Icing type classification via Near-infrared spectral irradiance and deep convolutional neural networks

WU Jun; LI Jiahao; LIU Guihang; CHEN Jiusheng; GUO Runxia

doi:null

您当前的位置：

首页 >

文章列表页 >

Icing type classification via Near-infrared spectral irradiance and deep convolutional neural networks

Optical Sensing and Measurement | 更新时间：2026-01-22

- Icing type classification via Near-infrared spectral irradiance and deep convolutional neural networks
- OPTICAL TECHNIQUE Vol. 52, Issue 1, Pages: 101-108(2026)
- 作者机构：
  
  1.中国民航大学航空工程学院，天津 300300
  2.中国民航大学电子信息与自动化学院，天津 300300
  3.中国民航大学工程技术训练中心，天津 300300
- 作者简介：
- 基金信息：
- DOI：
  CLC： TP391.41;TP183
- Received：11 July 2025，
  
  Revised：2025-12-27，
  
  Published：20 January 2026
- 稿件说明：
移动端阅览
吴军,李嘉昊,刘贵行等.基于近红外光谱辐照度和卷积神经网络的积冰分类方法[J].光学技术,2026,52(01):101-108.

WU Jun,LI Jiahao,LIU Guihang,et al.Icing type classification via Near-infrared spectral irradiance and deep convolutional neural networks[J].OPTICAL TECHNIQUE,2026,52(01):101-108.
吴军,李嘉昊,刘贵行等.基于近红外光谱辐照度和卷积神经网络的积冰分类方法[J].光学技术,2026,52(01):101-108. DOI：

WU Jun,LI Jiahao,LIU Guihang,et al.Icing type classification via Near-infrared spectral irradiance and deep convolutional neural networks[J].OPTICAL TECHNIQUE,2026,52(01):101-108. DOI：

摘要

飞机在运行过程中，特别是在寒冷的冬季和高湿度的环境条件下，在飞机起飞、降落等关键飞行阶段，飞机蒙皮表面积冰问题带来的危害尤为突出。提出了一种融合卷积神经网络（CNN）与近红外光谱分析的检测方法，用于实现对飞机蒙皮表面积冰类型的自动识别。构建了该方法的理论模型与实验方案，并提出了一种结合串联卷积模块与多尺度并联卷积路径的网络模型：IATNet模型。IATNet模型由1个一维卷积层（Conv0）、1个多尺度光谱特征编码模块（MSFE Module）、1个特征融合模块（FeatureFusion）、1个展平层（Flatten）和3个全连接层（D1、D2、D3）组成。经实验验证，IATNet模型在测试集上的识别准确率为98.15%，对比其他传统机器学习算法，识别准确率有显著提高。实验结果表明，检测方法可以精准识飞机蒙皮表面积冰类型，解决了传统飞机蒙皮积冰类型检测方案检测效率低、实时性差或易受环境干扰等问题。

Abstract

During aircraft operations

especially under cold winter conditions and high humidity environments

the hazards posed by surface icing on the aircraft skin are particularly prominent during critical flight phases such as takeoff and landing. In this context

a detection method integrating Convolutional Neural Networks （CNN） with near-infrared spectral analysis is proposed to enable automatic identification of aircraft skin icing types. The theoretical framework and experimental protocol for this approach are established

and a novel network architecture—termed IATNet—is introduced

which combines serial convolutional modules with multi-scale parallel convolutional paths. The IATNet model is composed of one one-dimensional convolutional layer （Conv0）

a multi-scale spectral feature encoding module （MSFE Module）

a feature fusion module （FeatureFusion）

a flattening layer （Flatten）

and three fully connected layers （D1

D3）. Experimental validation demonstrates that the IATNet model achieves an identification accuracy of 98.15% on the test set

which represents a significant improvement over other conventional machine learning algorithms. The experimental results indicate that this detection approach can accurately identify the types of icing on aircraft skin

effectively addressing the limitations of traditional detection methods

such as low efficiency

poor real-time performance

and susceptibility to environmental interference.

关键词

Keywords

references

周莉 , 徐浩军 , 龚胜科 , 等 . 飞机结冰特性及防除冰技术研究 [J ] . 中国安全科学学报 , 2010 , 20 （ 06 ）: 105 — 110 .

Zhou Li , Xu Haojun , Gong Shengke , et al . Study on aircraft icing characteristics and anti-/de-icing technology [J ] . Journal of Safety Science and Technology , 2010 , 20 （ 06 ）: 105 — 110 .

常士楠 . 大型飞机的防/除冰问题 [C ] //中国航空学会,中国工程院机械与运载工程学部. 大型飞机关键技术高层论坛暨中国航空学会2007年学术年会论文集 . 北京航空航天大学航空科学与工程学院 2007 : 469 — 475 .

Chang Shinan . Anti-/de-icing issues of large aircraft [C ] // Chinese Society of Aeronautics and Astronautics, Division of Mechanical and Vehicle Engineering of Chinese Academy of Engineering . Proceedings of the High-Level Forum on Key Technologies of Large Aircraft and 2007 Annual Conference of Chinese Society of Aeronautics and Astronautics . School of Aeronautical Science and Engineering, Beihang University , 2007 : 469 — 475 .

姜志国 , 韩冬兵 , 袁天云 , 等 . 基于全自动控制显微镜的自动聚焦算法研究 [J ] . 中国图象图形学报： 2004 , 9 （ 4 ）: 6 .

Jiang Zhiguo , Han Dongbing , Yuan Tianyun , et al . Study on auto focusing algorithm for automatic microscope [J ] . Journal of Image and Graphics , 2004 , 9 （ 4 ）: 396 — 401 .

Terrace S M , Bender K D , Sierra E A , et al . Comparison of human ice detection capabilities and ground ice detection system performance under post deicing conditions [J ] . Human Factors and Ergonomics Society Annual Meeting Proceedings , 2006 , 50 （ 17 ）: 2051 — 2055 .

李薇 , 陈迎春 , 张淼 , 等 . 基于光波导技术的新型飞机结冰探测系统 [J ] . 仪表技术与传感器 , 2009 ,（ 12 ）: 85 — 87 .

Li Wei , Chen Yingchun , Zhang Miao , et al . A novel aircraft icing detection system based on optical waveguide technology [J ] . Instrument Technique and Sensor , 2009 , （ 12 ）: 85 — 87 .

熊平 . 光纤式结冰传感器气象结冰探测应用研究 [D ] . 武汉 : 华中科技大学 2011 .

Xiong Ping . Research on application of fiber-optic icing sensor in meteorological icing detection [D ] . Wuhan : Huazhong University of Science and Technology , 2011 .

龚晓亮 . 结冰风洞中冰形生长的光学三维测量研究 [D ] . 西安 : 西北工业大学 , 2015 .

Gong Xiaoliang . Optical 3D measurement of ice shape growth in icing wind tunnel [D ] . Xi'anNorthwestern Polytechnical University , 2015 .

谭舒亚 , 葛俊锋 , 叶林 , 等 . 非接触式路面结冰探测技术研究进展 [J ] . 公路 , 2015 , 60 （ 06 ）: 244 — 249 .

Tan Shuya , Ge Junfeng , Ye Lin , et al . Research progress on non-contact road surface icing detection technology [J ] . Highway , 2015 , 60 （ 06 ）: 244 — 249 .

窦慧 , 张凌茗 , 韩峰 , 等 . 卷积神经网络的可解释性研究综述 [J ] . 软件学报 , 2024 （ 001 ）: 035 .

Dou Hui , Zhang Lingming , Han Feng , et al . A review of the interpretability of convolutional neural networks [J ] . Journal of Software , 2024 （ 001 ）: 035 .

陈国喜 , 刘忆森 , 周松斌 , 等 . 基于多分支空洞卷积网络的光谱定量分析 [J ] . 激光与光电子学进展 , 2023 , 6 （ 09 ）: 0930001 .

Chen Guoxi , Liu Yisen , Zhou Songbin , et al . Spectral quantitative analysis based on multi-branch dilated convolutional network [J ] . Progress in Laser and Optoelectronics , 2023 , 60 （ 9 ）: 0930001 .

朱御康 , 鲁昌华 , 张玉钧 , 等 . 基于波长注意力的多特征融合卷积神经网络的近红外光谱定量方法 [J ] . 光谱学与光谱分析 , 2024 , 44 （ 9 ）: 2607 — 2612 .

Zhu Yukang , Lu Changhua , Zhang Yujun , et al . Near-infrared spectral quantitative method based on wavelength attention multi-feature fusion convolutional neural network [J ] . Spectroscopy and Spectral Analysis , 2024 , 44 （ 9 ）: 2607 — 2612 .

刘兰军 , 翟永庆 , 郑俊俊 , 等 . 基于PSO-CNN的土壤氮含量可见/近红外光谱建模 [J ] . 光学技术 , 2021 , 47 （ 04 ）: 438 — 445 .

Liu Lanjun , Zhai Yongqing , Zheng Junjun , et al . Visible/near-infrared spectral modeling of soil nitrogen content based on PSO-CNN [J ] . Optical Technique , 2021 , 47 （ 04 ）: 438 — 445 .

焦青亮 , 刘明 , 于坤 , 等 . 基于卷积神经网络的光谱预处理方法 [J ] . 光谱学与光谱分析 , 2022 , 42 （ 1 ）: 6 .

Jiao Qingliang , Liu Ming , Yu Kun , et al . Spectral preprocessing method based on convolutional neural network [J ] . Spectroscopy and Spectral Analysis , 2022 , 42 （ 1 ）: 6 .

郑运 , 杨思雨 , 王涛 , 等 . 基于数据增强策略和卷积神经网络的近红外光谱分析研究 [J ] . 分析化学 , 2024 , 000 （ 9 ）: 11 .

Zheng Yun , Yang Siyu , Wang Tao , et al . Research on near-infrared spectral analysis based on data augmentation strategy and convolutional neural network [J ] . Analytical Chemistry , 2024 , 000 （ 9 ）: 11 .

王君 , 陈敏 , 董明利 , 等 . 多尺度特征融合的目标检测算法研究 [J ] . 光学技术 , 2022 , 48 （ 06 ）: 749 — 754 .

Wang Jun , Chen Min , Dong Mingli , et al . Research on object detection algorithm based on multi-scale feature fusion [J ] . Optical Technique , 2022 , 48 （ 06 ）: 749 — 754 .

傅鹏有 , 文岳 , 张雨柯 , 等 . 面向近红外光谱定量分析的深度学习建模与模型迁移 [J ] . 光谱学与光谱分析 , 2023 , 43 （ 1 ）: 10 .

Fu Pengyou , Wen Yue , Zhang Yuke , et al . Deep learning modeling and model transfer for near-infrared spectral quantitative analysis [J ] . Spectroscopy and Spectral Analysis , 2023 , 43 （ 1 ）: 10 .

王凡 . 融合CNN和Transformer的高分辨率遥感影像滑坡检测方法 [J ] . 遥感信息 , 2025 , 40 （ 05 ）: 158 — 165 .

Wang Fan . A high-resolution remote sensing image landslide detection method integrating CNN and Transformer [J ] . Remote Sensing Information , 2025 , 40 （ 5 ）: 158 — 165 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

OAM mode recognition system of vortex beams based on deep learning and FPGA

Improved neural network based demodulation of orbital angular momentum fiber-optic communication system

Research on temperature field measurement using Background-oriented schlieren based on hybrid Cross-correlation optical flow algorithm

Optimization and application of microspectroscopic fitting method in micro-region thin film characterization

Nonlinear impairment compensation method for Long-haul Fiber-optic communication system

Related Author

HUANG Dingjin

SHAO Ruikang

WU Tongqiao

LIU Yike

WANG Kunpeng

LV Hong

ZHANG Chen

MA Xuelian

Related Institution

College of Photoelectric Engineering, Xi 'an University of Technology, Xi 'an

School of Computer and Communication Engineering, Northeastern University

Division of Fundamental Education and Research, Guilin Institute of Information Technology

College of Mechanical and Electrical Engineering, Guilin Institute of Information Technology

College of Mechanical and Electrical Engineering, Guilin University of Electronic Technology

⁰