Pedestrian Detection in Low-Light Conditions: A Comprehensive Survey

TL;DR

This survey addresses the problem of pedestrian detection in low-light conditions, focusing on nighttime scenarios crucial for autonomous driving and safety. It provides a comprehensive taxonomy of methods, spanning handcrafted features, deep learning-based image fusion (early, halfway, and late), and hybrid approaches, while detailing influential datasets (notably KAIST) and benchmarking datasets such as NightOwls, FLIR, and SCUT. The authors analyze trends, datasets, and performance metrics across 118 papers, highlighting the dominance of cross-spectral fusion and the prevalence of deep learning, along with gaps like moving-camera multi-modal datasets and the need for explainable AI. The work aims to guide researchers and practitioners toward safer, more robust nighttime pedestrian detection and to inform future research directions in this safety-critical domain.

Abstract

Pedestrian detection remains a critical problem in various domains, such as computer vision, surveillance, and autonomous driving. In particular, accurate and instant detection of pedestrians in low-light conditions and reduced visibility is of utmost importance for autonomous vehicles to prevent accidents and save lives. This paper aims to comprehensively survey various pedestrian detection approaches, baselines, and datasets that specifically target low-light conditions. The survey discusses the challenges faced in detecting pedestrians at night and explores state-of-the-art methodologies proposed in recent years to address this issue. These methodologies encompass a diverse range, including deep learning-based, feature-based, and hybrid approaches, which have shown promising results in enhancing pedestrian detection performance under challenging lighting conditions. Furthermore, the paper highlights current research directions in the field and identifies potential solutions that merit further investigation by researchers. By thoroughly examining pedestrian detection techniques in low-light conditions, this survey seeks to contribute to the advancement of safer and more reliable autonomous driving systems and other applications related to pedestrian safety. Accordingly, most of the current approaches in the field use deep learning-based image fusion methodologies (i.e., early, halfway, and late fusion) for accurate and reliable pedestrian detection. Moreover, the majority of the works in the field (approximately 48%) have been evaluated on the KAIST dataset, while the real-world video feeds recorded by authors have been used in less than six percent of the works.

Figures (10)

• Figure 1: Challenges of detecting pedestrians at night (image taken from 2019 Traffic Safety conference nighttime visibility report by Texas A&M Transportation Institute).
• Figure 2: Distributions of the papers surveyed in the current research work that only focus on pedestrian detection at low-light scenarios from 2016-2023 (total: 118 papers).
• Figure 3: Instances of some datasets introduced for nighttime pedestrian detection. It should be noted that the collected image or video sequences were captured using various sensors.
• Figure 4: The overall diagram of night-time pedestrian detection methodologies.
• Figure 5: The primary classification of different nighttime pedestrian detection methodologies considered in this survey.