Feature Extraction and Region Growing Algorithm for Processing CT Scans of Engine Parts

doi:10.12382/bgxb.2021.0893

Abstract

Abstract:

Aiming at the problem that the influence of metal artifacts and noise in industrial computed tomography (CT) images will interfere with the accuracy and precision of part segmentation extraction, a feature extraction method based on standard deviation weight threshold and region growing is proposed for industrial CT images. A two-dimensional maximum between-class variance and two-dimensional minimum cross-entropy threshold segmentation algorithm based on standard deviation weight is proposed to remove the image background. Automatic selection of various sub-point regions is made based on the neighborhood mean of the image. The extraction of component features is completed based on the Scharr operator to calculate the gradient and improve the growth criterion. Experimental results demonstrate that compared with other region growing methods, our algorithm improves accuracy by 9.1% and achieves a maximum pixel accuracy close to 1. The dice score improves by 5.3% while the intersection over union is improves by 4.1% at maximum. Our feature extraction algorithm outperforms other region growing methods.

Key words: feature extraction, engine CT image, two-dimensional OUST method, two-dimensional minimum cross entropy, standard deviation weight

ZHANG Bin, LU Hongyi, LIU Shun, SANG Doudou, YANG Yucheng. Feature Extraction and Region Growing Algorithm for Processing CT Scans of Engine Parts[J]. Acta Armamentarii, 2023, 44(4): 1171-1180.

Figures/Tables 13

Fig.1 Standard deviation threshold and flowchart of the improved region growing algorithm

Table 1 Results of non-local mean filtering processing of engine conditioner and turbo CT slices

发动机部件	CT切片原始图像	滤波后图像
调节器
发动机涡轮

Fig.2 2D grayscale mean histogram of the grayscale-domain

Fig.3 Improved threshold segmentation

Fig.4 Flowchart of the regional growing method

Fig.5 Flowchart of the automatic selection process of seed points

Fig.6 Digital image neighborhood

Fig.7 Improved regional growing method

Fig.8 Raw CT images of the engine

Table 2 Extraction results of the improved algorithm

发动机部件	调节器	发动机涡轮
CT图像预先分割图
灰度值直方图
特征提取结果

Table 3 Comparison of four algorithms

发动机部件	传统区域生长法	最大熵法	文献[14]算法	本文算法
调节器
发动机涡轮

Fig.9 Design drawings for extracting ideal results

Table 4 Comparison of evaluation index values of different algorithms

算法	发动机部件	C_precision	P_accuracy	D_dice	I_intersection
传统区域生长法	调节器	0.955	0.998	0.957	0.918
	发动机涡轮	0.824	0.995	0.895	0.810
最大熵法	调节器	0.731	0.993	0.845	0.731
	发动机涡轮	0.920	0.998	0.958	0.920
文献[14]算法	调节器	0.899	0.997	0.942	0.890
	发动机涡轮	0.971	0.997	0.935	0.878
本文算法	调节器	0.981	0.998	0.962	0.927
	发动机涡轮	0.996	0.999	0.985	0.970

References 20

[1]	吕宁, 徐更光. 基于工业计算机断层成像的装药底隙无损检测方法研究[J]. 兵工学报, 2015, 36(1):157-162. doi: 10.3969/j.issn.1000-1093.2015.01.023
	LYU N, XU G G. Inspection of base separation of charge by industrial CT imaging[J]. Acta Armamentarii, 2015, 36(1):157-162. (in Chinese) doi: 10.3969/j.issn.1000-1093.2015.01.023
[2]	廖达海, 殷明帅, 罗宏斌, 等. 基于耦合去噪算法的航空发动机中Si3N4圆柱滚子表面缺陷的检测方法[J]. 兵工学报, 2022, 43(1):190-198.
	LIAO D H, YIN M S, LUO H B, et al. Detection and analysis of surface defects of Si3N4 cylindrical roller in aero-engine based on coupled denoising algorithm[J]. Acta Armamentarii, 2022, 43(1):190-198. (in Chinese)
[3]	卢洪义, 陈庆贵, 周红梅, 等. 固体火箭发动机CT图像条状伪影校正[J]. 航空动力学报, 2016, 31(10):2515-2521.
	LU H Y, CHEN Q G, ZHOU H M, et al. Streak artifacts correction of solid rocket motor’s CT image[J]. Journal of Aerospace Power, 2016, 31(10):2515-2521. (in Chinese)
[4]	黄鹏, 郑淇, 梁超. 图像分割方法综述[J]. 武汉大学学报(理学版), 2020, 66(6):519-531.
	HUANG P, ZHENG Q, LIANG C. Overview of image segmentation methods[J]. Journal of Wuhan University (Natural Science Edition), 2020, 66(6):519-531. (in Chinese)
[5]	吴一全, 吴加明, 占必超. 一种可有效分割小目标图像的阈值选取方法[J]. 兵工学报, 2011, 32(4):469-475.
	WU Y Q, WU J M, ZHAN B C. An effective method of threshold selection for small object image[J]. Acta Armamentarii, 2011, 32(4):469-475. (in Chinese)
[6]	朱敏, 卢洪义, 肖志斌, 等. 固体发动机CT图像的一种自动分割方法[J]. 固体火箭技术, 2008(2):201-204.
	ZHU M, LU H Y, XIAO Z B, et al. A kind of automatic segmentation method for solid motor CT image[J]. Journal of Solid Rocket Technology, 2008(2):201-204. (in Chinese)
[7]	于光辉, 卢洪义, 朱敏, 等. 基于引力模型的固体发动机CT图像边缘检测[J]. 核电子学与探测技术, 2012, 32(2):184-187.
	YU G H, LU H Y, ZHU M, et al. Edge detection of solid motor CT images based on gravity model[J]. Nuclear Electronics & Detection Technology, 2012, 32(2): 184-187. (in Chinese)
[8]	郑洲, 张学昌, 郑四鸣, 等. 基于区域增长与统一化水平集的CT肝脏图像分割[J]. 浙江大学学报(工学版), 2018, 52(12):2382-2396.
	ZGENG Z, ZHANG X C, ZHENG S M, et al. Liver segmentation in CT images based on region-growing and unified level set method[J]. Journal of Zhejiang University (Engineering Science), 2018, 52(12):2382-2396. (in Chinese)
[9]	张丽娟, 章润, 李东明, 等. 积神经网络交互分割肝脏CT图像[J]. 液晶与显示, 2021, 36(9):1294-1304.
	ZHANG L J, ZHANG R, LI D M, et al. Region-growing fully convolutional neural network interactive segmentation of liver CT images[J]. Chinese Journal of Liquid Crystals and Displays, 2021, 36(9):1294-1304. (in Chinese) doi: 10.37188/CJLCD.2020-0338 URL
[10]	谢勤岚, 潘先攀. 自适应形状约束Graph cuts算法在腹部CT图像分割中的应用[J]. 中南民族大学学报(自然科学版), 2019, 38(1):119-125.
	XIE Q L, PAN X P. Adaptive shape constraint based graph cuts and its application on CT images segmentation[J]. South-Central Minzu University(Natural Science Edition, 2019, 38(1):119-125. (in Chinese)
[11]	马月红, 孔梦瑶. 基于改进快速区域卷积网络的目标检测轻量化算法[J]. 兵工学报, 2021, 42(12):2664-2674. doi: 10.3969/j.issn.1000-1093.2021.12.014
	MA Y H, KONG M Y. A lightweight target detection algorithm based on the improved faster-RCNN[J]. Acta Armamentarii, 2021, 42(12):2664-2674. (in Chinese) doi: 10.3969/j.issn.1000-1093.2021.12.014
[12]	张玉燕, 李永保, 温银堂, 等. 基于Faster R-卷积神经网络的金属点阵结构缺陷识别方法[J]. 兵工学报, 2019, 40(11):2329-2335. doi: 10.3969/j.issn.1000-1093.2019.11.018
	ZHANG Y Y, LI Y B, WEN Y T, et al. Internal defect detection of metal three-dimensional multi-layer lattice structure based on faster R-CNN[J]. Acta Armamentarii, 2019, 40(11):2329-2335. (in Chinese) doi: 10.3969/j.issn.1000-1093.2019.11.018
[13]	TREMEAU A, BOREL N. A region growing and merging algorithm to color segmentation[J]. Pattern Recognition, 1997, 30(7):1191-1203. doi: 10.1016/S0031-3203(96)00147-1 URL
[14]	彭道刚, 尹磊, 戚尔江, 等. 基于OTSU和区域生长的电厂管道缺陷检测与分割[J]. 红外技术, 2021, 43(5):502-509.
	PENG D G, YIN L, QI E J, et al. Power plant pipeline defect detection and segmentation based on OTSU’s and region growing algorithms[J]. Infrared Technology, 2021, 43(5):502-509. (in Chinese)
[15]	余成波, 曾亮, 张林. 基于OTSU和区域生长的电气设备多点故障分割[J]. 红外技术, 2018, 40(10):1008-1012.
	YU C B, ZENG L, ZHANG L. Multi point fault segmentation of electrical equipment based on OTSU and region growth[J]. Infrared Technology, 2018, 40(10):1008-1012. (in Chinese)
[16]	万俊霞, 林珊玲, 梅婷, 等. 基于图像分割和动态直方图均衡的电润湿显示器图像增强算法[J]. 光子学报, 2022, 51(2):240-250.
	WAN J X, LIN S L, MEI T, et al. Image enhancement algorithm of electrowetting displays based on image segmentation and dynamic histogram equalization[J]. Acta Photonica Sinica, 2022, 51(2):240-250. (in Chinese)
[17]	YANG J, FAN J F, AI D N, et al. Local statistics and non-local mean filter for speckle noise reduction in medical ultrasound image[J]. Neurocomputing, 2016, 195: 88-95. doi: 10.1016/j.neucom.2015.05.140 URL
[18]	刘健庄, 栗文青. 灰度图像的二维 Otsu 自动阈值分割法[J]. 自动化学报, 1993, 19(1): 101-105.
	LIU J Z, LI W Q. The automatic thresholding of gray level pictures via two-dimensional Otsu method[J]. Ata Automatica Sinica, 1993, 19(1):101-105. (in Chinese)
[19]	LI C H, LEE C K. Minimum cross entropy thresholding[J]. Pattern Recognition, 1993, 26(4): 617-625. doi: 10.1016/0031-3203(93)90115-D URL
[20]	雷博, 范九伦. 灰度图像的二维交叉熵阈值分割法[J]. 光子学报, 2009, 38(6):1572-1576.
	LEI B, FAN J L. Two-dimensional cross-entropy thresholding segmentation method[J]. Acta Photonica Sinica, 2009, 38(6):1572-1576. (in Chinese)