强光环境智能汽车相机功能仿真测试

doi:10.19562/j.chinasae.qcgc.2025.12.010

摘要/Abstract

摘要：

由车辆远光灯、日光等光源产生的强光环境会对智能汽车相机功能产生较大影响，建立面向强光环境准确且可控的测试方法对提高智能汽车环境适应性具有重要意义。然而，基于实车道路或封闭场地进行强光实景测试存在测试成本高、周期长且场景难以精准可控复现的问题。本文提出一种面向强光环境的智能汽车相机功能仿真测试方法，通过几何-物理融合仿真模型模拟强光环境中的图像响应，并利用强光模拟图像测试智能汽车相机功能。首先，构建强光光源的几何位置和影响区域范围模型；其次，利用相机成像物理模型和强光光源物理特性确定强光影响对应的像素强度，基于强光影响区域和强光像素值将实车采集的无强光图像合成为强光模拟图像；最后，以真实强光环境测试结果作为真值，对比传统仿真软件、大模型方法和本文方法针对强光环境下相同相机功能的测试结果。实验结果表明，本文提出的方法能够提高强光环境仿真测试的性能，相较于传统仿真软件和大模型方法实现了更有效且准确的强光环境相机功能测试，同时相较于物理实景测试具有测试成本低、效率高且仿真模拟参数精准可控的优势。

关键词: 智能汽车, 相机功能测试, 强光环境, 几何-物理融合仿真模型

Abstract:

The strong light environment generated by vehicle high beams， sunlight， and other light sources can significantly affect the camera functions of intelligent vehicle. Establishing accurate and controllable testing methods for such strong light environment is crucial for improving the environmental adaptability of intelligent vehicles. However， real-world testing on roads or closed fields under strong light conditions faces challenges such as high testing cost， long cycles， and the difficulty of precisely replicating controlled scenarios. In this paper a simulation test method for intelligent vehicle camera function under strong light environment is proposed. It uses a geometric-physical fusion simulation model to simulate image responses in strong light environment， and employs strong light simulation images to test camera functions. Firstly， a geometric model of strong light environment imaging is constructed to determine the area affected by strong light. Secondly， the camera imaging physical model and the physical characteristics of strong light sources are used to determine the pixel intensity corresponding to the strong light effect. Based on the strong light impact region and pixel values， images collected from real vehicles under normal light conditions are combined to generate strong light simulation images. Finally， real-world test results under strong light environment are taken as ground truth， and the testing results for the same camera functions in strong light environment are compared with traditional simulation software， large model methods， and the method proposed in this paper. The experimental results show that the proposed method can improve the performance of strong light environment simulation testing. Compared to traditional simulation software and large model methods， it achieves more effective and accurate camera function testing under strong light conditions. Additionally， it offers the advantages of lower testing cost， higher efficiency， and precise controllability of simulation parameters， as compared to physical scene testing.

Key words: intelligent vehicle, camera function testing, strong light environment, geometric-physical fusion simulation model

黄殷梓,朱冰,赵健,张培兴,高质桐,韩嘉懿,范欣炜. 强光环境智能汽车相机功能仿真测试[J]. 汽车工程, 2025, 47(12): 2378-2386.

Yinzi Huang,Bing Zhu,Jian Zhao,Peixing Zhang,Zhitong Gao,Jiayi Han,Xinwei Fan. Simulation Test Method of Intelligent Vehicle Camera Function in Strong Light Environment[J]. Automotive Engineering, 2025, 47(12): 2378-2386.

图/表 20

图1

图2

图3

图4

图5

图6

图7

表1

车辆远光灯和日光物理特性"

物理特性参数	车辆远光灯	日光
$d Φ$	40 W	1 000 W/m³ $× d A$
$θ$ /（°）	10	45
$d w$ / $s r$	$5.48 × 10 - 4$	$9 × 10 - 3$
$d A$ / $m 2$	$1.26 × 10 - 5$	$1.26 × 10 - 5$
$L /$ （ $W ? m - 2 ? s r - 1$ ）	$5.9 × 109$	$1.6 × 106$

表1

表2

模拟相机物理参数及取值"

相机物理参数	数值
$D c$	4 mm
$f$	6.0 mm
$T$	20 ms
$g'$	1
$U$	0 V
$η$	12.89 mV
$S a t$	16 383
$B l a$	0
$g a i n R$	2.433 594
$g a i n B$	1.347 656

表2

图8

图9

图10

图11

图12

图13

图14

图15

图16

表3

图17

参考文献 18

[1]	苗作华，朱良建，赵成诚，等. 基于Retinex的低光照车位图像增强［J］. 汽车工程， 2023， 45（6）： 989-996.
	MIAO Z， ZHU L， ZHAO C， et al. Image enhancement of low-light parking space based on retinex［J］. Automotive Engineering， 2023， 45（6）： 989-996.
[2]	KANG S J， RYU K B， JEONG M S， et al. CAM-FRN： class attention map-based flare removal network in frontal-viewing camera images of vehicles［J］. Mathematics， 2023， 11（17）： 1-45.
[3]	CECCARELLI A， SECCI F. RGB cameras failures and their effects in autonomous driving applications［J］. IEEE Transactions on Dependable and Secure Computing， 2022， 20（4）： 2731-2745.
[4]	BOUDETTE N E. Fatal Tesla crash raises new questions about autopilot system［J］. The New York Times， 2018， 31.
[5]	隗寒冰，曹旭，赖锋.智能汽车环境感知算法测试评价系统开发［J］. 中国机械工程， 2018， 29（19）：2298-2305，2311.
	WEI H， CAO X， LAI F. Development of environmental perception algorithm test evaluation systems for intelligent vehicles［J］. China Mechanical Engineering， 2018， 29（19）：2298-2305，2311.
[6]	肖天昊.基于视觉感知和深度学习的车道保持控制方法研究［D］. 重庆：重庆理工大学， 2024.
[7]	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， LIU Z， YANG X， et al. Evaluation method for the penetration rate of perception system triggering conditions［J］. Automotive Engineering， 2024， 46（4）： 577-587.
[8]	张师源.面向自动驾驶测试的数字孪生系统研究与实现［D］. 西安：长安大学， 2023.
	ZHANG Shiyuan. Research and implementation of a digital twin system for automatic driving testing［D］. Xi’an：Chang’ an University， 2023.
[9]	朱冰，张培兴，赵健，等. 基于场景的自动驾驶汽车虚拟测试研究进展［J］.中国公路学报， 2019， 32（6）：1-19.
	ZHU B， ZHANG P， ZHAO J， et al. Review of scenario-based virtual validation methods for automated vehicles［J］. China Journal of Highway and Transport， 2019， 32（6）： 1-19.
[10]	PUN A， SUN G， WANG J， et al. LightSim： neural lighting simulation for urban scenes ［DB/OL］. （2023-09-11）［2024-08-08］. https：//arxiv.org/abs/2312.06654.
[11]	WU Y， HE Q， XUE T， et al. How to train neural networks for flare removal［C］. Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal， Canada： IEEE， 2021： 2239-2247.
[12]	MA S， AELTERMAN J. Lens flare-aware detector in autonomous driving［C］. VISIGRAPP： VISAPP. Rome， Italy： DBLP， 2024： 341-348.
[13]	DAI Y， LI C， ZHOU S， et al. Flare7K++： mixing synthetic and real datasets for nighttime flare removal and beyond ［DB/OL］. （2023-07-08）［2024-12-02］. https：//arxiv.org/abs/2306.04236.
[14]	NAYAR S K. Image formation ［EB/OL］. （2022-02-01）［2024-08-16］. https：//cave.cs.columbia.edu/Statics/monographs/Image%20Formation%20FPCV-1-1.pdf.
[15]	NAYAR S K. Image sensing ［EB/OL］. （2022-02-01）［2024-08-16］. https：//cave.cs.columbia.edu/Statics/monographs/Image%20Sensing%20FPCV-1-2.pdf.
[16]	SKORKA O， JASINSKI D， ISPASOIU R， et al. Method for quantifying image sensor susceptibility to chromatic flare artifacts［J］. Electronic Imaging， 2016， 28： 1-6.
[17]	LIU Y， ZHANG K， LI Y， et al. Sora： a review on background， technology， limitations， and opportunities of large vision models ［EB/OL］. （2024-04-17）［2024-12-16］. https：//arxiv.org/abs/2402.17177.
[18]	ZHANG Y， GUO Z， WU J， et al. Real-time vehicle detection based on improved YOLO v5［J］. Sustainability， 2022， 14（19）：