基于BP神经网络的长白落叶松人工林林分平均高预测

doi:10.3969/j.issn.1000-2006.201706012

国家林草科技领军期刊
中国精品科技期刊
中国高校百佳科技期刊
江苏省新闻出版政府奖期刊奖
RCCSE林学权威期刊（A+）
CSCD核心期刊
Scopus数据库收录期刊
中文核心期刊
SCD核心期刊

作者加群：102861116

微信公众号：南京林业大学学报

高级检索

南京林业大学学报（自然科学版） ›› 2018, Vol. 42 ›› Issue (02): 147-154.doi: 10.3969/j.issn.1000-2006.201706012

基于BP神经网络的长白落叶松人工林林分平均高预测

沈剑波¹, 雷相东^1*,李玉堂², 兰莹³

1.中国林业科学研究院资源信息研究所,北京 100091; 2.吉林省林业调查规划院,吉林长春 130022; 3. 吉林省长春市净月潭实验林场,吉林长春 130022

出版日期:2018-04-12 发布日期:2018-04-12
基金资助:
基金项目:国家林业公益性行业科研专项重大项目(201504303) 第一作者:沈剑波(lyshenjianbo@163.com),博士生。*通信作者:雷相东(xdlei@caf.ac.cn),研究员,博士。

Prediction mean height for Larix olgensis plantation based on Bayesian-regularization BP neural network

SHEN Jianbo¹, LEI Xiangdong^1*, LI Yutang², LAN Ying³

1.Institute of Resources Information Techniques,CAF, Beijing 100091, China; 2. Forest Inventory and Planning Institute of Jilin Province, Changchun 130022, China; 3.Jingyuetan Experimental Forest Farm of Changchun City Jilin Province, Changchun 130022, China

Online:2018-04-12 Published:2018-04-12

摘要/Abstract

摘要： 【目的】研究BP神经网络模型在树高预测中的应用,分析比较不同森林调查因子及不同神经网络训练算法对平均树高预测的影响,为树高预测提供新的方法。【方法】以吉林省长白落叶松人工林为对象,基于168块固定样地的314个观测数据,运用BP神经网络建模技术建立了林分平均树高生长模型。输入因子首先加入年龄,然后依次加入立地因子及林木竞争因子,分析立地因子及林木竞争因子对树高的影响。基于Matlab R2016b中的Sigmoid函数和线性函数为神经元的传递函数,分别采用贝叶斯正则化算法和Levenberg-Marquatdt算法(简称L-M算法)对网络进行训练,对比分析了贝叶斯正则化算法和L-M算法作为训练函数的差异。【结果】与L-M训练算法相比,贝叶斯正则化训练算法具有更好的泛化能力。模型中依次加入年龄、立地因子、林木竞争因子后,树高的拟合精度呈现出相同的上升趋势。采用贝叶斯正则化训练算法,当年龄作为输入因子时,决定系数R²为0.521 0,均方根误差(RMSE)为2.091 7,平均绝对误差(MAE)为1.627 6。加入立地因子后,决定系数R²提高至0.573 6,提高了10.10%,均方根误差(RMSE)为1.973 6,降低了5.65%,平均绝对误差(MAE)为1.579 7,降低了2.94%; 在此基础上,加入林木竞争因子后,决定系数R²为0.845 5,增长了47.40%, 均方根误差(RMSE)为1.187 9,下降了39.81%,平均绝对误差(MAE)为0.968 5,下降了38.69%。【结论】利用贝叶斯正则化BP神经网络可以准确地预测长白落叶松人工林的平均高。立地因子及林木竞争因子能够较好地提升林木生长预测的精度,且林木竞争因子对树高的影响明显大于立地因子。

Abstract: 【Objective】The neural network model was used for stand height prediction. Different types of forest survey factors and neural network training algorithm were examined. The results were expected to provide new method for stand height prediction.【Method】Stand mean height prediction model was developed based on the 314 observations from 168 permanent sample plots of Larix olgensis plantation in Jilin Province with BP neural network modeling technology. Firstly we input the age, then input the site factor and competitive factor to analyze their contributions to stand mean height. We applied Bayesian-regularization algorithm and Levenberg-Marquatdt algorithm for network training based on tansig function and purelin function in Matlab R2016b as neural transferring function. We compared and analyzed the differences of the two algorithms as training functions. 【Result】The analysis result showed that no matter which training algorithm was adopted, the estimation precision of stand mean height would be improved after the age, site factor and forest competitive factor were added in succession and Bayesian-regularization algorithm had better generalization ability than L-M training algorithm. The coefficient of determination R² was 0.521 0, RMSE was 2.091 7, MAE was 1.627 6 when Bayesian-regularization algorithm was adopted and age as input factor; when site factor was added subsequently, R², RMSE and MAE was 0.573 6(increased by 10.10%), 1.973 6(decreased by 5.65%), and 1.579 7(decreased by 2.94%), respectively; when forest competitive factor was added, R², RMSE and MAE was 0.845 5(increased by 47.40%), 1.187 9(decreased by 39.81%), and 0.968 5(decreased by 38.69%), respectively.【Conclusion】Bayesian regularization neural network modeling could be accurate for stand mean height prediction of Larix olgensis plantation. The precision was improved after site and forest competitive factor were taken into consideration and forest competitive factor gave more contribution than site factor.

中图分类号:

S758

沈剑波,雷相东,李玉堂,等. 基于BP神经网络的长白落叶松人工林林分平均高预测[J]. 南京林业大学学报（自然科学版）, 2018, 42(02): 147-154.

SHEN Jianbo, LEI Xiangdong, LI Yutang, LAN Ying. Prediction mean height for Larix olgensis plantation based on Bayesian-regularization BP neural network[J].Journal of Nanjing Forestry University (Natural Science Edition), 2018, 42(02): 147-154.DOI: 10.3969/j.issn.1000-2006.201706012.

参考文献

[1] 孟宪宇.测树学[M].3版.北京: 中国林业出版社,2006.
[2] 隋宏大.树高测量综合技术比较研究[D].北京:北京林业大学,2009. SUI H D. Comparative study on comprehensive technologies of tree height measurement[D].Beijing:Beijing Forestry University, 2009.
[3] 王明亮,唐守正.标准树高曲线的研制[J].林业科学研究,1997,10(3):36-41. WANG M L, TANG S Z. Research on universal height-diameter curves[J]. Forest Research, 1997,10(3):36-41.
[4] 丁贵杰.马尾松人工林标准树高曲线模型的研究[J]. 浙江林学院学报,1997,14(3):225-230. DING G J. Curve model of masson pine planted forest[J]. Journal of Zhejiang Forestry College, 1997,14(3):225-230.
[5] 吕勇.林木树高曲线模型研究[J].中南林学院学报,1997(4):86-89. LV Y. Models for tree height curves of Pinus massoniana and Cunninghamia lanceolata[J]. Journal of Central South Forestry University, 1997(4):86-89.
[6] 王明亮,李希菲. 非线性树高曲线模型的研究[J].林业科学研究, 2000, 13(1):74-79.DOI:10.3321/j.issn:1001-1498.2000.01.012. WANG M L, LI X F.Research on nonlinear height-diameter models[J]. Forest Research, 2000, 13(1):74-79.
[7] 曾翀,雷相东,刘宪钊,等. 落叶松云冷杉林单木树高曲线的研究[J].林业科学研究,2009, 22(2): 182-189.DOI:10.3321/j.issn:1001-1498.2009.02.006. ZENE C, LEI X D, LIU X Z, et al. Individual tree height-diameter curves of larch-spruce-fir forests[J]. Forest Research, 2009, 22(2): 182-189.
[8] 赵俊卉,亢新刚,刘燕.长白山主要针叶树种最优树高曲线研究[J]. 北京林业大学学报, 2009, 31(4):13-18.DOI:10.3321/j.issn:1000-1522.2009.04.003. ZHAO J H, KANG X G, LIU Y. Optimal height-diameter models for dominant coniferousspecies in Changbai Mountain, northeastern China[J].Journal of Beijing Forestry University, 2009, 31(4):13-18.
[9] 陈东升,孙晓梅,李凤日.基于混合模型的落叶松树高生长模型[J].东北林业大学学报, 2013,41(10):60-64.DOI:10.3969/j.issn.1000-5382.2013.10.013. CHEN D S, SUN X M, LI F R. Predicting models of tree height growth for larch based on mixed model[J].Journal of Northeast Forestry University, 2013,41(10):60-64.
[10] KNOWE S A. Effect of competition control treatments on height-age and height-diameter relationships in young Douglas-fir plantations[J]. Forest Ecology & Management, 1994, 67(1/2/3):101-111. DOI:10.1016/0378-1127(94)90010-8.
[11] CALEGARIO N, DANIELS R F, MAESTRI R, et al. Modeling dominant height growth based on nonlinear mixed-effects model: a clonal Eucalyptus plantation case study[J]. Forest Ecology and Management, 2005, 204(1):11-20. DOI:10.1016/j.foreco.2004.07.051.
[12] 李际平,姚东和.BP模型在单木树高与胸径生长模拟中的应用[J].中南林学院学报,1996,16(3):34-36. LI J P, YAO D H. Application of BP neural network model to the simulation of breast-height diameter and tree-height growth[J]. Journal of Central South Forestry University, 1996,16(3):34-36.
[13] 马天晓,赵晓峰,黄家荣,等. 基于人工神经网络的树高曲线模型研究[J].河南林业科技,2006,26(1):4-5.DOI:10.3969/j.issn.1003-2630.2006.01.002. MA T X, ZHAO X F, HUANG J R, et al. Research on the height-diameter model of tree with artificial neural network[J]. Journal of Henan Forestry Science and Technology, 2006,26(1):4-5.
[14] 徐志扬.基于BP神经网络的马尾松树高曲线模型[J].林业调查规划,2015,40(2):6-8. DOI:10.3969/j.issn.1671-3168.2015.02.002. XU Z Y. Height-diameter model for Pinus massoniana based on BP neural network[J].Forest Inventory and Planning, 2015, 40(2):6-8.
[15] 董云飞,孙玉军,王轶夫,等.基于BP神经网络的杉木标准树高曲线[J].东北林业大学学报,2014,42(7):154-156,165. DOI:10.3969/j.issn.1000-5382.2014.07.036. DONG Y F, SUN Y J, WANG Y F, et al. Generalized height-diameter model for Chinese fir based on BP neural network[J]. Journal of Northeast Forestry University,2014,42(7):154-156,165.
[16] 金星姬, 贾炜玮, 李凤日. 基于BP人工神经网络的兴安落叶松天然林全林分生长模型的研究[J]. 植物研究, 2008, 28(3):370-374. JIN X J, JIA W W, LI F R. Whole stand growth model for natural dahurian larch forests based on BP ANN[J]. Bulletin of Botanical Research, 2008, 28(3):370-374.
[17] 龙滔, 覃连欢, 叶绍明. 基于BP神经网络连栽桉树人工林生长量预测[J]. 东北林业大学学报, 2012, 40(5):122-125.DOI:10.3969/j.issn.1000-5382.2012.05.030. LONG T, QIN L H, YE S M. Prediction for the growth of eucalyptus plantations with continuous-planting rotations based on BP neural network[J].Journal of Northeast Forestry University, 2012, 40(5):122-125.
[18] 徐步强, 张秋良, 弥宏卓,等. 基于BP神经网络的油松人工林生长模型[J]. 东北林业大学学报, 2011, 39(12):33-35.DOI:10.3969/j.issn.1000-5382.2011.12.010. XU B Q, ZHANG Q L, MI H Z, et al. Growth model of pinus tabulaeformis plantation based on BP neural network[J]. Journal of Northeast Forestry University, 2011, 39(12):33-35.
[19] 肖兴威,王祝雄,陈雪峰,等.北京:国家森林资源连续清查技术规定[R]. 北京:国家林业局,2004.
[20] 张德丰.MATLAB神经网络应用设计[M].北京:机械工业出版社,2010.
[21] 周治平,孙子文,吴志健.采用L-M算法的JPEG图像隐写分析[J].计算机工程与应用, 2009, 45(14):113-115. DOI:10.3778/j.issn.1002-8331.2009.14.034. ZHOU Z P, SUN Z W, WU Z J.Steganalysis for JPEG images based on L-M algorithm[J].Computer Engineering and Applications, 2009, 45(14):113-115.
[22] SUN J, MAO H, LIU J, et al. The research of paddy rice moisture lossless detection based on L-M BP neural network[J]. Computer & Computing Technologies in Agriculture, 2008, 120(1):83-93. DOI:10.1007/978-1-4419-0211-5_46.
[23] 宋雷,黄藤,方剑,等.基于贝叶斯正则化BP神经网络的GPS高程转换[J].西南交通大学学报,2008,43(6):724-728.SONG L, HUANG T, FANG J, et al. Conversion of GPS height based on Bayesian regularization BP neural network[J].Journal of Southwest Jiaotong University, 2008,43(6):724-728.
[24] 彭开香,余尚志.基于贝叶斯神经网络的带钢厚度预测与控制[J].北京科技大学学报,2010,32(2):256-262. PENG K X, YU S Z. Prediction and control of strip thickness based on Bayesian neural networks[J].Journal of University of Science and Technology Beijing, 2010,32(2):256-262.
[25] MACKAY D J C. Bayesian interpolation[J]. Neural Computation, 1992, 4(3):415-447. DOI:10.1162/neco.1992.4.3.415.
[26] 刘春艳,凌建春,寇林元,等. GA-BP神经网络与BP神经网络性能比较[J].中国卫生统计, 2013, 30(2):173-176.DOI:10.3969/j.issn.1000-2006. LIU C Y, LING J C, KOU L Y, et al. Performance comparison between GA-BP neural network and BP neural network[J]. Chinese Journal of Health Statistics, 2013, 30(2):173-176.

基于BP神经网络的长白落叶松人工林林分平均高预测

Prediction mean height for Larix olgensis plantation based on Bayesian-regularization BP neural network

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

作者

读者

审稿

[1]	牛弘健, 刘文萍, 陈日强, 宗世祥, 骆有庆. 基于Resnet的林地无人机图像去雾改进算法[J]. 南京林业大学学报（自然科学版）, 2024, 48(2): 175-18.
[2]	高谢雨, 董利虎, 郝元朔. 基于TLS的抚育间伐对长白落叶松干形的影响[J]. 南京林业大学学报（自然科学版）, 2023, 47(6): 85-94.
[3]	李双娴, 陆鑫, 多杰才仁, 张怀清, 薛联凤, 云挺. 一种新的地面激光点云中树木叶面积计算方法[J]. 南京林业大学学报（自然科学版）, 2023, 47(5): 28-38.
[4]	张华聪, 谭新建, 喻龙华, 厉月桥, 陈永富, 刘仁, 张怀清. 基于增强Frost局部滤波及单木距离图重构标记的CHM树冠分割[J]. 南京林业大学学报（自然科学版）, 2023, 47(5): 9-18.
[5]	孙宇, 李凤日, 谢龙飞, 董利虎. 基于林分及地形因子的落叶松人工林林分生物量模型构建[J]. 南京林业大学学报（自然科学版）, 2023, 47(3): 129-136.
[6]	唐依人, 贾炜玮, 王帆, 孙毓蔓, 张颖. 基于TLS辅助的长白落叶松一级枝条生物量模型构建[J]. 南京林业大学学报（自然科学版）, 2023, 47(2): 130-140.
[7]	欧建德. 杈干现象对南方红豆杉树冠形态、生长和形质的影响[J]. 南京林业大学学报（自然科学版）, 2023, 47(2): 87-94.
[8]	王帆, 贾炜玮, 唐依人, 李丹丹. 基于TLS的红松树冠半径提取及其外轮廓模型构建[J]. 南京林业大学学报（自然科学版）, 2023, 47(1): 13-22.
[9]	程晓菲, 武刚. 基于高分辨率卫星影像的CV模型单木定位法[J]. 南京林业大学学报（自然科学版）, 2022, 46(5): 143-151.
[10]	聂璐毅, 董利虎, 李凤日, 苗铮, 谢龙飞. 基于两水平非线性混合效应模型的长白落叶松削度方程构建[J]. 南京林业大学学报（自然科学版）, 2022, 46(3): 194-202.
[11]	赵青, 黄菲, 陈晓辉, 林玉英, 邱荣祖, 巫志龙, 胡喜生. 乔木林丧失的时空变化及驱动力分析[J]. 南京林业大学学报（自然科学版）, 2022, 46(2): 227-235.
[12]	辛士冬, 姜立春, 穆林. 黑龙江省红松人工林林分乔木层可加性碳储量模型[J]. 南京林业大学学报（自然科学版）, 2022, 46(1): 115-121.
[13]	崔泽宇, 张怀清, 左袁青, 杨廷栋, 刘洋, 张京, 王林龙. 杉木三维模型各方向枝下高分布研究[J]. 南京林业大学学报（自然科学版）, 2022, 46(1): 81-87.
[14]	单凯丽, 臧颢, 潘萍, 宁金魁, 欧阳勋志. 赣中天然闽楠单木冠幅预测模型的研究[J]. 南京林业大学学报（自然科学版）, 2022, 46(1): 88-96.
[15]	卢军, 刘宪钊, 孟维亮, 李红军. 基于地面激光点云数据的单木三维重建方法[J]. 南京林业大学学报（自然科学版）, 2021, 45(6): 193-199.