Improved UNet++ for Tree Rings Segmentation of Chinese Fir CT Images

doi:10.3778/j.issn.1002-8331.2210-0212

Abstract

Abstract: In order to solve the problem that it is difficult to accurately segment tree rings with defects such as cracks, wormholes and knots. The medical CT is used as experimental equipment to reconstruct 125 CT images of Chinese fir transverse sections, and these images are used as the data set. Data set is expanded by pre-processing such as cutting, rotating and flipping CT images. An improved UNet++ model is proposed for tree rings segmentation. Convolutional blocks, downsampling layers, skip connections and upsampling layers have been added to the improved UNet++ model, and the learning depth is increased to 6 layers. The BCEWithLogitsLoss, ReLU and RMSProp are used as loss function, activation function and optimization function respectively. The improved UNet++ model is used to segment the tree rings of the transverse sections of Chinese fir reconstructed by CT, and the performance of the model is evaluated. The results show that the pixel accuracy of the improved UNet++ model is 97.81%, the dice coefficient is 98.89%, the intersection over union is 95.29%, and the mean intersection over union is 84.75%. The best segmentation result is obtained by fully extracting the characteristics in Chinese fir tree rings. Compared with the U-Net model and the UNet++ model, the improved UNet++ model makes the segmented tree rings complete and continuous, although most tree rings are cut by cracks and wormholes and cannot form a complete circular closed curve, fracture and noise are eliminated. The results show that the improved UNet++ model is not affected by defects such as cracks, knots and wormholes, and the segmentation results are very clear, which effectively solves mis-segmentation and under segmentation of dense tree rings under the interference of wormhole defects.

Key words: Chinese fir, transverse section, tree rings segmentation, CT image, UNet++ model

摘要： 为解决裂纹、虫孔和节子等缺陷影响下的年轮准确分割问题。以医疗CT为实验设备，重构125张杉木横切面CT图像为研究对象，经裁切、旋转、翻转等预处理实现数据扩充，提出改进UNet++模型用于年轮分割。改进UNet++模型采用增加卷积块、下采样层、跳跃连接和上采样层的方式，将学习深度增加至6层，以BCEWithLogitsLoss和ReLU分别作为损失函数和激活函数，RMSProp作为优化函数，对杉木横切面CT图像进行年轮分割，并对年轮分割性能进行评价。结果显示：改进UNet++模型对于杉木横切面CT图像的年轮分割的像素准确率为97.81%，骰子系数为98.89%，交并比为95.29%，平均交并比为84.75%，充分提取杉木年轮特征，分割效果最好。与U-Net模型和UNet++模型相比，改进UNet++模型在多数年轮被裂纹和虫孔切割，无法形成完整圆形闭合曲线的条件下，使分割的年轮具有很好的完整性和连续性，消除分割过程中的断裂和噪声现象；年轮分割结果不受裂纹、节子、虫孔等缺陷影响，结构非常清晰，有效解决多种缺陷干扰下的虫孔误分割和密集年轮欠分割等问题。

关键词: 杉木, 横切面, 年轮分割, CT图像, UNet++模型

LIU Shuai, GE Zhedong, LIU Xiaotong, GAO Yisheng, LI Yang, LI Mengfei. Improved UNet++ for Tree Rings Segmentation of Chinese Fir CT Images[J]. Computer Engineering and Applications, 2024, 60(5): 232-239.

刘帅, 葛浙东, 刘晓彤, 高宜生, 李阳, 李萌菲. 改进UNet++的杉木CT图像年轮分割[J]. 计算机工程与应用, 2024, 60(5): 232-239.

References

[1] POLK, PHILIPPE B, GERARD S. An automated method for tree-ring delineation based on active contours guided by DT-CWT complex coefficients in photographic images: application to Abies alba wood slice images[J]. Computers and Electronics in Agriculture, 2015, 118: 204-214.
[2] FANG K Y, HE M S, BAI M W, et al. The potential to use variations in tree-ring geometric center to estimate past wind speed change[J]. Natural Hazards Research, 2022, 2(2): 132-137.
[3] 宋馥杉, 方欧娅. 三江源国家公园大果圆柏生长衰退历史研究[J]. 森林与环境学报, 2019, 39(4): 386-392.
SONG F S, FANG O Y. Research on history of Juniperus tibetica growth declinein Three-River-Source National Park[J]. Journal of Forest and Environment, 2019, 39(4): 386-392.
[4] OKOCHI T, HOSHINO Y, FUJII H, et al. Nondestructive tree-ring measurements for Japanese oak and Japanese beech using micro-focus X-ray computed tomography[J]. Dendrochronologia, 2007, 24(2/3): 155-164.
[5] 王燕凤, 冯海林, 杜晓晨, 等. 一种对年轮图像双边滤波增强的树龄测量方法[J]. 林业工程学报, 2017, 2(5): 109-114.
WANG Y F, FENG H L, DU X C, et al. Improving the tree-age measurement by enhancing the tree-ring image with bilateral filter[J]. Journal of Forestry Engineering, 2017, 2(5): 109-114.
[6] 张露, 郭清宇, 梁静静, 等. 基于DOG算子的年轮图像边缘提取及树龄测量方法[J]. 中原工学院学报, 2019, 30(6): 50-54.
ZHANG L, GUO Q Y, LIANG J J, et al. Edge exteaction and tree age measurement method of annual ring image based on DOG operator[J]. Journal of Zhongyuan University of Technology, 2019, 30(6): 50-54.
[7] FABIJANSKA A, MALGORZATA D. DeepDendro-a tree rings detector based on a deep convolutional neural network[J]. Computers and Electronics in Agriculture, 2018, 150: 353-363.
[8] 宁霄, 赵鹏. 随机森林算法在树木年轮图像分割中的应用[J]. 林业工程学报, 2018, 3(4): 125-130.
NING X, ZHAO P. Image segmentation of tree ring based on the random forest algorithm[J]. Journal of Forestry Engineering, 2018, 3(4): 125-130.
[9] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[J]. Medical Image Computing and Computer-Assisted Intervention, 2015, 9351: 234-241.
[10] FABIJANSKA A, DANEK M, BARNIAK J, et al. Towards automatic tree rings detection in images of scanned wood samples[J]. Computers and Electronics in Agriculture, 2019, 140: 279-289.
[11] 宁霄, 赵鹏. 基于U-Net卷积神经网络的年轮图像分割算法[J]. 生态学杂志, 2019, 38(5): 1580-1588.
NING X, ZHAO P. Segmentation algorithm of annual ring image based on U-Net convolution network[J]. Chinese Journal of Ecology, 2019, 38(5): 1580-1588.
[12] ZHOU Z, RAHMAN S M M, TAJBAKHSH, et al. UNet++: a nested U-Net architecture for medical image segmentation[J]. Deep Learning Medical Image Analysis and Multimodal Learning for Clinical Decision Support, 2018, 11045: 3-11.
[13] 邢妍妍, 杨丰, 唐宇姣, 等. 融合型UNet++网络的超声胎儿头部边缘检测[J]. 中国图象图形学报, 2020, 25(2): 366-377.
XING Y Y, YANG F, TANG Y J, et al. Ultrasound fetal head edge detection using fusion UNet ++[J]. Journal of Image and Graphics, 2020, 25(2): 366-377.
[14] 荣亚琪, 张丽娟, 崔金利, 等. 基于NODE-UNet++和标记分水岭算法的红细胞图像分割[J]. 液晶与显示, 2022, 37(9): 1190-1198.
RONG Y Q, ZHANG L J, CUI J L, et al. Red blood cell image segmentation based on NODE-UNet++ and marker watershed[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(9): 1190-1198.
[15] LI Z, ZHANG H, LI Z Z, et al. Residual-attention unet++: a nested residual-attention U-NET for medical image segmentation[J]. Applied Sciences, 2022, 12(14): 7149.
[16] 崔少国, 文浩, 张宇楠, 等. 带有注意力机制的OCTA视网膜血管分割方法[J]. 计算机工程与应用, 2023, 59(18): 163-171.
CUI S G, WEN H, ZHANG Y, et al. Segmentation method of retinal vessels in OCTA with attention mechanism[J]. Computer Engineering and Applications, 2023, 59(18): 163-171.
[17] CHI J, ZHANG S, HAN X, et al. MID-UNet: multi-input directional UNet for COVID-19 lung infection segmentation from CT images[J]. Signal Processing: Image Communication, 2022, 108: 116835.
[18] WU J, ZHOU S, ZUO S, et al. U-Net combined with multi-scale attention mechanism for liver segmentation in CT images[J]. BMC Medical Informatics Decision Making, 2021, 21(1): 283.
[19] ZHOU Z W, RAHMAN S M M, NIMA T, et al. UNet++: redesigning skip connections to exploit multiscale features in image segmentation[J]. IEEE Transactions on Medical Imaging, 2020, 39(6): 1856-1868.
[20] 侯奕辰, 彭辉, 谢俊章, 等. 改进Unet++在脑肿瘤图像分割的研究[J]. 计算机工程与设计, 2022, 43(6): 1725-1731.
HOU Y C, PENG H , XIE J Z, et al. Research on segmentation of brain tumor image based on improved Unet++[J]. Computer Engineering and Design, 2022, 43(6): 1725-1731.
[21] 段杰, 崔志明, 沈艺, 等. 一种改进FCN的肝脏肿瘤CT图像分割方法[J]. 图学学报, 2020, 41(1): 100-107.
DUAN J, CUI Z M, SHEN Y, et al. A CT image segmentation method for liver tumor by an improved FCN[J]. Journal of Graphics, 2020, 41(1): 100-107.
[22] 杨阿庆, 薛月菊, 黄华盛, 等. 基于全卷积网络的哺乳母猪图像分割[J]. 农业工程学报, 2017, 33(23): 219-225.
YANG A Q, X Y J, H H S, et al. Lactating sow image segmentation based on fully convolutional networks[J]. Transactions of the Chinese Society of Agricultural Engineering, 2017, 33(23): 219-225.
[23] 杨蜀秦, 王鹏飞, 王帅, 等. 基于MHSA+DeepLab v3+的无人机遥感影像小麦倒伏检测[J]. 农业机械学报, 2022, 53(8): 213-219.
YANG S Q, WANG P F, WANG S, et al. Detection of wheat lodging in UAV remote sensing images based on Multi-Head Self-Attention DeepLab v3 +[J]. Transactions of the Chinese Society for Agricultural Machinery, 2022, 53(8): 213-219.