针对降雨模拟方法的机器视觉真实度对比研究

doi:10.19562/j.chinasae.qcgc.2025.03.007

Abstract

Abstract:

Given the high exposure and risk of rainfall as a trigger condition for visual perception systems， various rainfall simulation tests are the main research methods. However， the realism of rain simulation of different testing methods impacts the confidence in test conclusions. In this study indicators are selected to quantify the impact of rainfall on machine vision from the aspects of image quality and object detection. Using the numerical range and trend of index changes under real rainfall as a benchmark， the comparative study of the realism of different rainfall simulation methods in the dimension of machine vision is carried out. Additionally， in this study 1 950 images of no rain and various levels of real rainfall are collected to construct a dataset， so as to obtain statistical patterns of the impact of real rainfall on machine vision. Two simulated rainfall test sites， three simulation software， and one generative model are selected for rainfall simulation tests to compare and analyze the realism of different types of rainfall simulation methods horizontally. The results show that， in terms of image quality， simulation software and rainfall simulation equipment can better simulate the real rain in terms of DR value range and trend. Regarding target detection， simulation software and generative model are closer to real rainfall in terms of CC change values. Overall， in terms of realism， digital simulation of rainfall performs best， followed by physical rainfall simulation on site and generative model， providing a reference for testing the SOTIF of the visual perception system of intelligent and connected vehicles.

Key words: rainfall simulation, realism assessment, machine vision, visual perception system

Junyi Chen,Tian Xia,Zhenyuan Liu,Tong Jia,Xiaoyi Wang,Xuehan Ma,Xingyu Xing,Jianfeng Wu. A Comparative Study on the Machine Vision Realism of Rainfall Simulation Methods[J].Automotive Engineering, 2025, 47(3): 449-459.

Figures/Tables 31

数据集	有无雨标注	标注类型	是否真实	场景多样性	数量/ $103$ 张
KITTI^［17］	无		是	高	15
BDD100K^［18］	无		是	高	100
Cityscapes^［19］	无		是	高	25
CADC^［20］	无		是	高	7.50
Nuscense^［21］	有	有雨/无雨	是	高	40
Boreas^［22］	有	有雨/无雨	是	中	约880
Rain800^［23］	有	有雨/无雨	否（人工合成）	中	0.80
Rain12600^［24］	有	14种降雨	否（人工合成）	中	12.60
SPA^［25］	有	有雨/无雨	是	低	0.15
SIRR^［26］	有	有雨/无雨	是	低	1
NMRD （ours）	有	无雨/大雨/ 中雨/小雨	是	中	1.95

降雨等级	DR数值范围	DR趋势	MR数值范围	$?$ CC数值范围	综合真实度
场地物理模拟降雨A	★☆☆	☆☆	/	★☆☆	☆☆☆
场地物理模拟降雨B	★☆☆	★★	/	★☆☆	★☆☆
仿真数字模拟降雨A	★★★	★☆	/	★★☆	★★☆
仿真数字模拟降雨B	★★☆	☆☆	/	★★☆	★☆☆
仿真数字模拟降雨C	☆☆☆	★★	/	★☆☆	★☆☆
生成模型模拟降雨	☆☆☆	★☆	/	★★☆	★☆☆

References 29

1	International Organization for Standardization. Road vehicles-safety of the intended functionality： ISO 21448 ［S］. Geneva， Switzerland： ISO， 2022： 1.
2	HOSS M， SCHOLTES M， ECKSTEIN L. A review of testing object-based environment perception for safe automated driving ［J］.Aulomotive Innovation，2022，5（3）： 223-250.
3	陈君毅，刘镇源，杨雪珠，等.感知系统触发条件穿透率评估方法研究［J］.汽车工程，2024，46（4）：577-587.
	CHEN J Y， LIU Z Y， YANG X Z， et al. Evalution method for the penetration rate of perception system triggering conditions［J］. Automotive Engineering， 2024， 46（4）：577-587.
4	Working Party on Automated/Autonomous and Connected Vehicles. Proposal for a second iteration of the new assessment/test method for automated driving-master document ［Informal Document］［Z］. United Nations，2022.
5	OICA. Future certification of automated/autonomous driving systems ［EB/OL］. ［2019-01-28］. https：//unece.org/fileadmin/DAM/trans/doc/2019/wp29grva/GRVA-02-09e.pdf.
6	JIANG W， XING X Y， HUANG A， et al. Research on performance limitations of visual-based perception system for autonomous vehicle under severe weather conditions［C］. IEEE Intelligent Vehicles Symposium （IV）. IEEE， 2022： 586-593.
7	蒋渊德，孙朋朋，秦孔建，等.雨雪天气对自动驾驶视觉图像质量的影响［J］.中国公路学报， 2022， 35（3）： 307-316.
	JIANG D Y， SUN P P， QIN K J， et al. Influence of rain and snow on autonomous vehicles’ camera［J］. China Journal of Highway and Transport， 2022， 35（3）：307-316.
8	WEI Y， ZHANG Z， WANG Y， et al. Deraincyclegan： rain attentive cyclegan for single image deraining and rainmaking［J］. IEEE Transactions on Image Processing， 2021， 30： 4788-4801.
9	SON W， HA Y， OH T， et al. PG-based vehicle-in-the-loop simulation for system development and consistency validation［J］. Electronics， 2022， 11（24）： 4073.
10	XU W， SOULY N， BRAHMA P P. Reliability of GAN generated data to train and validate perception systems for autonomous vehicles［C］. 2021 IEEE Winter Conference on Applications of Computer Vision Workshops （WACVW）. IEEE， 2021： 171-180.
11	BRENNER J F， DEW B S， HORTON J B， et al. An automated microscope for cytologic research a preliminary evaluation［J］. Journal of Histochemistry & Cytochemistry， 1976， 24（1）： 100-111.
12	TENENBAUM J M. Accommodation in computer vision［M］. Stanford University， 1971.
13	REN S， HE K， GIRSHICK R， et al. Faster R-CNN： towards real-time object detection with region proposal networks［J］. IEEE Transactions on Pattern Analysis & Machine Intelligence， 2017， 39（6）：1137-1149.
14	REDMON J， DIVVALA S， GIRSHICK R， et al. You only look once： unified， real-time object detection［C］. 2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. IEEE， 2016： 779-788.
15	MA M， PANG H. SP-YOLOv8s： an improved YOLOv8s model for remote sensing image tiny object detection［J］. Appl. Sci.， 2023， 13： 8161.
16	JHONG S Y， CHEN Y Y， HSIA C H， et al. Nighttime object detection system with lightweight deep network for internet of vehicles［J］. Journal of Real-Time Image Processing， 2021， 18（4）： 1141-1155.
17	MENZE M， GEIGER A. Object scene flow for autonomous vehicles［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition， 2015： 3061-3070.
18	YU F， CHEN H， WANG X， et al. BDD100K： a diverse driving dataset for heterogeneous multitask learning［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition， 2020： 2636-2645.
19	CORDTS M， OMRAN M， RAMOS S， et al. The cityscapes dataset［C］. CVPR Workshop on the Future of Datasets in Vision. sn， 2015， 2.
20	PITROPOV M， GARCIA D E， REBELLO J， et al. Canadian adverse driving conditions dataset［J］. The International Journal of Robotics Research， 2021， 40（4-5）： 681-690.
21	CAESAR H， BANKITI V， LANG A H， et al. nuScenes： a multimodal dataset for autonomous driving［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition， 2020： 11621-11631.
22	BURNETT K， YOON D J， WU Y， et al. Boreas： a multi-season autonomous driving dataset［J］. arXiv preprint arXiv：， 2022.
23	ZHANG H， SINGDAGI V， PATEL V M. Image de-raining using a conditional generative adversarial network［J］. IEEE Transactions on Circuits and Systems for Video Technology， 2019， 30（11）： 3943-3956.
24	FU X， HUANG J， ZENG D， et al. Removing rain from single images via a deep detail network［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition， 2017： 3855-3863.
25	WANG T， YANG X， XU K， et al. Spatial attentive single-image deraining with a high quality real rain dataset［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition， 2019： 12270-12279.
26	WEI W， MENG D， ZHAO Q， et al. Semi-supervised transfer learning for image rain removal［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition， 2019： 3877-3886.
27	LIU Z Y， JIA T， XING X Y， et al. Hierarchical-level rain image generative model based on GAN［J］. arXiv preprint arXiv：， 2023.
28	中国气象服务协会.短时气象服务降雨量等级：T/CMSA 0013—2019［S］.2019.
29	ZHU J Y， PARK T， ISOLA P， et al. Unpaired image-to-image translation using cycle-consistent adversarial networks［C］. 2017 IEEE International Conference on Computer Vision （ICCV）. IEEE， 2017： 2223-2232.

参数类型	参数名称	参数设置
参数类型	参数名称	无雨场景	小雨场景	中雨场景	大雨场景
目标物参数	行人型号	walker.pedestrian.0016
	行人纵向距离	20 m
	雨	0	10%	20%	50%
	云	0	70%	80%	90%
	地面积水	0	50%	70%	100%
	湿度	0	10%	20%	40%

参数类型	参数名称	参数设置
参数类型	参数名称	无雨场景	小雨场景	中雨场景	大雨场景
目标物参数	行人型号	PedWithUmbrella
	行人纵向距离	20 m
	雨	0	30%	60%	90%
	云	0	90%	95%	100%
	地面干湿度	0	10%	40%	70%

参数类型	参数名称	参数设置
参数类型	参数名称	无雨场景	小雨场景	中雨场景	大雨场景
目标物参数	行人型号	Male_adult
	行人纵向距离	20 m
	雨	0	10%	25%	100%
	能见度	100%	100%	50%	15%
	地面反射	0	10%	20%	50%

降雨强度	?MR	?CC
小雨	0.000	0.009
中雨	0.000	-0.026
大雨	0.000	0.014

评价指标	DR			CC
降雨等级	小雨	中雨	大雨	小雨	中雨	大雨
场地物理模拟降雨A	0.085	0.008	0.017	0.047	0.008	0.007
场地物理模拟降雨B	0.002	0.003	0.002	0.004	0.004	0.006
仿真数字模拟降雨A	0.002	0.005	0.007	0.014	0.021	0.030
仿真数字模拟降雨B	0.002	0.002	0.002	0.010	0.021	0.030
仿真数字模拟降雨C	0.001	0.001	0.001	0.009	0.008	0.012
生成模型模拟降雨	0.003	0.001	0.003	0.015	0.053	0.030

A Comparative Study on the Machine Vision Realism of Rainfall Simulation Methods

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 31

References 29

Related Articles 4

Metrics

Comments

Recommended 0

调整显著性	P
降雨等级	小雨-中雨	小雨-大雨
场地物理模拟降雨A	0.220	0.000**
场地物理模拟降雨B	0.000**	0.000**
仿真数字模拟降雨A	0.041*	0.000**
仿真数字模拟降雨B	0.000**	0.000**
仿真数字模拟降雨C	0.000**	0.000**
生成模型模拟降雨	0.000**	0.000**

降雨等级		小雨	中雨	大雨
评价指标		?CC	?CC	?CC
真实降雨	25百分位数	-0.016	-0.067	-0.019
真实降雨	75百分位数	0.038	-0.003	0.071
场地物理模拟降雨A		√	↑	↑
场地物理模拟降雨B		↑	↑	√
仿真数字模拟降雨A		√	↑	√
仿真数字模拟降雨B		√	↑	√
仿真数字模拟降雨C		√	↑	↑
生成模型模拟降雨		√	↓	√

[1]	Yunbing Yan,Minghao Yue,Haiwei Li. Estimation of Road Side Slope Based on Fusion of Extended Kalman Filter and Machine Vision [J]. Automotive Engineering, 2024, 46(4): 605-616.
[2]	Yujie Yao,Yuhui Peng,Zehui Chen,Weikun He,Qing Wu,Wei Huang,Wenqiang Chen. An Improved YOLO Algorithm Supporting Anti-illumination Target Detection [J]. Automotive Engineering, 2023, 45(5): 777-785.
[3]	Zhu Yetian, Shen Hong, Zhang Wenjun, Lin Li, Wu Shaoteng, Ao Yunke,Wan Yu, Zhao Zhiqian, Pei Jingyu. Research and Development of Quality Inspection System for Inner Wall ofValve Core Holes of Vehicle Automatic Transmission Valve Body [J]. Automotive Engineering, 2020, 42(4): 484-490.
[4]	Wang Zhangu, Gao Song, Shao Jinju, Tan Derong, Sun Liang & Yu Jie. Front Vehicle Detection with Multi source Information Based on Deep Belief Network [J]. , 2018, 40(5): 554-560.