bilingual VMS signs of different configurations (length, complexity) impacted on driving performance. A driving simulator was used to allow full control over the scenarios, road environment and sign configuration and both longitudinal and lateral driver performance was assessed. Drivers were able to read one- and two-line monolingual signs and two-line bilingual signs without disruption to their driving behaviour. However, drivers significantly reduced their speed in order to read four-line monolingual and four-line bilingual signs, accompanied by an increase in headway to the vehicle in front. This implies that drivers are possibly reading the irrelevant text on the bilingual sign and various methods for reducing this effect are discussed.