MRD-Net: Multi-scale Refinement Dehazing Network for Autonomous Driving Perception Images

Author:

Wang Juan1,Wang Sheng1,Wu Minghu1,Yang Hao1,Cao Ye1,Hu Shuyao1,Shao Jixiang1,Zeng Chunyan1

Affiliation:

1. Hubei University of Technology

Abstract

Abstract

In hazy weather conditions, a significant accumulation of haze poses a severe challenge to the quality of image capture for autonomous driving systems, thereby heightening safety risks for autonomous vehicles. To tackle this issue, we propose the Multi-scale Refinement Dehazing Network (MRD-Net), an innovative architecture comprising a front-end module, a backbone module, and a tail-end module, specifically designed to eradicate haze with precision. To enhance the extraction of multi-scale features, the backbone module employs the Squeeze-Excitation Residual Dense Block (SRD). It not only learns the intricate multi-scale features of the image but also adaptively recalibrates the feature response of each feature map, ultimately bolstering the network's performance and resilience. The tail-end module, crafted with the Dilation Refinement Block (DRB), serves as a compensatory measure for any detail loss or pseudo-artifacts that might arise from the backbone module's operations. By incorporating this refinement block, the overall dehazing effect is further optimized. Empirical evaluations reveal that the proposed MRD-Net achieves impressive results, with a PSNR value of 28.12, an SSIM value of 0.964, and an LPIPS value of 0.032. These figures indicate that the network is adept at removing haze from images while preserving intricate details, ensuring the efficacy and reliability of autonomous driving systems in hazy environments.

Publisher

Research Square Platform LLC

Reference35 articles.

1. Wu, Junming and Li, Wanlin P-2.21: Research on multi-lidar detection technology for autonomous driving. 54: 558--561 https://doi.org/https://doi.org/10.1002/sdtp.16355, Autonomous driving, multi-lidar detection, Point cloud registration, \_eprint: https://sid.onlinelibrary.wiley.com/doi/pdf/10.1002/sdtp.16355, 2023, {SID} Symposium Digest of Technical Papers, S1, https://sid.onlinelibrary.wiley.com/doi/abs/10.1002/sdtp.16355

2. Dhruva, Meghna and Khared, Ishita and Adhia, Krishna and Mesvan, Nikite Real-Time Lane Detection and Object Recognition in Self-Driving Car using {YOLO} neural network and Computer Vision. 9: 2278--3075 https://doi.org/10.35940/ijitee.G1010.0597S20, 2020-05, International Journal of Innovative Technology and Exploring Engineering

3. Wang, Ziquan and Zhang, Yongsheng and Zhang, Zhenchao and Jiang, Zhipeng and Yu, Ying and Li, Li and Zhang, Lei {SDAT}-Former + +: A Foggy Scene Semantic Segmentation Method with Stronger Domain Adaption Teacher for Remote Sensing Images. 15(24) https://doi.org/10.3390/rs15245704, 2023, Remote Sensing, https://www.mdpi.com/2072-4292/15/24/5704, 2072-4292

4. Bissonnette, Luc R. Imaging through fog and rain. 31(5): 1045 -- 1052 https://doi.org/10.1117/12.56145, Aerosols, Atmospheric modeling, Atmospheric particles, Fiber optic gyroscopes, image propagation, Mass attenuation coefficient, modulation transfer function, Modulation transfer functions, multiple scattering, Particles, point spread function, Point spread functions, Receivers, Scattering, Publisher: {SPIE}, 1992, Optical Engineering, https://doi.org/10.1117/12.56145

5. Yoon, Sungan and Cho, Jeongho Deep Multimodal Detection in Reduced Visibility Using Thermal Depth Estimation for Autonomous Driving. 22(14) https://doi.org/10.3390/s22145084, 2022, Sensors, Recently, the rapid development of convolutional neural networks ({CNN}) has consistently improved object detection performance using {CNN} and has naturally been implemented in autonomous driving due to its operational potential in real-time. Detecting moving targets to realize autonomous driving is an essential task for the safety of drivers and pedestrians, and {CNN}-based moving target detectors have shown stable performance in fair weather. However, there is a considerable drop in detection performance during poor weather conditions like hazy or foggy situations due to particles in the atmosphere. To ensure stable moving object detection, an image restoration process with haze removal must be accompanied. Therefore, this paper proposes an image dehazing network that estimates the current weather conditions and removes haze using the haze level to improve the detection performance under poor weather conditions due to haze and low visibility. Combined with the thermal image, the restored image is assigned to the two You Only Look Once ({YOLO}) object detectors, respectively, which detect moving targets independently and improve object detection performance using late fusion. The proposed model showed improved dehazing performance compared with the existing image dehazing models and has proved that images taken under foggy conditions, the poorest weather for autonomous driving, can be restored to normal images. Through the fusion of the {RGB} image restored by the proposed image dehazing network with thermal images, the proposed model improved the detection accuracy by up to 22% or above in a dense haze environment like fog compared with models using existing image dehazing techniques., https://www.mdpi.com/1424-8220/22/14/5084, 1424-8220

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3