快速单曝光高动态范围高反光金属表面缺陷辨识算法研究

doi:10.3969/j.issn.1004-132X.2025.09.016

摘要/Abstract

摘要：

基于细节增强手段以及循环一致性生成对抗网络（CycleGAN）提出了一种快速单曝光高动态范围（HDR）的高反光金属表面缺陷辨识算法。将输入的低动态范围图像转换为HSV颜色空间，经导向滤波处理后得到亮度层和细节层；利用CycleGAN网络对这些层的动态范围分别进行扩展，并将扩展动态范围后的亮度层和细节层进行加权融合，经过滤波去噪后得到利于缺陷辨识的HDR图像；通过阈值分割、特征筛选、形态学处理等手段对HDR图像进行缺陷辨识。将该单曝光算法与经典的3种单曝光算法和1种多曝光算法进行实验对比分析，用峰值信噪比、图像熵、处理时间、灰度直方图、辨识结果5项指标综合评价实验结果。结果表明，该算法相较于其他3种单曝光算法能更好地解决过曝光问题，接近多曝光算法水平，且算法处理时间更短，能满足在线检测的需求，同时该算法挖掘图像细节信息的能力优于其他算法，辨识准确度更高。

关键词: 高反光金属, 单曝光高动态范围成像, 细节增强, 循环一致性生成对抗网络

Abstract:

A rapid single-exposure HDR defect recognition algorithm was proposed for highly reflective metal surfaces. This algorithm was based on detail enhancement techniques and CycleGAN. The input low dynamic range （LDR） images were first converted to HSV color space and processed with guided filtering to obtain luminance and detail layers. The CycleGAN network was then used to enhance the dynamic range of these layers separately. The enhanced luminance and detail layers were weighted and fused， followed by filtering and denoising to produce an HDR image suitable for defect recognition. Defects were identified in the HDR image using threshold segmentation， feature selection， and morphological processing. This single-exposure algorithm was experimentally compared with three classic single-exposure algorithms and one multi-exposure algorithm. The evaluation was based on five metrics： peak signal-to-noise ratio（PSNR）， image entropy， processing time， gray histograms， and recognition results. The experimental results indicate that the algorithm herein outperforms three other single-exposure algorithms in effectively addressing overexposure issues， achieving results comparable to multi-exposure algorithms. Additionally， it has a shorter processing time， making it suitable for online detection. Furthermore， this algorithm demonstrates superior capability in extracting image detail information compared to other algorithms， resulting in higher accuracy in recognition.

Key words: highly reflective metal, single-exposure high dynamic range（HDR） imaging, detail enhancement, cycle-consistent generative adversarial network （CycleGAN）

中图分类号:

TP391

贾维昊, 王鹏, 陈凯, 仝飞, 王国彪. 快速单曝光高动态范围高反光金属表面缺陷辨识算法研究[J]. 中国机械工程, 2025, 36(09): 2039-2046.

Weihao JIA, Peng WANG, Kai CHEN, Fei TONG, Guobiao WANG. Research on Rapid Single-exposure HDR Defect Recognition Algorithm for Highly Reflective Metal Surfaces[J]. China Mechanical Engineering, 2025, 36(09): 2039-2046.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.cmemo.org.cn/CN/10.3969/j.issn.1004-132X.2025.09.016

https://www.cmemo.org.cn/CN/Y2025/V36/I09/2039

图/表 10

图1 快速单曝光HDR表面缺陷辨识算法技术路线

Fig.1 Technical roadmap for rapid single exposure HDR surface defect recognition algorithm

图2 分离细节层

Fig.2 Separate detail layer

图3 提高动态范围

Fig.3 Enhance dynamic range

图4 缺陷辨识过程

Fig.4 Defect identify process

图5 各算法成像效果

Fig.5 Imaging effects of various algorithms

图6 PSNR对比结果

Fig.6 Comparison of PSNR

图7 图像熵对比结果

Fig.7 Comparison of image entropy

图8 处理时间对比结果

Fig.8 Comparison of processing time

图9 图像的灰度直方图

Fig.9 Gray histogram of image

图10 辨识结果

Fig.10 Identification results

参考文献 23

[1]	NEOGI N， MOHANTA D K， DUTTA P K. Defect Detection of Steel Surfaces with Global Adaptive Percentile Thresholding of Gradient Image［J］. Journal of the Institution of Engineers （India）：Series B， 2017， 98（6）：557-565.
[2]	GAIDHANE V H， HOTE Y V， SINGH V. An Efficient Similarity Measure Approach for PCB Surface Defect Detection［J］. Pattern Analysis and Applications， 2018， 21（1）：277-289.
[3]	TRUONG M T N， KIM S. Automatic Image Thresholding Using Otsu's Method and Entropy Weighting Scheme for Surface Defect Detection［J］. Soft Computing， 2018， 22（13）：4197-4203.
[4]	ASLAM Y， SANTHI N， RAMASAMY N， et al. RETRACTED ARTICLE：Localization and Segmentation of Metal Cracks Using Deep Learning［J］. Journal of Ambient Intelligence and Humanized Computing， 2021， 12（3）：4205-4213.
[5]	汤勃，孔建益，伍世虔. 机器视觉表面缺陷检测综述［J］. 中国图象图形学报， 2017， 22（12）：1640-1663.
	TANG Bo， KONG Jianyi， WU Shiqian. Review of Surface Defect Detection Based on Machine Vision［J］. Journal of Image and Graphics， 2017， 22（12）：1640-1663.
[6]	闵永智，岳彪，马宏锋，等. 基于图像灰度梯度特征的钢轨表面缺陷检测［J］. 仪器仪表学报， 2018， 39（4）：220-229.
	MIN Yongzhi， YUE Biao， MA Hongfeng， et al. Rail Surface Defects Detection Based on Gray Scale Gradient Characteristics of Image［J］. Chinese Journal of Scientific Instrument， 2018， 39（4）：220-229.
[7]	王柳，陈超，高楠，等. 基于自适应条纹投影的高反光物体三维面形测量［J］. 应用光学， 2018， 39（3）：373-378.
	WANG Liu， CHEN Chao， GAO Nan， et al. Three-dimensional Shape Measurement of High Reflective Objects Based on Adaptive Fringe-pattern Projection［J］. Journal of Applied Optics， 2018， 39（3）：373-378.
[8]	刘婷婷，王培光，张娜. 基于Zernike矩亚像素的高反光金属工件缺陷检测［J］. 激光与光电子学进展， 2019， 56（12）：121203.
	LIU Tingting， WANG Peiguang， ZHANG Na. Subpixel Defect Detection in Highly Reflective Workpieces Based on Zernike Moments［J］. Laser & Optoelectronics Progress， 2019， 56（12）：121203.
[9]	陆雅诺，陈炳才，陈德刚，等. 一种基于注意力模型的带钢表面缺陷识别算法［J］. 激光与光电子学进展， 2021， 58（14）：242-250.
	LU Yanuo， CHEN Bingcai， CHEN Degang， et al. Recognition Algorithm of Strip Steel Surface Defects Based on Attention Model［J］. Laser & Optoelectronics Progress， 2021， 58（14）：242-250.
[10]	李云飞. 基于改进K-means算法的工件表面缺陷分割算法研究［J］. 湖南工程学院学报（自然科学版）， 2021， 31（1）：53-58.
	LI Yunfei. Research on Workpiece Surface Defect Segmentation Algorithm Based on Improved K-means Algorithm［J］. Journal of Hunan Institute of Engineering （Natural Science Edition）， 2021， 31（1）：53-58.
[11]	谢舰，姚剑敏，严群，等. 基于深度学习的磁瓦表面缺陷分割与识别［J］. 液晶与显示， 2021， 36（5）：713-722.
	XIE Jian， YAO Jianmin， YAN Qun， et al. Segmentation and Recognition of Magnetic Tile Surface Defects Based on Deep Learning［J］. Chinese Journal of Liquid Crystals and Displays， 2021， 36（5）：713-722.
[12]	陈卓明. 基于多尺度分析的海洋石油机械零件表面缺陷分割［J］. 中国新技术新产品， 2022（24）：63-65.
	CHEN Zhuoming. Segmentation of Surface Defects of Offshore Oil Machinery Parts Based on Multi-scale Analysis［J］. New Technology & New Products of China， 2022（24）：63-65.
[13]	冯维，刘红帝，汤少靖，等. 基于HDRI的高反光金属表面缺陷检测方法研究［J］. 仪表技术与传感器， 2019（8）：112-116.
	FENG Wei， LIU Hongdi， TANG Shaojing， et al. Research on Defect Detection Method for High-reflective-metal Surface Based on HDRI［J］. Instrument Technique and Sensor， 2019（8）：112-116.
[14]	甘富升. 基于机器视觉的高反光金属表面缺陷检测与分类方法研究［D］. 成都：电子科技大学， 2023.
	GAN Fusheng. Research on High Reflective Metal Surface Defect Detection and Classification Method Based on Machine Vision［D］. Chengdu：University of Electronic Science and Technology of China， 2023.
[15]	张红英，朱恩弘，吴亚东. 一种基于细节层分离的单曝光HDR图像生成算法［J］. 自动化学报， 2019， 45（11）：2159-2170.
	ZHANG Hongying， ZHU Enhong， WU Yadong. High Dynamic Range Image Generating Algorithm Based on Detail Layer Separation of a Single Exposure Image［J］. Acta Automatica Sinica， 2019， 45（11）：2159-2170.
[16]	HE Kaiming， SUN Jian， TANG Xiaoou. Guided Image Filtering［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2013， 35（6）：1397-1409.
[17]	ZHU Junyan， 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）. Venice， 2017：2242-2251.
[18]	MERTENS T， KAUTZ J， van REETH F. Exposure Fusion：a Simple and Practical Alternative to High Dynamic Range Photography［J］. Computer Graphics Forum， 2009， 28（1）：161-171.
[19]	LEE C H， CHEN L H， WANG Weikang. Image Contrast Enhancement Using Classified Virtual Exposure Image Fusion［J］. IEEE Transactions on Consumer Electronics， 2012， 58（4）：1253-1261.
[20]	WANG T H， CHIU C W， WU Weichen， et al. Pseudo-multiple-exposure-based Tone Fusion with Local Region Adjustment［J］. IEEE Transactions on Multimedia， 2015， 17（4）：470-484.
[21]	PARK J S， SOH J W， CHO N I. Generation of High Dynamic Range Illumination from a Single Image for the Enhancement of Undesirably Illuminated Images［J］. Multimedia Tools and Applications， 2019， 78（14）：20263-20283.
[22]	王汇源. 数字图像通信原理与技术［M］. 北京：国防工业出版社， 2000.
	WANG Huiyuan. Principle and Technology of Digital Image Communication［M］. Beijing：National Defense Industry Press， 2000.
[23]	SHANNON C E. A Mathematical Theory of Communication［J］. Bell System Technical Journal， 1948， 27（3）：379-423.