基于改进的Faster R-CNN模型的树冠提取研究

doi:10.13466/j.cnki.lyzygl.2021.01.022

林业资源管理 ›› 2021›› Issue (1): 173-179.doi: 10.13466/j.cnki.lyzygl.2021.01.022

基于改进的Faster R-CNN模型的树冠提取研究

黄彦晓¹^,²(), 方陆明¹^,³, 黄思琪⁴, 高海力⁵, 杨来邦¹, 楼雄伟¹^,²()

1.浙江农林大学信息工程学院,杭州 311300
2.林业感知技术与智能装备国家林业和草原局重点实验,杭州 311300
3.浙江省林业智能监测与信息技术研究重点实验室,杭州 311300
4.浙江农林大学暨阳学院,浙江诸暨 311800
5.浙江农林大学林业与生物技术学院,杭州 311300

收稿日期:2020-10-07 修回日期:2020-12-18 出版日期:2021-02-28 发布日期:2021-03-30
通讯作者: 楼雄伟
作者简介:黄彦晓(1996- ),女,浙江宁波人,在读硕士,主要从事农业工程与信息技术研究。Email: 2667513147@qq.com
基金资助:
浙江省科技重点研发计划资助项目(2018C02013)

Research on Crown Extraction Based on Improved Faster R-CNN Model

HUANG Yanxiao¹^,²(), FANG Luming¹^,³, HUANG Siqi⁴, GAO Haili⁵, YANG Laibang¹, LOU Xiongwei¹^,²()

1. School of Information Engineering,Zhejiang A & F University,Hangzhou Zhejiang 311300,China
2. Key Laboratory of State Forestry and Grassland Administration on Forestry Sensing Technology and Intelligent Equipment,Hangzhou Zhejiang 311300,China
3. Key Laboratory of Forestry Intelligent Monitoring and Information Technology of Zhejiang Province,Hangzhou Zhejiang 311300,China
4. Jiyang College of Zhejiang A & F University,Zhuji Zhejiang 311800,China
5. College of Forestry and Biotechnology,Zhejiang A & F University,Hangzhou Zhejiang 311300,China

Received:2020-10-07 Revised:2020-12-18 Online:2021-02-28 Published:2021-03-30
Contact: LOU Xiongwei

摘要/Abstract

摘要：

树冠信息是森林资源调查中的重要内容。传统的树冠冠幅测量方法为实地调查,该方法测量结果在特定的地形和森林环境中误差较大,且人力消耗大、操作繁琐、耗时长。无人机影像技术和深度学习的发展为树冠测量提供了新的方法和实现思路。利用无人机获取了临安东部青山湖绿道两块纯水杉林样地的正射影像图,通过改进目前先进的目标检测方法 Faster R-CNN进行树冠的识别和冠幅的提取。基于改进的Faster R-CNN模型准确率和决定系数达到了92.92%和0.84,分别比改进前的模型提高了5.31%和0.12。这说明了无人机和目标检测技术识别树冠的可行性,这一方法和传统的调查方法相比,具有高效、便捷和低成本的优势。

关键词: 无人机影像, 树冠识别, 冠幅测量, 目标检测, Faster R-CNN

Abstract:

Canopy information is an important part of forest resources investigation.The traditional method of crown width measurement is through field survey,which may result in a significant error in the specific terrain and forest environment,along with great labor force,cumbersome and time-consuming operation procedure.The development of UAV imaging technology and machine learning provides a new method and realization idea for crown measurement.This paper employed UAV to obtain the orthophoto images of two pure Metasequoia glyptostroboides in the greenway of Qingshanhu in the east of Lin'an District.An advanced object detection method,Faster R-CNN,was improved to recognize the tree crown and extract the crown width.The Accuracy and R ² of the improved Faster R-CNN model are 92.92% and 0.84 respectively,which are 5.31% and 0.12 higher than those of the original model.This shows that the UAV and object detection technology are feasible to identify the tree crown.Compared with the traditional survey method,it has the advantages of high efficiency,convenience and low cost.

Key words: UAV images, crown recognition, crown width measurement, object detection, Faster R-CNN

中图分类号:

S758
TP751

黄彦晓, 方陆明, 黄思琪, 高海力, 杨来邦, 楼雄伟. 基于改进的Faster R-CNN模型的树冠提取研究[J]. 林业资源管理, 2021,(1): 173-179.

HUANG Yanxiao, FANG Luming, HUANG Siqi, GAO Haili, YANG Laibang, LOU Xiongwei. Research on Crown Extraction Based on Improved Faster R-CNN Model[J]. FOREST RESOURCES WANAGEMENT, 2021,(1): 173-179.

图/表 9

图1

图2

图3

图4

图5

图6

图7

表1

图8

参考文献 19

[1]	郭孝玉. 长白落叶松人工林树冠结构及生长模型研究[D]. 北京:北京林业大学, 2013.
[2]	Lee H, Slatton K C, Roth B E, et al. Adaptive clustering of airborne LiDAR data to segment individual tree crowns in managed pine forests[J]. International Journal of Remote Sensing, 2010,31(1):117-139.
[3]	李晓靖. 基于高分影像的面向对象分类与单木树冠提取研究[D]. 北京:北京林业大学, 2017.
[4]	刘玉锋, 潘英, 李虎. 基于高空间分辨率遥感数据的天山云杉树冠信息提取研究[J]. 国土资源遥感, 2019,31(4):112-119.
[5]	李永亮, 张怀清, 杨廷栋, 等. 基于自适应神经模糊系统的杉木冠幅估算方法[J]. 林业科学, 2019,55(11):45-51.
[6]	Erfanifard Y, Behnia N, Moosavi V. Tree crown delineation on UltraCam-D aerial imagery with SVM classification technique optimised by Taguchi method in Zagros woodlands[J]. International Journal of Image and Data Fusion, 2014,5(4):300-314.
[7]	Hearst M A, Dumais S T, Osuna E, et al. Support vector machines[J]. IEEE Intelligent Systems and their applications, 1998,13(4):18-28.
[8]	Wu Bin, Yu Bailang, Wu Qiusheng, et al. Individual tree crown delineation using localized contour tree method and airborne LiDAR data in coniferous forests[J]. International Journal of Applied Earth Observation and Geo Information, 2016,52:82-94.
[9]	Lecun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015,521(7553):436-444. doi: 10.1038/nature14539 pmid: 26017442
[10]	Ren S, He K, Girshick R, et al. Faster R-CNN:Towards real-time object detection with region proposal networks[J] //IEEE Trans Pattern Anal Mach Intell. 2017,39(6):1137-1149. pmid: 27295650
[11]	Everingham M, Van Gool L, Williams C K I, et al. The pascal visual object classes(voc)challenge[J]. International Journal of Computer Vision, 2010,88(2):303-338.
[12]	Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation [C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2014: 580-587.
[13]	Girshick R. Fast R-CNN [C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.
[14]	Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:2014,1409:1556.
[15]	He Kaiming, Zhang Xiangyu, Ren Shaoqing, et al. Deep residual learning for image recognition [C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.
[16]	Lin T Y, Dollár P, Girshick R, et al. Feature pyramid networks for object detection [C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 2117-2125.
[17]	Perez L, Wang J. The effectiveness of data augmentation in image classification using deep learning[J]. arXiv preprint arXiv:2017,1712:04621.
[18]	Deng Jia, Dong Wei, Socher R, et al. Imagenet:A large-scale hierarchical image database [C]//2009 IEEE conference on computer vision and pattern recognition.Ieee, 2009: 248-255.
[19]	Abadi M, Barham P, Chen J, et al. Tensorflow:A system for large-scale machine learning [C]//OSDI'16 Proceedings of the 12th USENIX Conference on Operating Systems Design and Implementation. 2016: 265-283.

指标	Faster R-CNN	Faster R-CNN_Res101_FPN
精确率/%	99	99.06
召回率/%	88.39	93.75
准确率/%	87.61	92.92
均方根误差	0.46	0.36
平均绝对误差	0.31	0.27
决定系数	0.72	0.84

基于改进的Faster R-CNN模型的树冠提取研究

Research on Crown Extraction Based on Improved Faster R-CNN Model

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 19

相关文章 3

编辑推荐

Metrics

本文评价

[1]	汪帅, 卢楠, 郑红, 李晖, 彭检贵, 张同, 魏彦华. 基于深度学习的野生动物视频目标检测[J]. 林草资源研究, 2024, 0(1): 34-40.
[2]	郑琰睿, 杨林剑, 李曙光, 张永炬. 基于深度学习的森林火灾烟雾监测[J]. 林业资源管理, 2023, 0(4): 150-160.
[3]	乔正年, 马骏, 徐雁南. 无人机遥感在林木冠幅提取中的应用[J]. 林业资源管理, 2019, 0(1): 78-84.