Enhancing the Safety of Autonomous Vehicles in Adverse Weather by Deep Learning-Based Object Detection

Biwei Zhang,Murat Simsek,Michel Kulhandjian,Burak Kantarci

DOI: https://doi.org/10.3390/electronics13091765

IF: 2.9

2024-05-03

Electronics

Abstract:Recognizing and categorizing items in weather-adverse environments poses significant challenges for autonomous vehicles. To improve the robustness of object-detection systems, this paper introduces an innovative approach for detecting objects at different levels by leveraging sensors and deep learning-based solutions within a traffic circle. The suggested approach improves the effectiveness of single-stage object detectors, aiming to advance the performance in perceiving autonomous racing environments and minimizing instances of false detection and low recognition rates. The improved framework is based on the one-stage object-detection model, incorporating multiple lightweight backbones. Additionally, attention mechanisms are integrated to refine the object-detection process further. Our proposed model demonstrates superior performance compared to the state-of-the-art method on the DAWN dataset, achieving a mean average precision (mAP) of 99.1%, surpassing the previous result of 84.7%.

engineering, electrical & electronic,physics, applied,computer science, information systems

What problem does this paper attempt to address?

The paper aims to address the challenges of object detection for autonomous vehicles under adverse weather conditions to improve their safety. Specifically, the authors propose an innovative deep learning-based approach to enhance the robustness of the object detection system by leveraging sensor data and an improved one-stage object detection model. The method is mainly improved through the following aspects:

1. **Multiple Lightweight Backbone Networks**: Introduced multiple lightweight backbone networks (such as ShuffleNetV2, GhostNet, VoVNet) combined with attention mechanisms (such as SE block, CBAM block, ECA block) to enhance detection performance in adverse environments.

2. **Transformer Prediction Head (TPH) Integration**: Integrated TPH into YOLOv5 to enhance object localization capabilities in adverse scenarios.

3. **Model Simplification and Optimization**: Employed techniques such as pruning, quantization, and distillation to simplify the framework, specifically reducing computational costs and model size for adverse weather-related issues.

4. **CBAM Enhancement**: Incorporated CBAM modules into YOLOv5 to improve the network's ability to recognize regions of interest in images and used a self-training classifier to further enhance object classification accuracy.

Through these improvements, the proposed model in the paper achieved a mean Average Precision (mAP) of 99.1% on the DAWN dataset, significantly surpassing the previous result of 84.7%, demonstrating strong performance in detecting objects under adverse weather conditions. These improvements are not limited to autonomous vehicles but are also applicable to drone object detection, pedestrian safety warning systems, and intelligent transportation systems that rely on vehicle-to-infrastructure communication.