基于CP-DeepLabv3+的玉米根系图像分割

doi:10.13304/j.nykjdb.2022.0996

摘要/Abstract

摘要：

利用微根管技术可以直接监测植物根系动态生长，并获取清晰根系图像，但土壤环境复杂、颗粒不均匀、细根数量多，图像分割时容易造成根系不连续，将土壤背景误认为根系。针对以上问题，提出了CP-DeepLabv3+算法进行图像分割。该算法引入坐标注意力机制（coordinate attention， CA），更精确地获得分割目标信息，使得分割目标边缘更加连续；在ASPP特征提取模块加入条纹池化（strip pooling，SP）分支，避免在相距较远的位置之间建立不必要的连接，提高图像分割精度。利用CP-DeepLabv3+算法对玉米根系数据集进行测试，结果显示，平均交并比（mean intersection over union，MIoU）值为82.95%，平均像素精确度（mean pixel accuracy，MPA）值为92.47%，相比于原始DeepLabv3+模型分别提高了3.69%、4.44%，表明该算法可有效分割玉米根系，对图像特征提取具有实际意义。

关键词: 玉米根系, 微根管法, 原位监测, CP-DeepLabv3+

Abstract:

Minirhizotron technique can directly monitor the dynamic growth and development of plant roots and be used to obtain clear root images. However， because of the complex soil environment， uneven particle size and large number of fine roots， it easily causes discontinuity of the divided roots and mistakes the soil background as the root. To solve the above problems， the CP-DeepLabv3+ algorithm was proposed to segment image. The coordinate attention mechanism （CA） was introduced to effectively segment the target location information and made the edge of the target more continuous. Strip pooling （SP） branch was added to ASPP feature extraction module to avoid unnecessary connections between distant locations and improve the accuracy of image segmentation. CP-DeepLabv3+ algorithm was applied to test maize root dataset. The results showed that the mean intersection-over-union （MIoU） value was 82.95%， the mean pixel accuracy （MPA） value was 92.47%， which was 3.69% and 4.44% higher than the original DeepLabv3+ model， respectively. This algorithm could effectively segment maize root and has practical significance for feature extraction.

Key words: maize root system, microroot canal method, in situ monitoring, CP-DeepLabv3+

中图分类号:

S513

赵亚凤, 王孟雪, 王德帅, 王冬冬, 李园, 胡峻峰. 基于CP-DeepLabv3+的玉米根系图像分割[J]. 中国农业科技导报, 2024, 26(3): 110-116.

Yafeng ZHAO, Mengxue WANG, Deshuai WANG, Dongdong WANG, Yuan LI, Junfeng HU. Maize Root Image Segmentation Based on CP-DeepLabv3+[J]. Journal of Agricultural Science and Technology, 2024, 26(3): 110-116.

图/表 8

图1 微根管采集图像装置

Fig. 1 Micro root canal image acquisition device

图2 CP-DeepLabv3+网络模型

Fig. 2 CP-DeepLabv3+network model

表1 软硬件环境配置

Table 1 Software and hardware environment configuration

实验环境 Experimental environment	版本配置 Version Configuration
操作系统 Operating system	Windows 10
CPU	Intel i5-8300
GPU	GeForce RTX 2080×2
CUDA	10.2
Python	3.7
框架 Frame	Pytorch 1.7.1

图3 图像预处理

Fig. 3 Pretreated image

表2 不同网络模型评估结果

Table 2 Evaluation results of different network models

模型 Model	交并比IoU/%					准确率 PA/%	平均准确率MPA/%
模型 Model	根系 Root	背景 Background			平均 MIoU/%	准确率 PA/%	平均准确率MPA/%
Unet	57.61		92.43	75.02		93.13	83.23
DeepLabv3+（Xception）	64.94		93.57	79.26		94.25	88.03
DeepLabv3+（CA）（MobileNetv2）	66.28		93.88	80.08		94.55	89.08
DeepLabv3+（CA+CFF）（ResNet）	66.42		94.04	80.23		94.52	89.95
DeepLabv3+（DenseASPP）（Xception）	66.49		94.23	80.36		94.60	89.98
CP-DeepLabv3+	70.18		95.72	82.95		96.58	92.47

图4 不同时期对应的网络模型分割效果注：方框所示为分割效果不同的部分。

Fig. 4 Network model segmentation effect corresponding to different periodsNote： The frame shows the parts with different segmentation effects.

图5 损失值曲线

Fig. 5 Loss value curve

图6 平均交并比曲线

Fig. 6 MIoU curve

参考文献 21

1	牛学礼,南志标.运用微根管技术研究草地植物细根的进展[J].草业学报,2017,26(11):205-215.
	NIU X L, NAN Z B. Rewiew of minirhizotron applications for study of fine roots in grassland [J]. Acta Pratac. Sin., 2017, 26(11):205-215.
2	DANJON F, REUBENS B. Assessing and analyzing 3D architecture of woody root systems, a review of methods and applications in tree and soil stability, resource acquisition and allocation [J]. Plant Soil, 2008, 303: 1-34.
3	BATES G H. A device for the observation of root growth in the soil [J]. Nature, 1937,139: 966–967.
4	廖荣伟,刘晶淼,安顺清,等.基于微根管技术的玉米根系生长监测[J].农业工程学报,2010,26(10):156-161.
	LIAO R W, LIU J M, AN S Q, et al.. Monitor of corn root growth in soil based on minirhizotron technique [J]. Trans. Chin. Soc. Agric. Eng., 2010,26(10):156-161.
5	李燕丽,王昌昆,卢碧林,等.基于微根管技术的盐胁迫下小麦根系生长原位监测方法[J].土壤学报,2021,58(3):599-609.
	LI Y L, WANG C K, LU B L, et al.. In-situ monitoring method of wheat root growth under salt stress using minirhizotron technique [J]. Acta Pedol. Sin., 2021, 58(3): 599-609.
6	刘凯,李文彬,赵玥,等.基于微根管图像的根系形态特征快速提取技术[J].中国水土保持科学,2021,19(4):129-136.
	LIU K, LI W B, ZHAO Y, et al.. Rapid extraction technology of the root morphological characteristics via minirhizotron image[J]. Sci. Soil Water Conservation, 2021, 19(4): 129-136.
7	李克新,李沐阳,薛瑞,等.林木幼苗根系 CT 序列图像分割[J].森林工程,2014,30 (1):25-29.
	LI K X, LI M Y, XUE R, et al.. CT slice image segmentation of the seedling roots [J]. Forest Eng., 2014,30 (1):25-29.
8	佘丽萱,康佳,王楠,等.一种新的棉花根系图像阈值分割方法[J].河北大学学报(自然科学版),2022,42(2):124-130.
	SHE L X, KANG J, WANG N, et al.. A new threshold segmentation method for cotton root images [J]. J. Hebei Univ. (Nat. Sci. ), 2022, 42(2): 124-130.
9	YASRAB R, ATKINSON J A, WELLS D M, et al.. RootNav 2.0: Deep learning for automatic navigation of complex plant root architectures [J/OL]. GigaScience, 2019,8(11): giz123 [2022-11-16]. .
10	WANG T, ROSTAMZA M, SONG Z, et al.. SegRoot: a high throughput segmentation method for root image analysis [J]. Comput. Electron. Agric., 2019, 162: 845-854.
11	WASSON A, BISCHOF L, ZWART A, et al.. A portable fluorescence spectroscopy imaging system for automated root phenotyping in soil cores in the field [J]. J. Exp. Bot., 2016,67(4): 1033-1043.
12	SMITH A G, PETERSEN J, SELVAN R, et al.. Segmentation of roots in soil with U-Net [J/OL]. Plant Methods, 2020,16:13 [2022-11-16]. .
13	申晨.基于深度学习的棉花原位根系分割技术研究[D].保定:河北农业大学,2021.
	SHEN C. Research on cotton in situ root segmentation technology based on deep learning [D]. Baoding: Hebei Agricultural University, 2021.
14	HOU Q B, ZHOU D Q, FENG J H. Coordinate attention for efficient mobile network design [C]// Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021.
15	HOU Q, ZHANG L, CHENG M M, et al.. Strip pooling: Rethinking spatial pooling for scene parsing [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 4003-4012.
16	CHEN L C, ZHU Y, PAPANDREOU G, et al.. Encoder-decoder with atrous separable convolution for semantic image segmentation [C]// Proceedings of the European conference on computer vision (ECCV). Computer Vision Foundation, 2018: 801-818.
17	黄聪,杨珺,刘毅,等.基于改进DeepLabV3+的遥感图像分割算法[J].电子测量技术,2022,45(21):148-155.
	HUANG C, YANG J, LIU Y, et al.. Remote sensing image segmentation algorithm based on improved DeeplabV3+ [J]. Electron. Meas. Technol., 2022,45(21):148-155.
18	李宁,张彦辉,尚英强,等.基于改进DeepLabv3+网络的风机叶片分割算法研究[J].电子技术应用,2022,48(9):108-113, 118.
	LI N, ZHANG Y H, SHANG Y Q,et al.. Research on fan blade segmentation algorithm based on improved DeepLabv3+ network [J]. Appl. Electron. Tech., 2022, 48(9):108-113, 118.
19	谭国金,欧吉,艾永明,等.基于改进DeepLabv3+模型的桥梁裂缝图像分割方法[J].吉林大学学报(工学版), 2022, 48(9):108-113, 118.
	TAN G J, OU J, AI Y M, et al.. Bridge crack image segmentation method based on improved DeepLabv3+ model [J]. J. Jilin Univ.(Eng. Technol.): 2022, 48(9):108-113, 118.
20	肖万欣,王延波,赵海岩,等.中国不同年代玉米根系形态生理功能特性的演化[J].辽宁农业科学,2022,(3):45-49.
	XIAO W X, WANG Y B, ZHAO H Y, et al.. Evolution of morphological and physiological function traits of corn root released in different ages of China[J].Liaoning Agric. Sci.,2022(3):45-49.
21	TIAN Y, YANG G D, WANG Z, et al.. Detection of apple lesions in orchards based on deep learning methods of CycleGAN and YOLOV3-dense [J/OL]. J. Sen., 2019, 7630926 [2022-11-16]. .