Enhanced YOLO-Based Vehicle Detection in Indian Traffic under Heavy Rainfall Using AI Based Image Detection

Abstract

Vehicle Detection has proven to be an important function for ITS, Surveillance systems, and autonomous vehicles. One of the efficient DL methods used for this purpose are YOLO algorithms, which include YOLO v2, v3, v5, v7, and v8. While there are continuous efforts being made to improve the efficiency of the system, bad weather conditions such as rains are still an issue in the process. Some of the issues caused by bad weather are rain, poor visibility, low contrast, reflections, and motion blur, which create difficulties in feature extraction and localization, thus decreasing the detection rate. Another issue is that more and more types of vehicles can now be seen in India like cars, buses, trucks, two-wheelers, auto rickshaws, and e-rickshaws. In this study, a review of YOLO based vehicle detection during heavy rainfall is carried out. Some of the areas included in the study are usage of YOLO-based algorithms for vehicle detection, real-time vehicle detection, weather dependent vehicle detection, image pre-processing techniques, namely deraining, adaptive image pre-processing, and retinex based pre-processing techniques. Alongside, the deficiencies of Indian traffic scenes have also been mentioned along with possible future work.

Introduction

The text discusses a YOLOv8-based vehicle detection system designed for Indian traffic under rainy conditions, motivated by the growing need for real-time intelligent traffic monitoring in ITS, autonomous driving, and smart surveillance.

It explains that while YOLO models (from YOLOv1 to YOLOv8) are widely used for fast and accurate object detection, they still struggle in adverse weather conditions like heavy rain, where issues such as blur, streak noise, low contrast, and reflections reduce detection accuracy. This challenge is more severe in India due to dense, heterogeneous traffic (cars, buses, trucks, bikes, auto-rickshaws, e-rickshaws).

To address this, the proposed approach integrates:

• AI-based image enhancement (deraining, contrast improvement, noise reduction) as preprocessing
• YOLOv8n model for real-time vehicle detection and classification
• A pipeline: Input → Enhancement → YOLOv8 Detection → Output

The methodology includes dataset collection (Indian traffic images with normal and rainy conditions), annotation using Roboflow, and evaluation using metrics like precision, recall, mAP, and IoU.

The literature review highlights YOLO’s evolution and its success in traffic applications, but also shows that most existing weather-robust models are tested on simulated datasets, not real Indian monsoon conditions.

The key research gap is the lack of:

• Real-world Indian rainy traffic datasets
• Integrated systems combining image enhancement + YOLO detection
• Robust, real-time (≥30 FPS) systems suitable for edge devices
• Handling of visually similar classes like bikes, scooters, and rickshaws

Conclusion

This paper has discussed different models applied for detection of vehicles in India in rainy seasons through the YOLO algorithm. It is important to understand the concept of object detection starting from fundamentals to latest YOLOv8 model.It is clearly understood from the review of literature that while there have been several effective real-time traffic surveillance solutions that use YOLO, most of them perform poorly when it comes to weather conditions such as heavy rainfall [14]. Moreover, heavy rain poses challenges in terms of poor visibility, motion blur, low contrast, and streaks of noise, all of which make it hard to detect and localize objects in the image. Furthermore, it is clear that this paper has pointed to several key gaps in research in the current literature. While there are many papers in the literature that are based on conventional and well-controlled datasets, only a handful of papers address issues related to complex and densely populated Indian traffic conditions. Also, no effort is made towards developing image enhancement techniques to improve detection in monsoon rainfall conditions . To address the above-stated issues, an AI-based theoretical framework has been formulated to facilitate effective image preprocessing and incorporate it in the YOLOv8 model. In particular, the proposed approach consists of data gathering from Google Script, annotations through Roboflow, AI-based image enhancement preprocessing, and object detection via YOLOv8. The implementation of such a technique would enable improving the effectiveness of vehicle recognition under adverse weather conditions. Therefore, the outcomes obtained during the study have indicated that the combination of preprocessing techniques and advanced deep learning models is crucial to guarantee their efficient performance within traffic settings. The proposed approach would be of great importance in developing weather-proof ITSs appropriate for Indian highways.

References

[1] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 779–788. [2] J. Redmon and A. Farhadi, “YOLO9000: Better, Faster, Stronger,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 7263–7271. [3] J. Redmon and A. Farhadi, “YOLOv3: An Incremental Improvement,” arXiv:1804.02767, 2018. [4] Z. Ge, S. Liu, F. Wang, Z. Li, and J. Sun, “YOLOX: Exceeding YOLO Series in 2021,” arXiv:2107.08430, 2021. [5] G. Jocher et al., “Ultralytics YOLOv5,” GitHub Repository, 2020. [Online]. Available: https://github.com/ultralytics/yolov5 [6] Ultralytics, “YOLOv8 Documentation,” 2023. [Online]. Available: https://docs.ultralytics.com [7] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 6, pp. 1137–1149, 2017. [8] X. Zhang, Y. Wang, and J. Li, “Image Deraining Using Deep Learning for Traffic Surveillance Systems,” IEEE Access, vol. 9, pp. 112345–112356, 2021. [9] Y. Li, R. T. Tan, X. Guo, J. Lu, and M. S. Brown, “Rain Streak Removal Using Layer Priors,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 2736–2744. [10] H. Wang, Q. Xie, and Y. Zhang, “Vehicle Detection in Adverse Weather Conditions Using Deep Learning,” Sensors, vol. 22, no. 4, pp. 1–15, 2022. [11] W. Luo, X. Wang, and X. Tang, “Pedestrian Detection in Adverse Weather Conditions,” IEEE Trans. Image Process., vol. 27, no. 1, pp. 1–14, 2018. [12] A. Bochkovskiy, C. Wang, and H. M. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” arXiv:2004.10934, 2020. [13] C. Wang, A. Bochkovskiy, and H. M. Liao, “YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors,” arXiv:2207.02696, 2022. [14] S. Garg and S. K. Nayar, “Vision and Rain,” Int. J. Comput. Vis., vol. 75, no. 1, pp. 3–27, 2007. [15] A. Mittal, T. Zickler, and W. T. Freeman, “Enhancing and Restoring Videos Distorted by Atmospheric Turbulence,” ACM Trans. Graph., vol. 25, no. 3, pp. 795–804, 2006. [16] E. H. Land and J. J. McCann, “Lightness and Retinex Theory,” J. Opt. Soc. Am., vol. 61, no. 1, pp. 1–11, 1971. [17] M. Teichmann et al., “MultiNet: Real-Time Joint Semantic Reasoning for Autonomous Driving,” in Proc. IEEE Intell. Veh. Symp. (IV), 2018, pp. 1013–1020.

Copyright

Copyright © 2026 Ms. Yadnika Bhanushali, Mr. Mohit Saxena, Ms. Pruthvi Khadpekar, Ms. Shrutika Shingate. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.