CoChat

Object Detection in Autonomous Vehicles under Adverse Weather: A Review of Traditional and Deep Learning Approaches

Open access

The DOI matches its Crossref record; 162 of the 162 references it cites were found in a registry or an index; 18 of the 89 sources its claims cite were checked for support.

Abstract

Enhancing the environmental perception of autonomous vehicles (AVs) in intelligent transportation systems requires computer vision technology to be effective in detecting objects and obstacles, particularly in adverse weather conditions. Adverse weather circumstances present serious difficulties for object-detecting systems, which are essential to contemporary safety procedures, infrastructure for monitoring, and intelligent transportation. AVs primarily depend on image processing algorithms that utilize a wide range of onboard visual sensors for guidance and decisionmaking. Ensuring the consistent identification of critical elements such as vehicles, pedestrians, and road lanes, even in adverse weather, is a paramount objective. This paper not only provides a comprehensive review of the literature on object detection (OD) under adverse weather conditions but also delves into the ever-evolving realm of the architecture of AVs, challenges for automated vehicles in adverse weather, the basic structure of OD, and explores the landscape of traditional and deep learning (DL) approaches for OD within the realm of AVs. These approaches are essential for advancing the capabilities of AVs in recognizing and responding to objects in their surroundings. This paper further investigates previous research that has employed both traditional and DL methodologies for the detection of vehicles, pedestrians, and road lanes, effectively linking these approaches with the evolving field of AVs. Moreover, this paper offers an in-depth analysis of the datasets commonly employed in AV research, with a specific focus on the detection of key elements in various environmental conditions, and then summarizes the evaluation matrix. We expect that this review paper will help scholars to gain a better understanding of this area of research.

Overview

CoChat Summary

TL;DR

This review surveys traditional and deep learning (DL) approaches to object detection (vehicles, pedestrians, road lanes) for autonomous vehicles under adverse weather, covering AV architecture, sensor technologies, datasets, and evaluation metrics. It concludes that DL methods are used far more intensively than traditional ones and perform better in challenging weather, while noting that the reasons for one-stage detectors' lower accuracy remain unclear.

Key findings

  • DL approaches were more frequently used than traditional approaches for all three detection tasks; vehicle detection was the most studied, followed by road lane and pedestrian detection (Figures 13–14).
  • One-stage detectors (e.g., YOLO series, SSD) were frequently used for vehicle detection, while two-stage detectors (e.g., Faster R-CNN) were frequently used for road lanes; for pedestrian detection both were applied equally (Figure 15).
  • Reported accuracies from reviewed works include: ResNet 100%, GoogleNet 65.50%, SqueezeNet 90.0% for vehicle detection in fog/low light/sun glare [131]; YOLOv5 with Transformer and CBAM at 94.7% mAP and 199.86 FPS [140]; Fast R-CNN at 98.44% recall, 94.20% accuracy, 90% precision [132]; Faster R-CNN at 95.16% on LISA, 91.20% on CDNet 2014, 89.48% on DAWN [133]; improved SSD at 92.18% mAP on KITTI at 15 ms/frame [134]; YOLOv3 at 93.66% average precision for nighttime detection [137]; Mask R-CNN improved by 4.66% to 85.05% in low light [147]; HOG-SVM pedestrian recognition at 98.00% on 2000 road photos [122].
  • Traditional methods (SIFT, Viola-Jones, Haar-like, HOG, SVM) are limited by manual feature engineering, sensitivity to lighting/occlusion/weather, and poor adaptability; DL learns representations automatically and generalizes better across weather conditions.

Methods

  • Design: narrative/systematic literature review of traditional and DL object detection approaches for AVs under adverse weather, covering AV architecture, sensor technologies (LiDAR, radar, ultrasonic, camera), datasets, and evaluation metrics.
  • Data: literature identified via systematic search across Springer Link, Elsevier, IEEE, MDPI, and Wiley Online Library, restricted to peer-reviewed journals and international conferences published 2001–2023.
  • Analysis: qualitative synthesis and tabulation of prior works (Tables 3–7), plus distribution counts of papers by approach and detector type (Figures 13–15); no primary experiments were conducted.

Limitations

  • Stated by the authors: the review focuses only on pedestrians, vehicles, and road lanes, omitting other objects (e.g., traffic signals, animals); reviewed algorithms struggle with small or occluded objects; some environmental factors (dust storms, heavy fog) were not extensively covered; rapid field evolution may mean recent advances are missed; simulated adverse-weather datasets may not replicate real-world complexity; ethical considerations and societal acceptance are not addressed.
  • Evident from the text: the provided full text is fragmented, with several sections (e.g., 6.1.2, 6.3, 7.1, 7.2, 8) appearing only partially or out of order, and some tables are garbled, so reported figures could not always be verified against their source context.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/a17030103

Cites

162 of 162 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 The global self-driving cars market size 2022 - 2032 in USD Million [3].
Figure 1 The global self-driving cars market size 2022 - 2032 in USD Million [3].CC-BY · From the PDF
Figure 2 Object identification strategies.
Figure 2 Object identification strategies.CC-BY · From the PDF
Figure 3 Paper organization, where OD= object detection.
Figure 3 Paper organization, where OD= object detection.CC-BY · From the PDF
Figure 4 SAE automation levels for AVs.
Figure 4 SAE automation levels for AVs.CC-BY · From the PDF
Figure 5 An architecture of AVs.
Figure 5 An architecture of AVs.CC-BY · From the PDF
Figure 6 Weather impacts on AVs.
Figure 6 Weather impacts on AVs.CC-BY · From the PDF
Figure 7 The basic framework for object detection systems.
Figure 7 The basic framework for object detection systems.CC-BY · From the PDF
Figure 8 Performance graph of traditional and DL approaches.
Figure 8 Performance graph of traditional and DL approaches.CC-BY · From the PDF
Figure 9 Primary structure of feature extraction for traditional and DL approaches.
Figure 9 Primary structure of feature extraction for traditional and DL approaches.CC-BY · From the PDF
Figure 10 Traditional and DL approaches.
Figure 10 Traditional and DL approaches.CC-BY · From the PDF
Figure 11 The architecture of Faster R-CNN [90].
Figure 11 The architecture of Faster R-CNN [90].CC-BY · From the PDF
Figure 12 Structure of YOLOv8.
Figure 12 Structure of YOLOv8.CC-BY · From the PDF
Figure 13 Overall percentage distribution of papers.
Figure 13 Overall percentage distribution of papers.CC-BY · From the PDF
Figure 14 Distribution of papers for traditional and DL approaches.
Figure 14 Distribution of papers for traditional and DL approaches.CC-BY · From the PDF
Figure 15 Paper distribution for one-stage and two-stage detectors.
Figure 15 Paper distribution for one-stage and two-stage detectors.CC-BY · From the PDF

Verification

Claims the paper makes

16 Valid · 6 Not valid · 70 Not checked · 38 Own results and interpretations

25 more claims

Create a free account to see all 130 claims and to verify the 25 references still open.

See all claims — free account
  • Probably not a substantial claim. 54
  • It may state two separate claims as one. 72
  • The quote only partly states this claim. 3

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

3 of 3

Cited works that were retracted

Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.

2 of 2

Citations

Cites

2 retracted

Research works printed but not linked: 6

  1. Chehri, A., Sharma, T., Debaque, B., Duclos, N., and Fortier, P. (2022). Sustainability in Energy and Buildings 2021, Springer.
  2. Pomerleau, D.A. NIPS’88: Proceedings of the 1st International Conference on Neural Information Processing Systems, 1 January 1988, MIT Press.
  3. Briefs, 2015, Mcity Grand Opening, Res. Rev., 46, 1
  4. Bellone, M., Ismailogullari, A., Müür, J., Nissin, O., Sell, R., and Soe, R.M. (2021). Towards Connected and Autonomous Vehicle Highways: Technical, Security and Social Challenges, Springer International Publishing.
  5. Jang, G., Park, J., and Kim, M. (2016). Advances in Computer Science and Ubiquitous Computing, Springer.
  6. Ghosh, R. (2021). Intelligent Computing and Communication Systems, Springer.

Cited by

See all 145

Metadata

Authors
Tahir, N. U. A., Zhang, Z., Asim, M., Chen, J., & ELAffendi, M.
Affiliations
  • Central South University (CN)
  • Guangdong University of Technology (CN)
  • Prince Sultan University (SA)
Corresponding author
  • Zhang, Zuping
  • Asim, Muhammad
Venue
Algorithms
Publisher
MDPI AG
Published
2024-02-26
Volume
17
Issue
3
Pages
103
ISSN
1999-4893 (ISSN-L)
OpenAlex
W4392169182
Cited by
145
References
159
Citations by year
2026: 54 · 2025: 77 · 2024: 14
FWCI
15.56
Type
journal article
Language
en
Open access
Locations
  • Algorithms (publishedVersion, cc-by)
  • Document Server@UHasselt (UHasselt) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
  • EIAS Data Science Lab, CCIS, Prince Sultan University (BOF23DOCBL11, 202208440309)
  • Special Research Fund (BOF) of Hasselt University (BOF23DOCBL11, 202208440309)
  • China Scholarship Council (BOF23DOCBL11, 202208440309)
Topics
  • Physical Sciences › Engineering › Automotive Engineering › Autonomous Vehicle Technology and Safety
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
Keywords
object detection, autonomous vehicles, adverse weather, environmental perception, vehicle detection, pedestrian detection, lane detection, deep learning object detection, benchmark dataset, evaluation metrics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.