Micro- and nano-scale feature extraction method for aviation silicon nitride turbine blades based on adaptive Perona-Malik enhancement and multi-scale Canny segmentation
-
摘要:
针对航空氮化硅涡轮叶片微纳尺度（10 nm～10 μm）特征图像中存在的噪声密集、边缘模糊与丢失问题，提出一种基于自适应Perona-Malik增强和变尺度Canny分割的耦合方法，实现航空氮化硅涡轮叶片微纳尺度特征的高精度、低损失提取。通过分析特征图像的梯度分布与噪声特性，构建基于梯度中位数与第90百分位数的扩散系数自适应机制，设计变尺度金字塔分层策略，在不同尺度下分别进行非极大值抑制与双阈值分割，最终通过加权融合还原至原尺度，实现多尺度边缘的综合提取与细化。增强后的图像结构相似性指数（SSIM）高达
0.9706 ，分割后的图像交并比（IoU）达到0.9369 ，有效改善了噪声干扰与边缘丢失导致的特征提取不完整问题，显著提升了航空氮化硅涡轮叶片微纳尺度特征的表征精度与缺陷分析能力，为叶片微纳缺陷精准识别与服役安全保障提供了可靠支撑。-
关键词:
- 航空氮化硅涡轮叶片 /
- 微纳尺度特征 /
- 自适应Perona-Malik增强 /
- 变尺度Canny分割 /
- 提取方法
Abstract:To address the issues of dense noise, blurred edges, and feature loss in images of aviation silicon nitride turbine blades with micro- and nano-scale features (10 nm—10 μm), a coupled method based on adaptive Perona-Malik enhancement and multi-scale Canny segmentation was proposed. This approach enabled high-precision, low-loss extraction of aviation silicon nitride turbine blades with micro- and nano- scale features. By analyzing the gradient distribution and noise characteristics of feature images, an adaptive mechanism based on the median gradient and 90th percentile diffusion coefficient was constructed. A multi-scale pyramidal hierarchical strategy was designed to perform non-maximum suppression and dual-threshold segmentation at different scales. Finally, through weighted fusion, the results were restored to the original scale, achieving comprehensive extraction and refinement of multi-scale edges. The enhanced image structure similarity index (SSIM) reached
0.9706 , while the intersection-over-union (IoU) of the segmented images achieved 0.9369. This effectively mitigated the issue of incomplete feature extraction caused by noise interference and edge loss, and significantly improved the characterization accuracy and defect analysis capability of aviation silicon nitride turbine blades with micro- and nano-scale features, thereby providing reliable support for the accurate identification of micro- and nano-scale defects in the blades and the guarantee of their service safety. -
图 1 航空氮化硅涡轮叶片微纳尺度特征图像采集平台设计
Figure 1. Design of an image acquisition platform for micro- and nano-scale features of aviation silicon nitride turbine blades
图 2 航空氮化硅涡轮叶片微纳尺度特征分析
Figure 2. Analysis of micro- and nano-scale features in aviation silicon nitride turbine blades
图 3 基于自适应Perona-Malik增强和变尺度Canny分割方法总流程
Figure 3. Overall workflow based on adaptive Perona-Malik enhancement and multi-scale Canny segmentation method
图 4 自适应Perona-Malik增强创新设计图
Figure 4. Innovative design diagram of adaptive Perona-Malik enhancement
图 5 变尺度Canny分割创新设计图
Figure 5. Innovative design diagram of multi-scale Canny segmentation
图 6 自适应Perona-Malik增强局部效果分析图
Figure 6. Local performance analysis diagram of adaptive Perona-Malik enhancement
图 7 自适应Perona-Malik增强整体效果分析图
Figure 7. Overall performance analysis diagram of adaptive Perona-Malik enhancement
图 8 变尺度Canny分割局部效果分析图
Figure 8. Local performance analysis diagram of multi-scale Canny segmentation
图 9 变尺度Canny分割整体效果分析图
Figure 9. Overall performance analysis diagram of multi-scale Canny segmentation
图 10 航空氮化硅涡轮叶片微纳尺度特征图像不同方法特征分离效果图
Figure 10. Feature separation performance comparison diagram of different methods for aviation silicon nitride micro- and nano-scale characterization images
表 1 硬件配置及图像采集参数
Table 1. Hardware configuration and image acquisition parameters
参数 数值或说明 显微镜型号 EM-30AXN 真空/加速电压/kV 1 放大倍率 5000 图像格式 PNG 
下载: 导出CSV
表 2 实验分析相关参数及其数值
Table 2. Experimental analysis parameters and their values
参数 数值或说明 图像分辨率 1280 像素×856像素配置环境 64 bit Windows11 CPU i5-13500H 机带RAM/GB 16.0 运行环境 MATLAB R2024b
下载: 导出CSV
表 3 不同梯度百分位数的增强效果对比
Table 3. Enhancement effect comparison of different gradient percentiles
评价指标 梯度百分位数 70 80 90 95 99 结构相似性指数
（SSIM）0.9215 0.9487 0.9706 0.9613 0.9372 直径一致性误差
（DCE）0.0136 0.0058 0.0022 0.0035 0.0089
下载: 导出CSV
表 4 方法相关指标展示
Table 4. Presentation of method-related metrics
方法 增强指标 分割指标 计算耗时/
ms结构相似性指数
（SSIM）直径一致性误差
（DCE）交并比
（IoU）戴斯相似系数
（Dice）AMF耦合本文分割 0.8065 0.4537 0.6695 0.8020 79.3 本文增强耦合WMS 0.9706 0.0022 0.5953 0.7463 73.5 本文方法 0.9706 0.0022 0.9369 0.9674 68.4
下载: 导出CSV
-
[1] WU Yuhou, GUO Jiancheng, ZHANG Xiaochen, et al. Research on vibration characteristics of silicon nitride 6206 full-ceramic bearing with different pre-tightening force and oil supply rate[J]. The International Journal of Advanced Manufacturing Technology, 2023, 127(9): 4943-4957. [2] NAZIR M H, AHMAD KHAN Z, SAEED A. Experimental analysis and modelling of c-crack propagation in silicon nitride ball bearing element under rolling contact fatigue[J]. Tribology International, 2018, 126: 386-401. doi: 10.1016/j.triboint.2018.04.030 [3] WANG Xiaolong, LI Songhua, FAN Shengqi, et al. Study on properties and tribological behaviors of silicon nitride-based Ti-DLC films[J]. Surface and Coatings Technology, 2025, 506: 132131. doi: 10.1016/j.surfcoat.2025.132131 [4] MCGARRITY K, TUMURUGOTI P, NING Kaijie, et al. Fractography of silicon nitride based ceramics to guide process improvements[J]. Journal of the European Ceramic Society, 2020, 40(14): 4746-4752. doi: 10.1016/j.jeurceramsoc.2020.02.017 [5] OHJI T, TATAMI J. Processing-structure-microscale properties of silicon nitride[J]. Ceramics International, 2024, 50(19): 37282-37290. doi: 10.1016/j.ceramint.2024.04.238 [6] HEGEDÜS N, BALÁZSI K, BALÁZSI C. Silicon nitride and hydrogenated silicon nitride thin films: a review of fabrication methods and applications[J]. Materials, 2021, 14(19): 5658. doi: 10.3390/ma14195658 [7] WEI Wanxin, LIANG Hanqin, SU Yunfeng, et al. Fretting damage behaviors of silicon nitride balls with different sintering aids and processes under grease lubrication[J]. Tribology International, 2022, 171: 107572. doi: 10.1016/j.triboint.2022.107572 [8] ZOU Wanyan, TONG Yuanjian, WANG Yu, et al. Characteristics of compressive failure behavior of polyacrylonitrile-based carbon fiber multifilament[J]. Polymer Composites, 2024, 45(1): 924-932. doi: 10.1002/pc.27826 [9] LEÓN-BECERRA J, HIDALGO-SALAZAR M Á, GONZÁLEZ-ESTRADA O A. Progressive damage analysis of carbon fiber-reinforced additive manufacturing composites[J]. The International Journal of Advanced Manufacturing Technology, 2023, 126(5): 2617-2631. doi: 10.1007/s00170-023-11256-w [10] YE C C, RU H Q, CHEN D L. Fatigue behavior of silicon nitride ceramics[J]. Ceramics International, 2023, 49(17): 28405-28414. doi: 10.1016/j.ceramint.2023.06.095 [11] JIANG Yi, HU Kun, ZHANG Xin, et al. A saturation channel detection method for surface defects of silicon nitride bearing rollers based on adaptive gamma correction-edge threshold segmentation coupling algorithm[J]. Materials Today Communications, 2023, 36: 106397. doi: 10.1016/j.mtcomm.2023.106397 [12] JIA Haiyuan, LIN Bin, LIU Zaiwei, et al. Laser-excited surface acoustic wave method for detecting subsurface damage of processed silicon nitride ceramics[J]. Ceramics International, 2024, 50(21): 42081-42091. doi: 10.1016/j.ceramint.2024.08.051 [13] PU Yezhuang, ZHAO Yugang, ZHAO Guoyong, et al. Experimental study on laser assisted machining of silicon nitride ceramics based on acoustic emission detection[J]. Optics & Laser Technology, 2025, 180: 111497. doi: 10.1016/j.optlastec.2024.111497 [14] HIGA E, SOMETANI M, HIRAI H, et al. Electrically detected magnetic resonance study on interface defects at nitrided Si-face, a-face, and m-face 4H-SiC/SiO2 interfaces[J]. Applied Physics Letters, 2020, 116(17): 171602. doi: 10.1063/5.0002944 [15] ZWANENBURG E A, NORMAN D G, QIAN C, et al. Effective X-ray micro computed tomography imaging of carbon fibre composites[J]. Composites Part B: Engineering, 2023, 258: 110707. doi: 10.1016/j.compositesb.2023.110707 [16] MA Mengyuan, WANG Zhongxin, GAO Zhihao, et al. Ultrasonic phased array testing and identification of multiple-type internal defects in carbon fiber reinforced plastics based on convolutional neural network[J]. Materials, 2025, 18(2): 318. doi: 10.3390/ma18020318 [17] RASHIDI A, OLFATBAKHSH T, CRAWFORD B, et al. A review of current challenges and case study toward optimizing micro-computed X-ray tomography of carbon fabric composites[J]. Materials, 2020, 13(16): 3606. doi: 10.3390/ma13163606 [18] YU Dongling, ZHANG Huiling, ZHANG Xiaohui, et al. Si3N4 ceramic ball surface defects’ detection based on SWT and nonlinear enhancement[J]. Mathematical Problems in Engineering, 2021, 2021(1): 4922315. doi: 10.1155/2021/4922315 [19] YU Dongling, ZHU Zuoxiang, MIN Jianliang, et al. Multi-scale decomposition enhancement algorithm for surface defect images of Si3N4 ceramic bearing balls based on stationary wavelet transform[J]. Advances in Applied Ceramics, 2021, 120(1): 47-57. doi: 10.1080/17436753.2020.1858010 [20] JIANG Yi, TANG Mengtao, DONG Wenjie, et al. Microcrack feature extraction method for chaotic optical surface of silicon nitride ceramic bearing roller based on multi-scale wavelet transform enhancement and optimized PSO-FCM coupling[J]. AIP Advances, 2025, 15: 015214. doi: 10.1063/5.0244948 [21] WANG Nannan, ZHANG Yongxia. Adaptive and fast image superpixel segmentation approach[J]. Image and Vision Computing, 2021, 116: 104315. doi: 10.1016/j.imavis.2021.104315 [22] LEI Tao, JIA Xiaohong, LIU Tongliang, et al. Adaptive morphological reconstruction for seeded image segmentation[J]. IEEE Transactions on Image Processing, 2019, 28(11): 5510-5523. doi: 10.1109/TIP.2019.2920514 [23] HUANG J T, TING C H. Deep learning object detection applied to defect recognition of memory modules[J]. The International Journal of Advanced Manufacturing Technology, 2022, 121(11): 8433-8445. doi: 10.1007/s00170-022-09716-w [24] CHO C, LEE Y H, PARK J, et al. A self-spatial adaptive weighting based U-Net for image segmentation[J]. Electronics, 2021, 10(3): 348. doi: 10.3390/electronics10030348 [25] CHEN Liping, GAO Jinhui, LOPES A M, et al. Adaptive fractional-order genetic-particle swarm optimization Otsu algorithm for image segmentation[J]. Applied Intelligence, 2023, 53(22): 26949-26966. doi: 10.1007/s10489-023-04969-8 -
点击查看大图
计量
- 文章访问数: 244
- HTML浏览量: 90
- PDF量: 33
- 被引次数: 0