Nighttime Driving Research Articles

Shadow Removal for Enhanced Nighttime Driving Scene Generation

Autonomous vehicles depend on robust vision systems capable of performing under diverse lighting conditions, yet existing models often exhibit substantial performance degradation when applied to nighttime scenarios after being trained exclusively on daytime data. This discrepancy arises from the lack of fine-grained details that characterize nighttime environments, such as shadows and varying light intensities. To address this gap, we introduce a targeted approach to shadow removal designed for driving scenes. By applying Partitioned Shadow Removal, an enhanced technique that refines shadow-affected areas, alongside image-to-image translation, we generate realistic nighttime scenes from daytime data. Experimental results indicate that our augmented nighttime scenes significantly enhance segmentation accuracy in shadow-impacted regions, thereby increasing model robustness under low-light conditions. Our findings highlight the value of Partitioned Shadow Removal as a practical data augmentation tool, adapted to address the unique challenges of applying shadow removal in driving scenes, thereby paving the way for improved nighttime performance in autonomous vehicle vision systems.

Nighttime scene understanding with label transfer scene parser

Semantic segmentation plays a crucial role in traffic scene understanding, especially in nighttime conditions. This paper tackles the task of semantic segmentation in nighttime scenes. The largest challenge of this task is the lack of annotated nighttime images to train a deep learning-based scene parser. The existing annotated datasets are abundant in daytime conditions but scarce in nighttime due to the high cost. Thus, we propose a novel Label Transfer Scene Parser (LTSP) framework for nighttime scene semantic segmentation by leveraging daytime annotation transfer. Our framework performs segmentation in the dark without training on real nighttime annotated data. In particular, we propose translating daytime images to nighttime conditions to obtain more data with annotation in an efficient way. In addition, we utilize the pseudo-labels inferred from unlabeled nighttime scenes to further train the scene parser. The novelty of our work is the ability to perform nighttime segmentation via daytime annotated labels and nighttime synthetic versions of the same set of images. The extensive experiments demonstrate the improvement and efficiency of our scene parser over the state-of-the-art methods with a similar semi-supervised approach on the benchmark of Nighttime Driving Test dataset. Notably, our proposed method utilizes only one-tenth of the amount of labeled and unlabeled data in comparison with the previous methods. Code is available at https://github.com/danhntd/Label_Transfer_Scene_Parser.git.

An Adaptive Semantic Segmentation Network for Adversarial Learning Domain Based on Low-Light Enhancement and Decoupled Generation

Nighttime semantic segmentation due to issues such as low contrast, fuzzy imaging, and low-quality annotation results in significant degradation of masks. In this paper, we introduce a domain adaptive approach for nighttime semantic segmentation that overcomes the reliance on low-light image annotations to transfer the source domain model to the target domain. On the front end, a low-light image enhancement sub-network combining lightweight deep learning with mapping curve iteration is adopted to enhance nighttime foreground contrast. In the segmentation network, the body generation and edge preservation branches are implemented to generate consistent representations within the same semantic region. Additionally, a pixel weighting strategy is embedded to increase the prediction accuracy for small targets. During the training, a discriminator is implemented to distinguish features between the source and target domains, thereby guiding the segmentation network for adversarial transfer learning. The proposed approach’s effectiveness is verified through testing on Dark Zurich, Nighttime Driving, and CityScapes, including evaluations of mIoU, PSNR, and SSIM. They confirm that our approach surpasses existing baselines in segmentation scenarios.

Smart Energy Public Street Lighting System

Public street lighting illuminates roadways for two, three, and four-wheeled vehicles. Most roads have street lighting for safety and aesthetics. The security feature improves nighttime driver sight. Motorists need appropriate illumination to reduce accidents and crime. Lighting design and placement affect the street’s and city’s nighttime attractiveness. Public street lighting lights utilize uncontrolled electricity, resulting in excessive expenditures. We need a design that effectively manages and uses electrical energy. The IoT-based Public Street Lighting system employing solar panels as battery chargers may be utilized for electrical energy planning. The ESP8266 module provides control, a Wi-Fi module provides Wi-Fi, the relay module provides an electric switch to turn on or off public street lighting lights using the intelligent energy public street lighting application, and solar panels charge the batteries. The system has LDR sensor monitoring, ultrasonic sensors, and a Nodemcu ESP8266 microprocessor. The light sensor module detects light intensity. The Smart Energy, Public Street Lighting app lets officers monitor the process in real time. This program will monitor public street lighting and analyze and graph the data.

Improving Nighttime Driving-Scene Segmentation via Dual Image-Adaptive Learnable Filters

Semantic segmentation on driving-scene images is vital for autonomous driving. Although encouraging performance has been achieved on daytime images, the performance on nighttime images are less satisfactory due to the insufficient exposure and the lack of labeled data. To address these issues, we present an add-on module called dual image-adaptive learnable filters (DIAL-Filters) to improve the semantic segmentation in nighttime driving conditions, aiming at exploiting the intrinsic features of driving-scene images under different illuminations. DIAL-Filters consist of two parts, including an image-adaptive processing module (IAPM) and a learnable guided filter (LGF). With DIAL-Filters, we design both unsupervised and supervised frameworks for nighttime driving-scene segmentation, which can be trained in an end-to-end manner. Specifically, the IAPM module consists of a small convolutional neural network with a set of differentiable image filters, where each image can be adaptively enhanced for better segmentation with respect to the different illuminations. The LGF is employed to enhance the output of segmentation network to get the final segmentation result. The DIAL-Filters are light-weight and efficient and they can be readily applied for both daytime and nighttime images. Our experiments show that DAIL-Filters can significantly improve the supervised segmentation performance on ACDC_Night and NightCity datasets, while it demonstrates the state-of-the-art performance on unsupervised nighttime semantic segmentation on Dark Zurich and Nighttime Driving testbeds. Codes and models are available at https://github.com/wenyyu/IA-Seg.

Driving in the dark: Deciphering nighttime driver detection of free-ranging roadside wildlife

Driving in the dark: Deciphering nighttime driver detection of free-ranging roadside wildlife

Exploring the Causal Impact of Age and Nighttime Driving on Road Traffic Injuries among Elderly Drivers: A Bayesian LASSO Approach.

To determine the causal relationship between aging and nighttime driving and the odds of injury among elderly drivers. In this cross-sectional study, 5460 car accidents were investigated from 2015 to 2016. The data were extracted from the Iranian Integrated Road Traffic Injury Registry System. Pedestrian accidents, motorcycle crashes, and fatalities were excluded from the study. To account for major confounders, Bayesian-LASSO, and treatment-effect cutting-edge approaches were used. Overall, 801 injuries (14.67%) were evaluated. The results of the univariable analysis indicated that aging and nighttime had adverse effects on the odds of road traffic injuries (RTIs), even after adjusting for the effect of other variables, these effects remained statistically significant. According to a newly developed approach, the overall effects of aging and nighttime were significantly and directly correlated with the odds of being injured for older adults (both p<0.001). Our findings indicated that drivers over 75 years old experienced 23% higher injury odds (OR=1.23, 95% CI:1.11 to 1.39; p<0.001), while driving at night increased the odds by 1.78 times (OR=1.78, 95% CI:1.51 to 1.83; p<0.001). Aging and nighttime driving are significant risk factors for RTIs among elderly drivers. This highlights the importance of implementing targeted interventions to enhance road safety for this vulnerable population. Furthermore, the use of advanced Bayesian-LASSO and treatment-effect statistical methods highlights the importance of utilizing sophisticated methodologies in epidemiological research to effectively capture and adjust for potential confounding factors.

A One-Stage Domain Adaptation Network With Image Alignment for Unsupervised Nighttime Semantic Segmentation.

In this paper, we tackle the problem of semantic segmentation for nighttime images that plays an equally important role as that for daytime images in autonomous driving, but is also much more challenging due to very poor illuminations and scarce annotated datasets. It can be treated as an unsupervised domain adaptation (UDA) problem, i.e., applying other labeled dataset taken in the daytime to guide the network training meanwhile reducing the domain shift, so that the trained model can generalize well to the desired domain of nighttime images. However, current general-purpose UDA approaches are insufficient to address the significant appearance difference between the day and night domains. To overcome such a large domain gap, we propose a novel domain adaptation network "DANIA" for nighttime semantic image segmentation by leveraging a labeled daytime dataset (the source domain) and an unlabeled dataset that contains coarsely aligned day-night image pairs (the target daytime and nighttime domains). These three domains are used to perform a multi-target adaptation via adversarial training in the network. Specifically, for the unlabeled day-night image pairs, we use the pixel-level predictions of static object categories on a daytime image as a pseudo supervision to segment its counterpart nighttime image. We also include a step of image alignment to relieve the inaccuracy caused by the misalignment between day-night image pairs by estimating a flow to refine the pseudo supervision produced by daytime images. Finally, a re-weighting strategy is applied to further improve the predictions, especially boosting the prediction accuracy of small objects. The proposed DANIA is a one-stage adaptation framework for nighttime semantic segmentation, which does not train additional day-night image transfer models as a separate pre-processing stage. Extensive experiments on Dark Zurich and Nighttime Driving datasets show that our DANIA achieves state-of-the-art performance for nighttime semantic segmentation.

Thermal Infrared Image Colorization for Nighttime Driving Scenes With Top-Down Guided Attention

Benefitting from insensitivity to light and high penetration of foggy environments, infrared cameras are widely used for sensing in nighttime traffic scenes. However, the low contrast and lack of chromaticity of thermal infrared (TIR) images hinder the human interpretation and portability of high-level computer vision algorithms. Colorization to translate a nighttime TIR image into a daytime color (NTIR2DC) image may be a promising way to facilitate nighttime scene perception. Despite recent impressive advances in image translation, semantic encoding entanglement and geometric distortion in the NTIR2DC task remain under-addressed. Hence, we propose a toP-down attEntion And gRadient aLignment based generative adversarial network, referred to as PearlGAN. A top-down guided attention module and an elaborate attentional loss are first designed to reduce the semantic encoding ambiguity during translation. Then, a structured gradient alignment loss is introduced to encourage edge consistency between the translated and input images. In addition, pixel-level annotation is carried out on a subset of FLIR and KAIST datasets to evaluate the semantic preservation performance of multiple translation methods. Furthermore, a new metric is devised to evaluate the geometric consistency in the translation process. Extensive experiments demonstrate the superiority of the proposed PearlGAN over other image translation methods for the NTIR2DC task. The source code and labeled segmentation masks will be available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/FuyaLuo/PearlGAN/</uri> .

Contrast Sensitivity and Night Driving in Older People: Quantifying the Relationship Between Visual Acuity, Contrast Sensitivity, and Hazard Detection Distance in a Night-Time Driving Simulator.

Purpose(i) To assess how well contrast sensitivity (CS) predicts night-time hazard detection distance (a key component of night driving ability), in normally sighted older drivers, relative to a conventional measure of high contrast visual acuity (VA); (ii) To evaluate whether CS can be accurately quantified within a night driving simulator.Materials and MethodsParticipants were 15 (five female) ophthalmologically healthy adults, aged 55–81 years. CS was measured in a driving simulator using Landolt Cs, presented under static or dynamic driving conditions, and with or without glare. In the dynamic driving conditions, the participant was asked to simultaneously maintain a (virtual) speed of 60 km/h on a country road. In the with glare conditions, two calibrated LED arrays, moved by cable robots, simulated the trajectories and luminance characteristics of the (low beam) headlights of an approaching car. For comparison, CS was also measured clinically (with and without glare) using a Optovist I instrument (Vistec Inc., Olching, Germany). Visual acuity (VA) thresholds were also assessed at high and low contrast using the Freiburg Visual Acuity Test (FrACT) under photopic conditions. As a measure of driving performance, median hazard detection distance (MHDD) was computed, in meters, across three kinds of simulated obstacles of varying contrast.ResultsContrast sensitivity and low contrast VA were both significantly associated with driving performance (both P < 0.01), whereas conventional high contrast acuity was not (P = 0.10). There was good correlation (P < 0.01) between CS measured in the driving simulator and a conventional clinical instrument (Optovist I). As expected, CS was shown to decrease in the presence of glare, in dynamic driving conditions, and as a function of age (all P < 0.01).ConclusionContrast sensitivity and low contrast VA predict night-time hazard detection ability in a manner that conventional high contrast VA does not. Either may therefore provide a useful metric for assessing fitness to drive at night, particularly in older individuals. CS measurements can be made within a driving simulator, and the data are in good agreement with conventional clinical methods (Optovist I).

Determination of Speed-Dependent Roadway Luminance for an Adequate Feeling of Safety at Nighttime Driving

The purpose of this work is to determine as a function of velocity the minimal roadway luminance that is required to be judged as being bright enough for a driver to perform a nighttime driving task with an adequate feeling of safety. In this context, it shall also be evaluated which areas of the vehicle forefield are most crucial for the driver’s general brightness perception. A field study with 23 subjects and dimmable LED headlights was conducted, in which the subjects were given the task to assess their perceived brightness for different luminance levels caused by the headlights’ low-beam distribution in the vehicle’s forefield on a 5-step rating scale. The experiments were repeated for three different driving velocities of 0 km h−1 (static case), 30 km h−1, and 60 km h−1, respectively. Results for the static case indicate that, for the roadway to be perceived as bright enough by 50% of the subjects, an average roadway luminance of 0.88 cd m−2 is required in an area up to 32 m in front of the vehicle. Furthermore, a significant effect of driving speed is observed. For example, at 60 km h−1, the luminance must be increased to 1.54 cd m−2 to be still perceived as sufficiently bright by 50% of the subjects.

Nighttime effectiveness of the pedestrian hybrid beacon, rectangular rapid flashing beacon, and LED-embedded crossing sign

Introduction: A large majority of pedestrian fatal crashes occurred during the nighttime. The focus of this research was to identify if the following pedestrian crossing treatments were more or less effective at night: pedestrian hybrid beacon (PHB), rectangular rapid flashing beacon (RRFB), or LED-embedded crossing warning sign (LED-Em). Method: For each treatment, two statistical evaluations were used on the staged pedestrian data: ANCOVA models that considered per site mean yield rates and logistic regression that considered the individual driver response to the crossing pedestrian. Results: For the PHB, essentially no difference was found between the very high daytime and nighttime driver yielding values. The research found RRFBs to be more effective at night, and the LED-Em to be more effective during the day. Using the results from the logistic regression evaluation, higher driver yielding was observed at LED-Em sites in the lower speed limit group (30 or 35 mph (48.3 or 56.3 kph), with 2 lanes (rather than 4 lanes), with narrow lanes of 10.5 or 11 ft (3.2 or 3.4 m) widths (rather than 11.5 or 12 ft (3.5 or 3.7 m) widths), and lower hourly volumes. The results from the ANCOVA model for LED-Ems also showed a statistically significant difference for yield lines (higher yielding when present). Conclusions: This analysis represents the only known study to date on the effectiveness of pedestrian crossing treatments at night. Practical Applications: This study provides additional support for the PHB as a treatment because the PHB was found to be highly effective during the nighttime as well as the daytime. The value of using advance yield lines was also demonstrated. The findings offer a caution regarding the use of the LED-Em treatment on higher speed, higher volume, or wider roads.

Drinking and Night-Time Driving May Increase the Risk of Severe Health Outcomes: A 5-Year Retrospective Study of Traffic Injuries among International Travelers at a University Hospital Emergency Center in Thailand.

Road traffic injury (RTI) is a leading cause of death in developing countries. This burden affects not only locals, but also international travelers. Data on international travelers with RTIs in Thailand, especially from a medical perspective, are limited. This study aimed to analyze the factors associated with severe health outcomes following RTIs among international travelers at a university hospital emergency center in Thailand from January 2015 to December 2019. The retrieved data consisted of demographics, risks, preventive factors, and health outcomes. The severity of outcome was classified as fatality, hospitalization, or non-severe. A multinomial logistic regression model was used to identify the possible determinants of severity of health outcome among international travelers with RTI. A total of 720 travelers with RTIs (69% males; 82.5% were Southeast Asian) were included, with a mean age of 28.5 years. Of these, 144 (20%) had severe health outcomes: 64 (9%) fatalities and 80 (11%) hospitalizations. The level of severity of outcome was not associated with travelers’ demographics, but was associated with conventional risk factors, i.e., motorcycle use, alcohol/drug use, night-time driving, and less use of seatbelt/helmet. In a multinomial logistic regression analysis, alcohol drinking (adjusted odds ratio (AOR) 2.53, 95% confidence interval (CI) 1.41–4.55) and night-time driving (AOR 2.54, 95% CI 1.36–4.75) were associated with hospitalization. Patients who had a history of tetanus vaccination were less likely to die (AOR 0.37, 95% CI 0.17–0.81). In conclusion, one-fifth of RTIs resulted in severe health outcomes, and 9% were fatal. Road safety campaigns in Thailand should target travelers of all nationalities. Interventions that enhance travelers’ safety practices and proper preparation for road accidents should be explored further.

An Analysis of the Impact of Attentional Momentum Effect on Driver’s Ability of Awareness During Night-time Driving

This study aims to demonstrate that driver’s awareness during driving at night can be improved by attentional momentum effect. Driver’s ability to be aware of pedestrians is often degraded due to the inhibition of return (hereafter, “IOR”). The IOR is the tendency that exogenous attention is suppressed toward the direction to which the driver previously referred. Previous studies have suggested that attentional momentum effect (hereafter, “AME”) can relieve the effect of IOR. For example, IOR phenomenon appears when driver’s exogenous attention is localized on roadside stores with high-illumination intensity. Focusing on the night-time driving, this study investigates the impact through a detection task based on dual-task setting. As a result, it was observed that the target detection was delayed under IOR condition, and the delay in target detection was improved under AME condition. This result supports that attentional momentum effect led to improve the driver’s ability to be aware of objects during night time driving.

0635 Vigilance Observations - Learning from Nighttime Driving Behaviours

0635 Vigilance Observations - Learning from Nighttime Driving Behaviours

Nighttime Driving in Older Adults: Effects of Glare and Association With Mesopic Visual Function.

To examine the associations between nighttime driving performance of older drivers and photopic, mesopic, and glare-based tests of visual function. Participants included 26 older drivers (71.8 ± 6.3 years), with minimal or no eye disease, but who reported vision-related nighttime driving difficulties. Nighttime driving performance was assessed on a closed-road circuit, which included intermittent glare. An overall driving performance score was calculated based on detection of signs, pedestrians, wooden animals and road markings, lane-keeping, and avoidance of low contrast hazards. Visual function tests included photopic and mesopic visual acuity (VA) and contrast sensitivity (CS). Tests of glare (Berkeley Glare and Aston Halometer) and mesopic motion sensitivity were also assessed. Regression analyses were used to explore the associations between these vision measures and nighttime driving performance. The overall driving performance score was significantly reduced by intermittent glare (P = 0.002); notably, pedestrian detection decreased by 38% in the presence of intermittent glare (P < 0.001). Overall driving scores were most strongly associated with motion sensitivity (P = 0.001) and mesopic high contrast VA (P = 0.002), rather than photopic or glare-based tests. Motion sensitivity accounted for more than twice the variation in driving performance compared to photopic high contrast VA (29% vs. 14%). Glare reduced several aspects of nighttime driving performance. Mesopic tests of visual function, including motion sensitivity and mesopic high contrast VA, were more strongly associated with nighttime driving performance than photopic high-contrast VA. These results highlight the potential importance of nonstandard vision tests for assessing older drivers' visual capacity to drive at night.

Effects of Intersection Lighting Design on Nighttime Visual Performance of Drivers

ABSTRACTNighttime crashes at intersections present a major traffic safety issue in the United States. Existing approaches to intersection lighting design do not account for a driver’s visual performance or the potential interactive effects of vehicle headlamps and roadway lighting. For effective design lighting at intersections, empirical research is required to evaluate the effects of lighting configuration (part of the intersection illuminated) and lighting levels on nighttime driver visual performance. The current study had two goals: First, to quantify visual performance in three lighting configurations (illuminating the intersection box, approach, or both) and second, to determine what lighting levels within each lighting configuration support the best visual performance. The study involved a target detection task completed at night on a realistic roadway intersection. Illuminating the intersection box led to superior visual performance, as indicated by longer target detection distances, fewer missed targets, and more targets identified within a safe stopping distance. For this lighting configuration, visual performance plateaued between 7 and 10 lx of mean intersection illuminance. These results have important implications for the design of intersection lighting at isolated/rural intersections, specifically that illuminating the intersection box is an effective strategy to increase nighttime visual performance for a wider range of driver ages and could also be an energy-efficient solution.

Association between Sleep Disordered Breathing and Nighttime Driving Performance in Mild Cognitive Impairment.

The effect of sleep disordered breathing (SDB) on driving performance in older adults has not been extensively investigated, especially in those with mild cognitive impairment (MCI). The aim of this study was to examine the relationship between severity measures of SDB and a simulated driving task in older adults with and without MCI. Nineteen older adults (age ≥50) meeting criteria for MCI and 23 age-matched cognitively intact controls underwent neuropsychological assessment and a driving simulator task in the evening before a diagnostic sleep study. There were no differences in driving simulator performance or SDB severity between the two groups. In patients with MCI, a higher oxygen desaturation index (ODI) was associated with an increased number of crashes on the simulator task, as well as other driving parameters such as steering and speed deviation. Poorer driving performance was also associated with poorer executive functioning (set-shifting) but the relationship between ODI and crashes was independent of executive ability. While driving ability did not differ between older adults with and without MCI, oxygen saturation dips in MCI were related to worse driving performance. These results suggest that decreased brain integrity may render those with SDB particularly vulnerable to driving accidents. In older adults, both cognition and SDB need to be considered concurrently in relation to driving ability. (JINS, 2017, 23, 502-510).

Recommended Guidelines for Nighttime Overhead Sign Visibility

Guidelines for the provision of effective nighttime performance of overhead signs were developed. Relevant policies and guidelines with regard to sign lighting currently provide little useful information to determine when sign lighting is needed, and the reference material available to practitioners is out of date. Two complementary nighttime visibility studies were conducted, which were designed to produce results useful in developing updated guidelines for overhead sign visibility. The first was conducted on a closed course and investigated the legibility distances of three sign legend and background configurations under various sign lighting treatments. The second was conducted on the open road. It investigated the effects of sign luminance and visual complexity on the distance at which a driver can read overhead signs during a recognition task. The combined findings were used to develop revised guidelines designed to provide adequate nighttime visibility of overhead signs. The proposed guidelines are based on the needs of nighttime motorists and have been formatted specifically for AASHTO’s Roadway Lighting Design Guide, which is being updated. The revised chapter on roadway sign lighting was provided to the AASHTO task force responsible for revisions. The guidelines include a list of recommended retroreflective sheeting materials that can be used to meet nighttime driver needs for specific complexity levels.

Do Nighttime Driving Restrictions Reduce Criminal Participation Among Teenagers? Evidence from Graduated Driver Licensing

To date, all 50 states and the District of Columbia have a three-stage Graduated Driver Licensing (GDL) system that phases in driving privileges for teenagers. GDL laws effectively impose a statutory driving curfew and a limitation on the number of passengers in motor vehicles. Both the timing of motor vehicle access and a limitation on the peer influences available in a motor vehicle could significantly affect the production of criminal behavior. Using the Uniform Crime Reports 1995-2011 and a triple differences approach, we find that the implementation of GDL decreased criminal participation by six percent among teenagers between the ages of 16 and 17, as measured by arrests. These effects are larger in magnitude in states where the nighttime driving curfew is required for a longer period of time. We also show that GDL plays an important role in reducing crime in periods of low gasoline prices, a time when teen driver prevalence would otherwise have been high. These results suggest that there is another benefit to states for adopting GDL laws and provide insight into the production of teenage crime.