一种基于GC-U-Net模型的立木胸径自动测量方法

doi:10.11707/j.1001-7488.LYKX20240617

摘要/Abstract

摘要：

目的: 针对传统测量立木胸径过程中费时耗力、易受人为因素影响、设备价格昂贵等问题，提出一种基于智能手机和深度学习的立木胸径自动测量方法。方法: 首先，为达到低成本的立木胸径自动测量需求，采用智能手机获取立木单目RGB图像；然后，为精准提取立木轮廓，提出一种基于GC-U-Net模型的立木图像分割算法，在传统U-Net分割模型基础上，集成VGG16和CBAM增强模型对立木树干特征的识别能力；最后，基于摄影测量原理，构建立木胸径测量模型，利用分割后的树干图像快速准确地计算出立木胸径。结果: 相较传统U-Net模型，GC-U-Net模型的平均交并比（mIoU）提升4.38%，平均像素准确率（mPA）提升6.08%，召回率（Recall）提升4.85%；相比Deeplabv3+、PSPNet、SegNet分割模型，GC-U-Net模型可取得更好的树干分割结果，mIoU分别提升6.04%、6.52%、11.0%，mPA分别提升7.93%、7.36%、12.31%，Recall分别提升6.88%、7.83%、11.08%；树干直径测量平均相对误差仅2.37%，拟合度达0.91。结论: 与传统立木胸径测量方法相比，本研究方法仅通过智能手机获取立木单目图像即可自动获取胸径，全程无需昂贵的专业设备；图像采集过程中也不需要参照物，简化了现场测量的复杂性，提高了测量效率；同时，GC-U-Net模型还可确保立木胸径测量的准确度。

关键词: GC-U-Net, 树干分割, 胸径测量, 智能手机

Abstract:

Objective: This study aims to address the problems of time-consuming and labor-intensive, susceptible to human error, and expensive equipment in the traditional process of measuring the diameter at breast height (DBH) of standing trees, for which an automatic method for measuring DBH of standing trees based on smartphone and deep learning is proposed. Method: First, in order to meet the demand for low-cost automatic measurement of standing tree DBH, a smartphone was used to acquire monocular RGB images of standing trees. Then a standing tree image segmentation algorithm based on GC-U-Net model was proposed to accurately extract the contours of standing trees. Based on the traditional U-Net segmentation model, the model's ability to recognize the features of the trunk of a standing tree was enhanced by integrating the VGG16 and CBAM attention mechanisms. Finally, based on photogrammetric principles, a standing tree diameter measurement model was constructed, and used to quickly and accurately calculate DBH of the standing trees using the segmented image. Result: The comparative experimental results showed that the GC-U-Net model improved the mean intersection ratio (mIoU) by 4.38%, the mean pixel accuracy (mPA) by 6.08%, and the recall by 4.85% over the traditional U-Net model. Compared with the PSPNet, SegNet, and Deeplabv3+ segmentation models, the GC-U-Net model also obtained better trunk segmentation, with mIoU being improved by 6.04%, 6.52%, and 11.0%, mPA being improved by 7.93%, 7.36%, and 12.31%, and Recall being improved by 6.88%, 7.83%, and 11.08%, respectively. The average relative error of trunk diameter measurement was only 2.37%, and the goodness-of-fit reached 0.91. Conclusion: Compared with the traditional method of measuring DBH, the method in this study can automatically obtain the DBH value of standing trees by only acquiring a monocular image of standing trees from a smartphone, which reduces the cost of measurement equipment. The image acquisition process does not require a reference, which simplifies the complexity of on-site measurements and improves the efficiency of the measurements. At the same time, the proposed GC-U-Net model ensures the measurement accuracy of the standing tree DBH.

Key words: GC-U-Net, trunk segmentation, measurement of diameter at breast height, mobile phone

中图分类号:

S758

常乐,杜晓晨,冯海林,李颜娥,黄坚钦. 一种基于GC-U-Net模型的立木胸径自动测量方法[J]. 林业科学, 2025, 61(4): 20-32.

Le Chang,Xiaochen Du,Hailin Feng,Yan’e Li,Jianqin Huang. An Automatic Measurement of Standing Tree Diameter at Breast Height Based on the GC-U-Net Model[J]. Scientia Silvae Sinicae, 2025, 61(4): 20-32.

图/表 15

图1

图2

图3

图4

图5

表1

图6

图7

表2

图8

表3

图9

图10

图11

图12

参考文献 0

	陈日强, 李长春, 杨贵军, 等. 无人机机载激光雷达提取果树单木树冠信息. 农业工程学报, 2020, 36 (22): 50- 59. doi: 10.11975/j.issn.1002-6819.2020.22.006
	Chen R Q, Li C C, Yang G J, et al. UAV airborne LiDAR extraction of single-tree canopy information for fruit trees. Transactions of the Chinese Society of Agricultural Engineering, 2020, 36 (22): 50- 59. doi: 10.11975/j.issn.1002-6819.2020.22.006
	管昉立, 徐爱俊. 基于智能手机与机器视觉技术的立木胸径测量方法. 浙江农林大学学报, 2018, 35 (5): 892- 899.
	Guan F L, Xu A J. Measurement method of standing tree diameter at breast height based on smartphone and machine vision technology. Journal of Zhejiang A&F University, 2018, 35 (5): 892- 899.
	李美琪, 刘美玲, 王　璇, 等. 八角林病虫害遥感识别模型. 林业科学, 2024, 60 (11): 128- 138. doi: 10.11707/j.1001-7488.LYKX20230559
	Li M Q, Liu M L, Wang X, et al. Remote sensing identification model for diseases and pests in star anise forests. Scientia Silvae Sinicae, 2024, 60 (11): 128- 138. doi: 10.11707/j.1001-7488.LYKX20230559
	潘　玺, 李　康, 杨　忠. 基于卷积神经网络的近红外光谱与数字图像特征信息融合木材树种识别. 林业科学, 2024, 60 (12): 136- 145.
	Pan X, Li K, Yang Z. Fusion of near-infrared spectral and digital image feature information based on convolutional neural network for wood species recognition. Scientia Silvae Sinicae, 2024, 60 (12): 136- 145.
	田智康, 葛浙东, 郑焕琪, 等. 基于深度学习的75种阔叶材微观辨识方法. 林业科学, 2024, 60 (10): 94- 103. doi: 10.11707/j.1001-7488.LYKX20240285
	Tian Z K, Ge Z D, Zheng H Q, et al. A deep learning-based method for the microscopic identification of 75 broadleaf species. Scientia Silvae Sinicae, 2024, 60 (10): 94- 103. doi: 10.11707/j.1001-7488.LYKX20240285
	Afify H M, Mohammed K K, Hassanien A E. An improved framework for polyp image segmentation based on SEGNET architecture. International Journal of Imaging Systems and Technology, 2021, 31 (3): 1741- 1751. doi: 10.1002/ima.22568
	Al−lQubaydhi N, Alenezi A, Alanazi T, et al. Deep learning for unmanned aerial vehicles detection: a review. Computer Science Review, 2024, 51, 100614. doi: 10.1016/j.cosrev.2023.100614
	Badrinarayanan V, Kendall A, Cipolla R. SegNet: a deep convolutional encoder−decoder architecture for image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (12): 2481- 2495. doi: 10.1109/TPAMI.2016.2644615
	Cao L J, Zheng X Y, Fang L M. The semantic segmentation of standing tree images based on the YoloV7 deep learning algorithm. Electronics, 2023, 12 (4): 929. doi: 10.3390/electronics12040929
	Cui Z Y, Wang X Y, Liu N Y, et al. Ship detection in large-scale SAR images via spatial shuffle-group enhance attention. IEEE Transactions on Geoscience and Remote Sensing, 2020, 59 (1): 379- 391.
	Du X C, Feng H L, Hu M Y, et al. Three-dimensional stress wave imaging of wood internal defects using TKriging method. Computers and Electronics in Agriculture, 2018, 148, 63- 71. doi: 10.1016/j.compag.2018.03.005
	Du X C, Zheng Y L, Feng H L. Stress wave hybrid imaging for detecting wood internal defects under sparse signals. Forests, 2024a, 15 (7): 1139. doi: 10.3390/f15071139
	Du X C, Zheng Y L, Feng H L. Optimizing sensor positions in the stress wave tomography of internal defects in hardwood. Forests, 2024b, 15 (3): 465. doi: 10.3390/f15030465
	Faghihi A, Fathollahi M, Rajabi R. Diagnosis of skin cancer using VGG16 and VGG19 based transfer learning models. Multimedia Tools and Applications, 2023, 83 (19): 57495- 57510. doi: 10.1007/s11042-023-17735-2
	García-Hidalgo M, García-Pedrero Á, Rozas V, et al. Tree ring segmentation using UNEt TRansformer neural network on stained microsections for quantitative wood anatomy. Frontiers in Plant Science, 2024, 14, 1327163. doi: 10.3389/fpls.2023.1327163
	Guo S J, Zhao Z X, Guo L Y, et al. A Method for measuring the absolute position and attitude parameters of a moving rigid body using a monocular camera. Applied Sciences, 2023, 13 (21): 11863. doi: 10.3390/app132111863
	Hong Y, Liu S J, Li Z P, et al. Airborne single-photon LiDAR towards a small−sized and low−power payload. Optica, 2024, 11 (5): 612- 618. doi: 10.1364/OPTICA.518999
	Hui Z Y, Lin L, Jin S G, et al. A reliable DBH estimation method using terrestrial LiDAR points through polar coordinate transformation and progressive outlier removal. Forests, 2024, 15 (6): 1031. doi: 10.3390/f15061031
	Kumar S, Furuhashi H. Long-range measurement system using ultrasonic range sensor with high-power transmitter array in air. Ultrasonics, 2017, 74, 186- 195. doi: 10.1016/j.ultras.2016.10.012
	Li G H, Weng X, Du X C, et al. Stress wave velocity patterns in the longitudinal-radial plane of trees for defect diagnosis. Computers and Electronics in Agriculture, 2016, 124, 23- 28. doi: 10.1016/j.compag.2016.03.021
	Liu Y Y, Bai X T, Wang J F, et al. Image semantic segmentation approach based on DeepLabV3 plus network with an attention mechanism. Engineering Applications of Artificial Intelligence, 2024, 127, 107260. doi: 10.1016/j.engappai.2023.107260
	Niu Z Y, Zhong G Q, Yu H. A review on the attention mechanism of deep learning. Neurocomputing, 2021, 452, 48- 62. doi: 10.1016/j.neucom.2021.03.091
	Ouyang D, He S, Zhang G Z, et al. 2023. Efficient multi-scale attention module with cross−spatial learning. ICASSP 2023—2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 1−5.
	Pan X R, Ge C J, Lu R, et al. 2022. On the integration of self-attention and convolution. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 815−825.
	Qiao Y C, Hu Y H, Zheng Z Z, et al. A diameter measurement method of red jujubes trunk based on improved PSPNet. Agriculture, 2022, 12 (8): 1140. doi: 10.3390/agriculture12081140
	Shi L J, Wang G Y, Mo L F, et al. Automatic segmentation of standing trees from forest images based on deep learning. Sensors, 2022, 22 (17): 6663. doi: 10.3390/s22176663
	Siddique N, Paheding S, Elkin C P, et al. U-Net and its variants for medical image segmentation: a review of theory and applications. IEEE Access, 2021, 9, 82031- 82057. doi: 10.1109/ACCESS.2021.3086020
	Tang H H, Li X B, Meng L, et al. Process modeling and optimization in laser drilling of bulk metallic glasses based on GABPNN and machine vision. Optics & Laser Technology, 2024, 172, 110502.
	Van Dyk D A, Meng X L. The art of data augmentation. Journal of Computational and Graphical Statistics, 2001, 10 (1): 1- 50. doi: 10.1198/10618600152418584
	Wan D H, Lu R S, Shen S Y, et al. Mixed local channel attention for object detection. Engineering Applications of Artificial Intelligence, 2023, 123, 106442. doi: 10.1016/j.engappai.2023.106442
	Wang S, Li R, Li H, et al. An automated method for stem diameter measurement based on laser module and deep learning. Plant Methods, 2023, 19 (1): 68. doi: 10.1186/s13007-023-01045-7
	Williams C, Falck F, Deligiannidis G, et al. A unified framework for U-Net design and analysis. Advances in Neural Information Processing Systems, 2023, 36, 27745- 27782.
	Woo S, Park J, Lee J Y, et al. 2018. Cbam: Convolutional block attention module. Proceedings of the European Conference on Computer Vision (ECCV), 3−19.
	Yan L F, Liu D W, Xiang Q, et al. PSP net-based automatic segmentation network model for prostate magnetic resonance imaging. Computer Methods and Programs in Biomedicine, 2021, 207, 106211. doi: 10.1016/j.cmpb.2021.106211
	Yu B, Yang L, Chen F. Semantic segmentation for high spatial resolution remote sensing images based on convolution neural network and pyramid pooling module. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2018, 11 (9): 3252- 3261. doi: 10.1109/JSTARS.2018.2860989
	Zhang J H, Gong J L, Zhang Y F, et al. Weed Identification in maize fields based on improved Swin-UNet. Agronomy, 2023, 13 (7): 1846. doi: 10.3390/agronomy13071846
	Zhu H Y, Gowen A, Feng H L, et al. Deep spectral-spatial features of near infrared hyperspectral images for pixel-wise classification of food products. Sensors, 2020, 20 (18): 5322. doi: 10.3390/s20185322

配置Configuration	参数Parameter
中央处理器Central processing unit (CPU) 图形处理器Graphics processing unit (GPU) 开发环境Development environment 加速环境Accelerated environment 模型优化器Model optimizer 训练轮数Epochs 学习率Learning rate	Intel(R) Core (TM) i5-13490F NVIDIA GeForce RTX4060Ti Pycharm CUDA12.1 CUDNN8.2.1 Adam 200 0.01

模型Model	mIoU（%）	mPA（%）	Recall（%）
DeepLabv3+	80.04	81.59	81.41
PSPNet	79.56	82.16	80.46
SegNet	75.09	77.21	77.21
U-Net	81.7	83.44	83.44
Swin-UNet	73.9	76.12	76.12
GC-U-Net	86.08	89.52	88.29

模型改进Model improvement	mIoU（%）	mPA（%）	$ \text{Rec}\text{all} $（%）
Axial	83.15	86.38	86.38
SE	82.92	85.79	85.21
CBAM	85.19	89.52	88.29
Axial+SE	82.06	84.61	84.61
Axial+CBAM	80.97	83.46	83.46
SE+CBAM	84.55	86.85	86.85
Axial+SE+CBAM	81.34	84.86	83.44
SA	84.52	87.18	87.18
PPM	84.89	87.62	87.62

[1]	刘霞, 凌成星, 陈永富, 刘华, 贺振平, 李泽江, 孙维娜, 马志杰, 由海霞, 吕文, 赵峰, 曾浩威, 王鑫淼. 鄂尔多斯地区13种典型灌木的可加性生物量模型及含碳率[J]. 林业科学, 2025, 61(4): 117-128.
[2]	吴璧芸,雷相东,何潇,李玉堂. 长白山天然针阔混交林优势高估计方法及立地质量评价[J]. 林业科学, 2025, 61(2): 85-92.
[3]	曾伟生,杨学云,蒲莹. 天然落叶松林木冠幅模型与林分密度模型及其关系[J]. 林业科学, 2025, 61(2): 93-100.
[4]	罗光成,何潇,雷相东,吴璧芸,向玮. 长白落叶松人工林优势高广义代数差分生长模型[J]. 林业科学, 2024, 60(12): 1-12.
[5]	韩新生,王彦辉,于澎涛,李振华,于艺鹏,王晓. 宁夏六盘山北部华北落叶松林树高与胸径生长的多因子响应耦合模型构建[J]. 林业科学, 2024, 60(11): 13-24.
[6]	张雨田,石军南,张怀清,吴炳伦. 洞庭湖湿地植被时空动态及其驱动力分析[J]. 林业科学, 2024, 60(8): 1-13.
[7]	张宇,张怀清,安锋,蒋玲,云挺. 基于计算机模拟模型的林木冠层太阳短波辐射定量分析方法[J]. 林业科学, 2024, 60(4): 16-30.
[8]	于帅,蔡体久,张丕德,任铭磊,张海宇,琚存勇. 边缘校正方法对空间结构参数影响的尺度效应[J]. 林业科学, 2023, 59(10): 57-65.
[9]	杨晓慧,吴金卓,刘浩然,钟浩,林文树. 基于UAV-LiDAR的人工林林分郁闭度估测[J]. 林业科学, 2023, 59(8): 12-21.
[10]	屈彦成,江怡航,姜彦妍,张建国,罗安利,张雄清. 基于胸高处边材面积、胸径和冠基部直径的杉木单木叶生物量预测模型[J]. 林业科学, 2023, 59(7): 106-114.
[11]	马颢铜,金光泽,刘志理. 小兴安岭红松胸高断面积生长随树龄的变化规律及主要影响因素[J]. 林业科学, 2023, 59(7): 96-105.
[12]	何培,王君杰,辛士冬,张兹鹏,姜立春. 天然樟子松和兴安落叶松树干削度方程4种建模方法比较[J]. 林业科学, 2023, 59(6): 28-35.
[13]	曾伟生,蒲莹,杨学云,易善军. 我国5种主要人工林乔木层碳储量生长模型及其气候驱动分析[J]. 林业科学, 2023, 59(3): 21-30.
[14]	郭旭展,陈巧,张晓芳,洪亮,尤媛媛,唐守正,符利勇. 基于无人机高分辨率影像的油松新造林健康树冠提取[J]. 林业科学, 2022, 58(10): 111-120.
[15]	黄宏超,谢栋博,段光爽,张状,张海江,符利勇. 华北落叶松和白桦半参数树高曲线模型[J]. 林业科学, 2022, 58(10): 101-110.