Processing math: 100%
  • Scopus收录期刊
  • CSCD(核心库)来源期刊
  • 中文核心期刊
  • 中国科技核心期刊
  • F5000顶尖学术来源期刊
  • RCCSE中国核心学术期刊
高级检索

融合多光谱频域特征的坡垒相对叶绿素含量预测

袁莹, 王雪峰, 石蒙蒙, 王鹏, 陈星京

袁莹, 王雪峰, 石蒙蒙, 王鹏, 陈星京. 融合多光谱频域特征的坡垒相对叶绿素含量预测[J]. 北京林业大学学报, 2023, 45(11): 42-52. DOI: 10.12171/j.1000-1522.20230113
引用本文: 袁莹, 王雪峰, 石蒙蒙, 王鹏, 陈星京. 融合多光谱频域特征的坡垒相对叶绿素含量预测[J]. 北京林业大学学报, 2023, 45(11): 42-52. DOI: 10.12171/j.1000-1522.20230113
Yuan Ying, Wang Xuefeng, Shi Mengmeng, Wang Peng, Chen Xingjing. Prediction of relative chlorophyll content in Hopea hainanensis based on multispectral frequency domain features[J]. Journal of Beijing Forestry University, 2023, 45(11): 42-52. DOI: 10.12171/j.1000-1522.20230113
Citation: Yuan Ying, Wang Xuefeng, Shi Mengmeng, Wang Peng, Chen Xingjing. Prediction of relative chlorophyll content in Hopea hainanensis based on multispectral frequency domain features[J]. Journal of Beijing Forestry University, 2023, 45(11): 42-52. DOI: 10.12171/j.1000-1522.20230113

融合多光谱频域特征的坡垒相对叶绿素含量预测

基金项目: 海南省院士创新平台科研专项(YSPTZX202001),国家自然科学基金项目(32071761)。
详细信息
    作者简介:

    袁莹。主要研究方向:林业信息技术应用。Email:paru_salt@163.com 地址:100091北京市海淀区香山路东小府1号中国林业科学研究院资源信息研究所

    责任作者:

    王雪峰,博士,研究员。主要研究方向:林业信息技术。Email:xuefeng@ifrit.ac.cn 地址:同上。

  • 中图分类号: S758;Q949;S758.6

Prediction of relative chlorophyll content in Hopea hainanensis based on multispectral frequency domain features

  • 摘要:
    目的 

    研究坡垒叶绿素含量的多光谱图像预测法,探讨融合多光谱频域特征预测叶绿素含量的可行性,为坡垒叶绿素含量的无损监测提供有效工具。

    方法 

    采用将植被指数与传统的阈值分割法相结合的方式,去除坡垒冠层多光谱图像的背景,以F1为分割精度评价指标,确定最适多光谱图像分割方法。对分割后的坡垒冠层多光谱图像,精准提取空域特征(植被指数与纹理特征),并引入3种频域特征,然后基于相关性分析和Lasso算法筛选图像特征,以便携式叶绿素仪测得的SPAD值作为相对叶绿素含量实测值,确定与坡垒SPAD值强相关的优选特征及合适的建模特征组合,结合偏最小二乘回归(PLSR)、随机森林(RF)和极限梯度提升(XGBoost)模型,分别建立多光谱空、频、融合特征模型并进行精度验证,确定适用于幼龄坡垒SPAD值预测的模型形式。

    结果 

    差值植被指数与Kapur阈值结合的分割方法获得了最高分割精度,评价指标F1达到0.917,为坡垒冠层多光谱图像的最适分割方法。多光谱图像空频域特征表现了与坡垒SPAD值的显著相关性,其中相关性最强的特征为修正叶绿素吸收反射率植被指数,相关系数达到−0.780,为基于单图像特征预测SPAD值的优选特征。在3种频域特征中,小波特征与SPAD值的相关性表现最优。因此,小波变换为优选坡垒多光谱图像频域变换方法。基于不同图像特征构建的SPAD值预测模型,性能表现排序为单频域特征模型 < 单空域特征模型 < 融合特征模型,最适建模算法为RF、XGBoost算法。基于RF的融合特征模型为最适模型,检验R2达到0.791,较单空域特征模型的检验R2提高了7.9%。

    结论 

    引入3种频域特征能够提高坡垒SPAD值预测精度,且基于RF的融合特征模型获得了较高的预测精度,因此融合多光谱空频域特征并结合机器学习算法,可作为一种有效的幼龄坡垒相对叶绿素含量预测工具,有利于坡垒培育经营工作的智能化发展。

    Abstract:
    Objective 

    This paper studies the multispectral image-based estimation method for chlorophyll content in Hopea hainanensis, so as to explore the feasibility of fusing multispectral frequency domain features to estimate chlorophyll content, and provide an effective tool for nondestructive monitoring of chlorophyll content in H. hainanensis.

    Method 

    By combining vegetation index with traditional threshold segmentation methods, the background of multispectral images of H. hainanensis was removed, and the optimal segmentation method was determined by F1 as the segmentation accuracy evaluation index. Then, based on the segmented multispectral image, spatial domain features (vegetation index and texture features) were extracted, and three frequency domain features were introduced. The measured value of relative chlorophyll content (SPAD value) was measured using a portable chlorophyll analyzer SPAD. And based on correlation analysis and Lasso algorithm, image features were filtered to determine the preferred features which were strongly correlated with SPAD value of H. hainanensis. Finally, based on partial least-squares regression (PLSR), random forest (RF) and XGBoost algorithm, multispectral spatial domain, frequency domain and fusion feature models were established, and precision verification was conducted to determine the optimal model form for SPAD value estimation of young H. hainanensis.

    Result 

    The segmentation method combining DVI and Kapur threshold achieved the highest segmentation accuracy, with F1 of 0.917. Therefore, it was the most suitable segmentation method for H. hainanensis canopy multispectral images. Many spatial and frequency domain features of multispectral images exhibited significant correlations with the SPAD values of H. hainanensis. The most correlated feature was the modified chlorophyll absorption reflectivity index, with a correlation coefficient of −0.780. It was the preferred feature for estimating SPAD values based on single image features. Among the three frequency domain features, the correlation performance of wavelet features was the best. Therefore, wavelet transform was the preferred frequency domain transformation method for slope barrier multispectral images. The SPAD value estimation models constructed with different image features were sorted by performance as single frequency domain feature model < single spatial domain feature model < fused feature model, and the corresponding optimal modeling algorithms were RF and XGBoost, respectively. The fusion feature model based on RF was the optimal model, with a test R2 of 0.791, which was 7.9% higher than the test R2 of a single spatial feature model.

    Conclusion 

    The estimation accuracy of H. hainanensis SPAD values can be improved by introducing three frequency domain features, and the fusion feature model based on RF can achieve good estimation accuracy. Therefore, integrating multispectral spatial and frequency domain features with machine learning algorithms can be used as an effective tool for estimating the relative chlorophyll content of young H. hainanensis, which is conducive to the intelligent development of H. hainanensis cultivation and management.

  • 2017年9月23日中共中央国务院批复了《北京城市总体规划(2016年—2035年)》,要求“优化城市功能和空间布局,坚持疏解整治促提升,坚决拆除违法建筑,加强对疏解腾退空间的利用和引导,腾笼换鸟、留白增绿”。提出“疏解北京非首都功能,改善生态环境,建设和谐宜居之都”“健全市域绿色空间体系,建设森林城市,让森林进入城市的目标”。森林城市建设要求重视生物多样性保护,满足居民休闲游憩需求,传播生态文化知识。从20世纪60年代中期开始,国外一些林学家,提出在市区和郊区发展城市森林,把森林引入城市[1]的理念。结合国情,我国学者也从多学科多角度对城市森林进行概念定义、实践探究。关注的重点也从城市森林林木本身扩大到城市森林协调城市生态环境、维持生态系统良性循环等方面。作为城市生态系统的重要组成部分, 城市森林在维护碳氧平衡、净化环境、调节气候、涵养水源和保护生物多样性等方面具有重要作用[2-4]。大力发展城市森林,使城市与森林和谐共存,已经成为新世纪世界生态城市的发展方向[5]

    《北京城市总体规划(2016年—2035年)》通过后,北京市将拆违腾退土地用于“留白增绿”,作为绿色空间为城市提供生态服务,增加绿色游憩场所。计划拆除违法建设4 000万m2以上,腾退占地3 974 hm2,其中用于“留白增绿”1 986 hm2[6]。本研究定义“留白增绿”背景下的城市森林概念为:在城市地域范围内拆违腾退土地上,结合地块尺度、周边用地类型、总体规划定位三大要素所建立的能够针对性发挥城市森林生态、生活、生产功能,符合城市森林景观风貌的林木群落及其环境所构成的森林生态系统。如何在“留白增绿”的背景下充分利用拆违腾退土地,建设与城市发展背景相契合的城市森林成为新的议题。

    本研究对2 039个,共计1 951.77 hm2的“留白增绿”地块进行梳理统计,地块面积从数平方米到数十公顷不等;空间区位遍布北京各个市辖区;地块类型涵盖公园绿地、防护绿地、生产绿地、农林用地、其他非建设用地等;部分地块位于总体规划中市域绿色空间的城市公园环、郊野公园环、环首都森林湿地公园环、历史文化名城保护区、文化保护带、市级通风廊道等;周边用地类型各异,涵盖居住用地、商业服务设施用地、物流仓储用地等。

    城市森林的类型因划分标准的不同而呈现多样化,本文以城市森林主要功能为标准对城市森林类型进行探讨。以往的研究表明,从降低空气细菌含量方面考虑,绿地面积大于0.09 hm2才能形成“森林内环境”,为居民提供良好的游憩地[7];鸟类栖息地面积范围至少要大于1.5 hm2,且在1.5~30 hm2之间都具有较高的生态效益比[8],10~35 hm2的公园可为大部分鸟类提供家园[9]。李延明等[10]的研究表明,规模大于3 hm2且绿化覆盖率达到60%以上的集中绿地, 能够发挥较好的缓解城市热岛效应的功能;而任志彬[11]的研究表明城市森林斑块面积对热环境效应作用效率最大的区间为2.8~25 hm2。本文基于以上研究结论,对2 039个“留白增绿”地块尺度进行统计(图 1),其中0.09~1.5 hm2范围的地块比例高达49.5%,说明大部分“留白增绿”地块能够起到提供良好游憩服务的功能,1.5 hm2以上的地块占比15.8%,能够发挥居民游憩、鸟类栖息、缓解热岛效应等不同功能。

    图  1  “留白增绿”不同尺度地块的数量
    Figure  1.  Classification of different scales of "leave blank apace and increase green space"

    结合不同尺度城市森林主导功能、场地周边用地类型、总体规划定位三大要素对城市森林功能类型进行探讨。在划分生态、生活、生产3个一级分类的基础上,提出栖息生境型、环境调节型、景观游憩型、科普教育型、康养服务型、设施防护型、经济生产型共7个二级分类(图 2)。

    图  2  “留白增绿”背景下的城市森林功能类型划分
    Figure  2.  Urban forest function type division under the background of "leave blank space and increase green space"

    城市化进程迅猛推进后,建设城市森林成为协调人与自然关系、发展生态城市的重要手段。然而,城市森林建设受到诸多因素影响,如何确定城市森林主导功能类型成为一大难点。因此对“留白增绿”背景下的城市森林功能类型进行探讨,以期为城市森林营建提供目标指引。

    《北京城市总体规划(2016年—2035年)》提出建设森林环绕的生态城市,提升生态空间品质;要求加强城市生态建设,重视生物多样性保护。城市化进程造成了城市生物多样性锐减,重视和保护生物多样性成为城市生态化发展的重点。因此,本文主要对以提高物种多样性为主要目标的栖息生境型城市森林进行营建策略研究。

    生境又称栖息地,是指生物的居住场所,即生物个体、种群或群落能在其中完成生命过程的空间[12]。本文所提到的栖息生境是指动物发生取食、繁殖、筑巢等行为的主要生存环境。多样化的生境包含更多适宜物种生存的条件[13],生境多样性的营建是实现城市生物多样性再生的关键技术之一[14]

    为保证生态系统的良性循环,需要考虑构建完整的食物链。城市人工生境可分为人工陆地生境和人工水体生境两大类。人工生境营造中主要针对的生物类型有鱼类、底栖类、两栖类、爬行类、鸟类、哺乳类等。在栖息生境型城市森林营建中,应考虑在以上生物类型中选择目标类群,并针对其中的目标物种营建栖息生境。

    鱼类、底栖类、两栖类、爬行类、哺乳类对生境的需求相对较为简单,在有一定绿化面积、一定浅水区域、植物群落结构丰富、土壤有机质充足的环境中能够普遍适应。鸟类处于城市生态系统的顶端[15],其种类及数量在一定程度上决定了其他等级生物的生存状态,在一定程度上代表了生态系统的物种多样性及生态系统的健康程度[16-17],同时鸟类可以作为环境监测的指标[18],是城市生态系统综合质量的指示生物类群[19]。因此,本研究以鸟类为主要目标物种,从林地生境、水体生境两种主要生境类型提出栖息生境型城市森林营建策略。

    植物多样性是影响鸟类多样性的重要生态因子[20],同时植被也是影响城市森林景观风貌的重要自然因素,乡土植物能够充分适应当地环境和气候条件,体现地区植物区系景观风貌,创造“景观文化”的本土化[21]。乡土植物的应用有助于奠定城市森林景观风貌的整体基调。《国家森林城市评价指标:LY/T2004—004林》中提出,乡土树种应占城市绿化树种使用数量的80%以上。北京五环内记录到的包含279种乡土植物在内的共536种植物[22]及北京市《乡土植物资源发展名录》推荐的82种乔灌草植被为提高城市森林乡土植物比例和奠定森林景观风貌的基础提供了参考。结合国家森林城市评价指标,本研究认为在北京市城市森林营造中应保持乡土植物比例在80%以上。

    空间层次越复杂、植物群落越丰富,提供给鸟类的栖息场所和食物越多[23]。通过改进植物群落的树种组成和结构可以有效提高城市绿地鸟类的多样性水平[24]。乔木层和灌木层的结构和组成对栖息地质量起到决定性作用[25]。大面积多树种的阔叶林,可提高鸟类繁殖群落的多样性[26]。植被结构的中、下层绿化,以丰富的垂直结构增加了环境次级结构的多样性[27],为鸟类提供了更多繁殖、摄食和隐蔽的空间[26]。陈自新等[28]对北京市绿地生态效益调研分析后指出,北京市绿地中所用乔木(株)、灌木(株)、草本(m2)、绿地(m2)最适合的种植比例不少于1:6:21:29。本研究认为在栖息生境型城市森林营造中,应在上述结论的基础上,适当增加阔叶林、灌草层植物的比例,控制每29 m2绿地上乔木(株)、灌木(株)、草本(m2)比例在1:6:21以上。

    植被郁闭度可以作为衡量栖息地的重要指标[29]。不同郁闭度的乔、灌、草植物群落保证了植物在单元空间中的互相作用,更加充分的吸收利用自然资源。完全郁闭的树林,阻碍鸟类的飞行出入,因此成林时郁闭度应小于90%[30],纯针叶林郁闭度应小于85%[23]。本研究认为城市森林应控制整体林地植物群落郁闭度在70%~85%左右,为鸟类提供更适宜的生活空间。结合黄越[31]对北京鸟类生境类型的调查研究,本文提出10种不同郁闭度的二级生境类型(表 1),对应不同乔灌草搭配比例及高度。除草地外,其他9种二级生境可结合常绿、落叶、混交构成,形成27种三级生境类型。共形成28种不同郁闭度三级生境类型。

    表  1  不同郁闭度二级生境类型
    Table  1.  Secondary habitat types with different canopy closure
    一级类型
    Primary classification
    二级类型
    Secondary classification
    说明
    Description
    乔木林
    Tree forest
    疏林草地
    Sparse forest-grass land
    无灌木或灌木覆盖度小于15%,乔木覆盖度在30%~70%之间
    The shrub-free or shrub coverage is less than 15%, and the tree coverage is between 30% and 70%
    密林草地
    Jungle forest-grass land
    无灌木或者灌木覆盖度小于15%,乔木覆盖度在70%~90%之间
    The shrub-free or shrub coverage is less than 15%, and the tree coverage is between 70% and 90%
    疏林灌丛
    Sparse forest-shrubbery
    灌木覆盖度在30%~70%之间,乔木覆盖度在30%~70%之间
    Shrub coverage is between 30% and 70%, and the tree coverage is between 30% and 70%
    密林灌丛
    Jungle forest-shrubbery
    灌木覆盖度在30%~70%之间,乔木覆盖度在70%~90%之间
    Shrub coverage is between 30% and 70%, and the tree coverage is between 70% and 90%
    疏林灌草地
    Sparse forest-shrubbery-grass land
    灌木覆盖度在15%~30%之间,乔木覆盖度在30%~70%之间
    Shrub coverage is between 15% and 30%,and the tree coverage is between 30% and 70%
    密林灌草地
    Jungle forest-shrubbery-grass land
    灌木覆盖度在15%~30%之间,乔木覆盖度在70%~90%之间
    Shrub coverage is between 15% and 30%, and the tree coverage is between 70% and 90%
    复层林
    Multiple layer forest
    灌木覆盖度大于15%,乔木覆盖度大于30%
    Shrub coverage is more than 15% and tree coverage is more than 30%
    灌丛
    Shrubbery
    矮灌丛
    Short shrubbery
    灌丛高度小于1 m,灌木覆盖度大于30%,乔木覆盖度小于30%
    Shrub height is less than 1 m, coverage is more than 30%, and tree coverage is less than 30%
    灌丛
    Shrubbery
    灌丛高度大于1 m,独干灌木高度小于1.5 m,灌木覆盖度大于30%,乔木覆盖度小于30%
    The shrub higher than 1 m, the height of branch independent plant is less than 1.5 m, the shrub coverage is more than 30%, and the tree coverage is less than 30%
    草地Lawn 草地
    Lawn
    以草本植物为主,木本植物覆盖度小于15%
    Mainly herbaceous, woody plant coverage is less than 15%
    下载: 导出CSV 
    | 显示表格

    食源树种的种类和数量对鸟类物种多样性产生影响[32]。木本植物挂果期长短与鸟类多度存在显著相关性[20]。乡土树种果实成熟期与鸟类繁殖期或迁徙期基本一致,本地鸟类经过长期的自然选择,更倾向于在乡土树种上取食、栖息和停留[33]。在栖息生境型城市森林营造中应考虑增加能够为鸟类提供食物(如种子、花芽、果实和浆果等)的树种[34],并保证四季均衡供应[32]及食源树种在生境范围内的均匀分布。通过对前人研究的总结,发现因调查样地尺度各异、环境条件复杂多变、植物配置情况多样、鸟类取食喜好多样,难以对最优食源树种组团类型及比例进行量化。故本研究仅提出增加食源树种比例、兼顾食源树种季相搭配及均衡分布的策略,而不对具体比例进行限定。

    增加常绿乔木的比例能为鸟类提供良好的栖息场所,提高鸟类多样性[35],同时常绿针叶植物能够在冬季为鸟类提供一定的食源。根据李淑凤[36]对北京市公园绿地植物配置的研究,常绿树一般占落叶树的30%~40%即可,即常绿落叶比例约为23:77~29:71。在栖息生境型城市森林营造中,为保证冬季鸟类觅食及隐蔽需求,将栖息生境城市森林中常绿落叶植物比例提高到4:6,为鸟类创造更适宜的栖息条件。

    水域生态的复杂多样影响鸟类群落多样性[37]。为增加环境多样性,在保持原有水域的基础上适当增加各种形式的水体,以满足鸟类的生存需要[27]。通过曲折蜿蜒的岸线,提供港湾、浅滩、半岛等栖息条件,营造有近圆形核心、弯曲边界和边缘指状突起的水系形态[38](图 3)。创造丰富栖息生境的同时增加物种传播的可能。

    图  3  水系形态示意图
    Figure  3.  Schematic of water system

    在水域面积大于5 hm2时,建议在水中设置岛屿,供鸟类栖息[31]。研究表明,随着岛屿面积增大,单位面积内岸线周长迅速减小,当岛屿面积超过3 000 m2时,岸线变化趋于平缓[38](图 4)。创造多种生境类型,需控制岸线与岛屿面积比例。因此,在栖息生境型城市森林营造中应控制单个岛屿面积在3 000 m2以下,且有大小变化。控制岛屿与水体岸线距离在10 m以上[39]

    图  4  岛屿周长面积比与岛屿面积关系
    此图改绘自参考文献[38]。
    Figure  4.  Relationship between the ratio of circumference to area of an island with island area
    This figure is redrawn from reference[38].

    不同种水鸟在觅食和营巢上对水深有着不同的需求,涉禽和游禽对水深条件要求相对较高。在选定以涉禽和游禽为水体生境主要目标物种的前提下,认为北京地区更适合通过低水位营造野生动物栖息地[40],水深通常应不大于1 m[41]。在水域范围内,分别营造0~10 cm、10~20 cm、20~30 cm的水深环境满足涉禽生境需求,营造30~100 cm、100 cm以上水深满足游禽生境需求,同时控制整体水深不超过6 m(图 5),并根据选定目标种中游禽、涉禽所占比例,分配不同水深的水体面积。

    图  5  水体深度示意图
    Figure  5.  Schematic diagram of water depth

    水生植物的多样性对鸟类群落的影响较大[37]。结合前人的研究,不同鸟类对水生植物覆盖度有不同要求,但普遍低于70%。在栖息生境型城市森林营造中,保留至少1/3的水面供鸟类起飞或降落[42],2/3的水面分别以0%~30%、30%~50%、50%~70%的覆盖度种植水生植物(图 6),并根据不同物种的需求,构建不同覆盖度水体环境,保证生境类型的多样性。

    图  6  不同水生植物覆盖度示意图
    Figure  6.  Schematic representation of aquatic plants with different coverage

    硬质驳岸及硬化池底极不利于水生生物的生存,影响部分肉食性水鸟的取食。在栖息生境型城市森林营造中,为便于鸟类觅食、隐藏等活动,驳岸坡度应尽量控制在10:1或更小,营造一定的裸露滩涂和砂石驳岸[41]。以透气性较好的煤渣垫层、有机介质等铺设池底,使土壤渗透率保持在0.025~0.35 cm/h[38]。在保证一定面积水域常水位水深的基础上,充分利用自然降水变化形成季节性水体环境(图 7)。

    图  7  驳岸示意图
    Figure  7.  Schematic diagram of the revetment

    本文以“留白增绿”地块中位于东南五环与京沪高速交汇处的横街子地块为研究对象(图 8)。横街子地块由3块“留白增绿”地块组成,总面积62.6 hm2,处于北京市第一道绿化隔离带范围内。场地内部有较好的水源条件,因拆迁所产生的约9.2 hm2低地可以作为水系营造使用。李湛东等[43]的研究结果表明,横街子地块及周边区域植物多样性较低。周边绿地数量较多、功能较为完善,能满足游人休闲游憩的需求。结合前文对城市森林功能类型的探讨,认为横街子地块适宜营建以提高生物多样性为主要目标的栖息生境型城市森林。

    图  8  研究对象区位
    Figure  8.  Location of the research object

    以前文的营建策略为基础,在横街子栖息生境型城市森林营造中,以鸟类作为生境营建的主要目标类群。根据黄越[31]对北京市圆明园、中山公园等26个城市公园鸟类生境的调查研究,林鸟中红嘴蓝鹊(Urocissa erythroryncha)、金翅雀(Carduelis sinica)和斑鸫(Turdus eunomus)可以作为针叶林生境代表物种;白头鹎(Pycnonotus sinensis)、四声杜鹃(Cuculus micropterus)、大杜鹃(Cuculus canorus)、丝光椋鸟(Sturnus sericeus)、黄眉柳莺(Phylloscopus inornatus)、燕雀(Fringilla montifringilla)可以作为阔叶林生境代表种;红喉姬鹟(Ficedula parva)、褐柳莺(Phylloscopus fuscatus)和黄喉鹀(Emberiza elegans)可以作为灌丛生境的代表种;水鸟中苍鹭(Ardea cinerea)、草鹭(Ardea purpurea)、大白鹭(Ardea alba)可以作为不同营巢习性的代表种;黑水鸡(Gallinula chloropus)、鸳鸯(Aix galericulata)、小鸊鷉(Tachybaptus ruficollis)可以分别作为水岸、近水及深水的代表种;近水鸟普通翠鸟是溪流状近水生境的代表种。在横街子栖息生境型城市森林营造中以上述代表物种作为目标种营建栖息生境。

    横街子地块紧邻东南五环与京沪高速,有较大的噪声干扰。研究表明针阔混交林对实时交通噪声有较好的衰减作用[44],其中阔叶乔木降噪能力优于针叶乔木[45]。乔灌层宽度对噪音衰减值的影响最为显著,在30 m内随着林带宽度的增加, 降噪效果更加明显[46]。为避免快速交通对城市森林内部的干扰,在场地临近快速路一侧设置30 m针阔混交林并搭配灌草层作为隔离带。在场地内部营建生境核心斑块(图 9)。

    图  9  总平面图
    Figure  9.  Design plan

    参照前文栖息生境型城市森林营建策略,在林地生境营造中,以北京市乡土植物为主要树种结合北京市《乡土植物资源发展名录》推荐的82种乔灌草植被作为横街子城市森林植物选择名录,搭配食源树种,控制不同类型植物群落郁闭度(图 10),控制常绿落叶比例约为4:6(图 11),形成栖息生境型城市森林基底。在此基础上依据林地生境目标鸟类的生活习性营建栖息生境(表 2)。

    图  10  不同郁闭度群落分布示意图
    Figure  10.  Schematic diagram of the distribution of plant communities with different canopy closure
    图  11  常绿及落叶植物分布示意图
    Figure  11.  Schematic representation of evergreen and deciduous plant distribution
    表  2  林地鸟类生境营造
    Table  2.  Woodland bird habitat construction conditions
    林地类型
    Forest land type
    鸟的种类
    Species of birds
    生境需求
    Habitat requirement
    针叶林
    Coniferous forest
    金翅雀Carduelis sinica、斑鸫Turdus eunomus 保证上层林木有一定盖度且生境下层较开敞, 草本植物丰富[49]
    Ensure that the upper trees have a certain degree of coverage and the lower layer of the habitat is open with rich herbs[49]
    红嘴蓝鹊Urocissa erythroryncha 连续常绿林面积至少4 hm2[31]
    Continuous evergreen forest area is of at least 4 ha[31]
    阔叶林
    Broadleaved forest habitat
    燕雀Fringilla montifringilla 拾取草籽为食,喜上层乔木主干清晰,下层灌木稀疏的群落[47]
    Picking up grass seeds for food, like the environment with clear branches of upper trees and lower sparse shrub community[47]
    白头鹎Pycnonotus sinensis、丝光椋鸟Sturnus sericeus 喜主干清晰下层开阔的阔叶乔木群落,灌木下层稀疏的植物或体量小的球形灌木[49]。距地面大多2~3 m筑巢,亦有筑在6~6.5 m高大乔木上[30]
    Like the environment with clear branches and broadleaved trees, lower sparse shrub community or small spherical shrub[49]. Nesting on places 2-3 m higher than the ground, and nesting on 6-6.5 m tall trees[30]
    四声杜鹃Cuculus micropterus、大杜鹃Cuculus canorus 喜乔木主干通直,下层相对开阔,以阔叶林为主。搭配树龄较长的针叶树。喜既不影响飞行又能覆盖地表的丰富小灌木及草本空间[47]
    Like trees with straight trunk, the lower layer is relatively open, mainly broadleaved forest. Pair with longer-aged conifers. Like rich shrubs and herbaceous spaces cover the surface but do not affect flight activities[47]
    黄眉柳莺Phylloscopus inornatus 喜株型挺拔整齐的高大乔木, 偏好落叶林和常绿林,如落叶密林草地和常绿密林灌丛草地[31]
    Like straight and neat trees, preference deciduous forest and evergreen forest, such as deciduous jungle grassland and evergreen jungle shrub grassland[31]
    灌丛Shrub 黄喉鹀Emberiza elegans、红喉姬鹟Ficedula parva、褐柳莺Phylloscopus fuscatus 喜中等灌木和低矮的自然地被[30],常活动在林缘以及溪流沿岸的疏林与灌丛[48]
    Like medium shrubs and low herbs[30], of ten active in forests and shrubs along the forest margins and streamlined[48]
    下载: 导出CSV 
    | 显示表格

    根据场地基础条件,场地内部有约9.2 hm2水面能够营造水体生境。通过曲折蜿蜒的岸线,提供港湾、浅滩、半岛等栖息条件,并营造多个小型岛屿(图 12)。在此基础上依据水体生境目标鸟类的生活习性设计水体深度及水生植被覆盖率(表 3)。

    图  12  水体生境示意图
    Figure  12.  Schematic representation of evergreen and deciduous plants distribution
    表  3  水体鸟类生境营造
    Table  3.  Aquatic habitat construction conditions
    类别
    Category

    Family

    Species
    生境需求
    Habitat requirement
    水鸟
    Water birds
    鹭科
    Ardeidae
    大白鹭
    Ardea alba
    喜在近水域或水中岛屿高树上营巢[49]。水域内需沉水、挺水植物覆盖率40%~60%之间,高于1 m的植物占60%左右,木本植物、挺水植物4 m以内越高越好。陆地植物7 m内越高越适宜[30]
    Like to nest on high trees in the water or on the island[49].The coverage of aquatic plants in the waters is between 40% and 60%, plants above 1 m is about 60%. Woody plants and aquatic plants should be within 4 m and plants on land should within 7 m, the higher, the better [30]
    水鸟
    Water birds
    鹭科
    Ardeidae
    草鹭
    Ardea purpurea
    喜栖息在水边灌丛或芦苇(Phragmites communis)沼泽[49]。水域内需沉水、挺水植物覆盖率40%~60%之间,高于1 m的植物占60%左右,木本植物、挺水植物4 m以内越高越好。陆地植物7 m内越高越适宜[30]
    Like to inhabit in the shrubs or the reeds.The coverage of aquatic plants in the waters is between 40% and 60%, plants above 1m is about 60%. Woody plants and aquatic plants should be within 4 m and plants on land should within 7 m, the higher, the better [30]
    水鸟
    Water birds
    鹭科
    Ardeidae
    苍鹭
    Ardea cinerea
    喜栖息于有大片芦苇和水生植物的浅水域[49]。水域内需沉水、挺水植物覆盖率40%~60%之间,高于1 m的植物占60%左右,木本植物、挺水植物4 m以内越高越好。陆地植物7 m内越高越适宜[30]
    Like to inhabit in shallow waters with large reeds and aquatic plants.The coverage of aquatic plants in the waters is between 40% and 60%, plants above 1 m is about 60%. Woody plants and aquatic plants should be within 4m and plants on land should within 7 m, the higher, the better [30]
    水鸟
    Water birds
    鸭科
    Anatidae
    黑水鸡
    Gallinula chloropus
    栖息于富有树木、芦苇和水生挺水植物遮蔽的淡水水域,不喜欢很开阔的场所[49]。栖息地水域植被覆盖率50%~75%为宜,其中30%~50%灌木,40%~70%挺水植物,0%~10%乔木,以及25%水面,水深0.9 m以内水域中宜有倒伏树干[30]
    Inhabit in freshwater covered with trees, reeds and aquatic plants, do not like very open space[49]. The vegetation coverage 50%-75%, with 30%-50% shrubs, 40%-70% aquatic plants, 0%-10% trees, and 25% water surface. There should be fallen tree trunks in the water within 0.9 m[30]
    水鸟
    Water birds
    鸭科
    Anatidae
    鸳鸯
    Aix galericulata
    喜在水深不超过2 m的浅水区觅食,水域植被覆盖率50%~75%为宜,近水区域结合少量乔灌,以木本上层结构为主[30]
    Like to eat in shallow waters with a water depth no more than 2 m, vegetation coverage in the waters should is 50%-75%, near-water area combined with a small amount of trees and shrubs, mainly based on woody superstructure[30]
    水鸟
    Water birds
    鸭科
    Anatidae
    小鸊鷉Tachybaptus ruficollis 在芦苇、香蒲(Typha orientalis Presl)、灯芯草(Juncus effusus L.)等挺水植被区域活动,挺水植物占整个水面比例小于30%,栖息地生态岛上高度0.6~10 m的植被覆盖率50%以上为宜,距离人类活动不宜小于400 m[30]
    Activities in the watery vegetation areas such as reeds, cattails, and rushes. Vegetation coverage is less than 30%. The coverage of trees with a height of 0.6-10 m on the habitat ecological island is more than 50%. Human activities should be outside the distance of 400 m[30]
    近水鸟
    Near- water birds
    翠鸟科
    Alcedinidae
    普通翠鸟Alcedo atthis 栖息于有灌丛或疏林、水清澈而缓流的小河、溪涧、湖泊以及灌溉渠等水域[49]。岸际15 m以内浮水植物、原木、岩石等阻碍越少越好[30]
    Inhabited in rivers, streams, lakes and irrigation canals with shrubs or sparse forests, clear waters and slow streams[49]. The less obstructions such as floating plants, logs and rocks within 15 m of the coast, the better[30]
    下载: 导出CSV 
    | 显示表格

    城市森林功能的多样性、城市环境条件的异质性决定了未来城市森林建设模式的多样性。本文将“留白增绿”背景下的城市森林定义为在城市地域范围内拆违腾退土地上,结合地块尺度、周边用地类型、总体规划定位三大要素所建立的能够针对性发挥城市森林生态、生活、生产功能,符合城市森林景观风貌的林木群落及其环境所构成的森林生态系统。根据“留白增绿”地块尺度、周边用地类型及总体规划定位要求对可能营造的城市森林功能类型进行探讨,在生态型、生活型、生产型3个一级分类的基础上,提出栖息生境型、环境调节型、景观游憩型、科普教育型、康养服务型、设施防护型、经济生产型共7个二级分类,以期为城市森林营建提供目标指引。多种功能类型的城市森林将成为城市生态环境的重要组成部分。

    在此基础上,基于“留白增绿”的城市发展背景,选择目标物种,依据目标物种所需的主要生境类型,提出栖息生境型城市森林营建策略。主要分为林地生境营建策略、水体生境营建策略。林地生境营建策略包括增加乡土树种比例、调整植物群落结构、控制植物群落郁闭度、增加食源植物比例、增加常绿乔木比例;水体生境营建策略包括丰富水体形式、控制水体深度、控制水生植物覆盖度、改良驳岸类型。以横街子栖息生境型城市森林为例,根据场地内部条件,选择目标物种,结合上述林地生境、水体生境营建策略,为目标物种营造适宜栖息的生境条件。

    但需要指出的是,本文是基于提高生物多样性为目标的栖息生境型城市森林所进行的营建策略研究,其建设策略更多地是以适宜生物栖息为原则提出的。营建城市森林是落实城市生态文明建设的重要途径,但城市森林功能并不唯一,其环境调节、生活游憩、生产防护等功能也是城市森林营建时需要考虑和重视的。因此,针对不同的场地条件,我们需要全方位考虑并择优选择和建立不同功能类型的城市森林。

  • 图  1   冠层多光谱图像分割

    Figure  1.   Multispectral image segmentation

    图  2   基于多光谱图像特征的λ迭代过程

    MSE为均方误差;λ为调整参数。MSE represents mean square error. λ represents tuning parameter.

    Figure  2.   Iterative process of λ based on multispectral image features

    图  3   基于多光谱空域特征的建模和检验结果

    Figure  3.   Modeling and verification results based on multispectral spatial domain features

    图  4   基于多光谱频域特征的建模和检验结果

    Figure  4.   Modeling and verification results based on multispectral frequency domain features

    图  5   基于多光谱融合特征的建模和检验结果

    Figure  5.   Modeling and verification results based on multispectral fusion features

    表  1   植被指数及其计算公式

    Table  1   Vegetation index and its calculation formula

    植被指数
    Vegetation index
    计算式
    Calculation formula
    植被指数
    Vegetation index
    计算式
    Calculation formula
    归一化红边绿指数
    Normalized difference
    red edge-green index (NDIreg)[14]
    NDIreg = (RRERG)/(RRE + RG) 红边植被指数
    Red-edge vegetation index (REVI)[17]
    REVI = RNIR/RRE – 1
    归一化红边红指数
    Normalized difference
    red edge-red index (NDIrer)[14]
    NDIrer =(RRERR)/(RRE + RR) 修正叶绿素吸收反射率植被指数
    Modified chlorophyll absorption
    reflectivity index (MCARI)[18]
    MCARI =
    [RRERR – 0.2(RRERG)] RRE/RR
    增强型植被指数
    Enhanced vegetation
    index (EVI)[15]
    EVI = 2.5(RNIRRR)/(RNIR + 6RR – 7.5RB + 1) 改进简单比值植被指数
    Modified simple ratio vegetation
    index (MSR)[19]
    MSR=(RNIR/RR1)/RNIR/RR+1
    重归一化差异指数
    Renormalized difference
    vegetation index (RDVI)[16]
    RDVI=(RNIRRR)/RNIR+RR 修正植被指数
    Modified vegetation index (MVI)[20]
    MVI=(RNIRRR)/(RNIR+RR)+0.5
    注: RRRGRBRNIRRRE分别表示红、绿、蓝、近红外、红边波段反射率。 Note: RR, RG, RB, RNIR, RRE indicate reflectivity of red, green, blue, near infrared, and red edge bands, respectively.
    下载: 导出CSV

    表  2   频域特征

    Table  2   Frequency domain features

    方法
    Method
    特征
    Feature
    命名
    Naming
    特征
    Feature
    命名
    Naming
    特征
    Feature
    命名
    Naming
    特征
    Feature
    命名
    Naming
    傅里叶变换
    Fourier transform
    能量–均值
    Energy-mean
    Bj-EnmFFT 熵–均值
    Entropy-mean
    Bj-EntmFFT 惯性矩–均值
    MOI-mean
    Bj-MOImFFT 相关–均值
    Correlation-mean
    Bj-CormFFT
    能量–标准差
    Energy-SD
    Bj-EnsFFT 熵–标准差
    Entropy-SD
    Bj-EntsFFT 惯性矩–标准差
    MOI-SD
    Bj-MOIsFFT 相关–标准差
    Correlation-SD
    Bj-CorsFFT
    小波变换
    Wavelet transform
    能量–均值
    Energy-mean
    Bj-EnmWT 熵–均值
    Entropy-mean
    Bj-EntmWT 惯性矩–均值
    MOI-mean
    Bj-MOImWT 相关–均值
    Correlation-mean
    Bj-CormWT
    能量–标准差
    Energy-SD
    Bj-EnsWT 熵–标准差
    Entropy-SD
    Bj-EntsWT 惯性矩–标准差
    MOI-SD
    Bj-MOIsWT 相关–标准差
    Correlation-SD
    Bj-CorsWT
    里斯变换
    Riesz transform
    能量–均值
    Energy-mean
    Bj-EnmR 熵–均值
    Entropy-mean
    Bj-EntmR 惯性矩–均值
    MOI-mean
    Bj-MOImR 相关–均值
    Correlation-mean
    Bj-CormR
    能量–标准差
    Energy-SD
    Bj-EnsR 熵–标准差
    Entropy-SD
    Bj-EntsR 惯性矩–标准差
    MOI-SD
    Bj-MOIsR 相关–标准差
    Correlation-SD
    Bj-CorsR
    注:j = {1, 2, 3, 4, 5},对应的Bj分别为R、G、B、NIR、RE,即红、绿、蓝、近红外、红边波段。Notes: j = {1, 2, 3, 4, 5}, Bj indicates R, G, B, NIR, RE, namely, red, green, blue, near infrared, and red edge bands.
    下载: 导出CSV

    表  3   坡垒多光谱图像分割结果评价

    Table  3   Evaluation of segmentation results for slope barrier multispectral images

    分割精度指标
    Segmentation accuracy indicator
    RVI + Kapur DVI + Kapur NDVI + Kapur
    精确率 Accuracy (P) 0.816 0.883 0.863
    召回率 Recall (R) 0.980 0.952 0.971
    F1 0.890 0.917 0.914
    下载: 导出CSV

    表  4   多光谱空域特征的描述统计和相关系数

    Table  4   Descriptive statistics and correlation coefficients of multispectral spatial domain features

    特征类别
    Feature category
    特征
    Feature
    均值 ± 标准差
    Mean ± SD
    相关系数
    Correlation coefficient (r
    特征
    Feature
    均值 ± 标准差
    Mean ± SD
    r
    植被指数
    Vegetation index
    MCARI0.58 ± 0.30−0.780***NIR0.48 ± 0.11−0.588***
    NDIrer0.41 ± 0.10−0.662***RDVI0.38 ± 0.10−0.555***
    RE0.41 ± 0.11−0.658***NDVI30.52 ± 0.11−0.495***
    DVI30.33 ± 0.10−0.632***RVI33.33 ± 0.95−0.477***
    DVI10.31 ± 0.10−0.615***MVI0.99 ± 0.06−0.400***
    纹理特征
    Texture feature
    B-Entm0.06 ± 0.050.427***B-MOIs0.01 ± 0.010.349***
    R-Entm0.00 ± 0.000.416***RE-Corm0.12 ± 0.03−0.347***
    G-Cors0.00 ± 0.000.403***NIR-Ens4.64 ± 1.150.347**
    B-MOIm0.00 ± 0.000.400***NIR-Corm4.70 ± 1.170.340***
    B-Corm0.01 ± 0.010.375***RE-Ents4.72 ± 1.180.340***
    注:*表示图像特征与SPAD值在0.05水平上显著相关,**表示二者在0.01水平上显著相关,***表示二者在0.001水平上显著相关。下同。Notes: * indicates significant correlation between image features and SPAD values at 0.05 level, ** indicates significant correlation at 0.01 level, *** indicates significant correlation at 0.001 level. The same below.
    下载: 导出CSV

    表  5   多光谱频域特征的描述统计和相关系数

    Table  5   Descriptive statistics and correlation coefficients of multispectral frequency domain features

    特征类别
    Feature category
    特征
    Feature
    均值 ± 标准差
    Mean ± SD
    r特征
    Feature
    均值 ± 标准差
    Mean ± SD
    r
    傅里叶特征
    Fourier feature
    G-CorsFFT0.05 ± 0.03−0.210*
    小波特征
    Wavelet feature
    G-EnmWT0.92 ± 0.08−0.401***RE-EnmWT0.92 ± 0.07−0.337***
    G-CorsWT2.10 ± 1.57−0.378***B-EntmWT0.18 ± 0.110.326***
    G-EnsWT0.30 ± 0.260.367***G-EntmWT0.19 ± 0.160.325***
    B-CorsWT2.46 ± 1.56−0.357***NIR-EnmWT0.92 ± 0.07−0.309***
    B-EnsWT0.28 ± 0.200.341***B-EnmWT0.93 ± 0.06−0.308***
    里斯特征
    Riesz feature
    B-CorsR0.01 ± 0.01−0.415***R-EnmR0.16 ± 0.06−0.288**
    B-EnmR0.18 ± 0.06−0.373***B-MOIsR0.04 ± 0.020.277**
    RE-MOImR0.01 ± 0.00−0.346***NIR-EnsR2.46 ± 0.390.277**
    NIR-EnmR0.20 ± 0.07−0.301***G-EnmR0.15 ± 0.04−0.276**
    RE-EnmR0.17 ± 0.06−0.300***RE-EnsR2.58 ± 0.360.270**
    下载: 导出CSV

    表  6   筛选后多光谱图像特征的应用方差膨胀因子(VIF)值

    Table  6   Variance inflation factor (VIF) values of filtered multispectral image features

    特征类别
    Feature category
    特征
    Feature
    VIF特征
    Feature
    VIF特征
    Feature
    VIF
    多光谱空域特征
    Multispectral spatial domain feature
    RE0.705NDIrer0.229B-Entm0.729
    RVI40.609MCARI2.941RE-Ens1.243
    DVI32.129B-Ens2.947RE-Corm0.249
    多光谱频域特征
    Multispectral frequency domain feature
    B-EnmWT0.042G-EnmWT0.017B-MOIsR0.012
    B-EnsWT0.042NIR-EnmWT0.013B-CorsR0.013
    B-CorsWT0.011RE-EnmWT0.016RE-MOImR0.010
    下载: 导出CSV

    表  7   基于多光谱空域特征的预测模型评价

    Table  7   Evaluation of prediction model based on multispectral spatial domain features

    模型
    Model
    建模 Modeling 检验 Testing
    R2 MAE RMSE R2 MAE RMSE
    MLR 0.650 4.195 5.282 0.643 4.761 5.775
    PLSR 0.649 4.209 5.286 0.677 4.532 5.493
    RF 0.945 1.597 2.090 0.733 4.003 4.997
    XGBoost 0.868 2.430 3.239 0.747 3.821 4.860
    下载: 导出CSV

    表  8   基于多光谱频域特征的预测模型评价

    Table  8   Evaluation of prediction model based on multispectral frequency domain features

    模型
    Model
    建模 Modeling 检验 Testing
    R2 MAE RMSE R2 MAE RMSE
    MLR 0.369 5.990 7.092 0.402 0.424 7.470
    PLSR 0.369 5.900 7.092 0.403 6.321 7.467
    RF 0.866 2.696 3.274 0.412 6.035 7.409
    XGBoost 0.968 1.169 1.609 0.405 5.586 7.455
    下载: 导出CSV

    表  9   基于多光谱融合特征的预测模型评价

    Table  9   Evaluation of prediction model based on multispectral fusion features

    模型
    Model
    建模 Modeling 检验 Testing
    R2 MAE RMSE R2 MAE RMSE
    MLR 0.700 3.806 4.894 0.537 5.074 6.576
    PLSR 0.699 3.813 4.898 0.594 4.899 6.159
    RF 0.947 1.552 2.056 0.791 3.431 4.414
    XGBoost 0.984 0.822 1.126 0.769 3.86 4.641
    下载: 导出CSV
  • [1]

    Ly V, Nanthavong K, Pooma R, et al. Hopea hainanensis[Z/OL]. IUCN Red List of Threatened Species, 2018: e.T32357A2816074[2023−04−11]. https://dx.doi.org/10.2305/IUCN.UK.2018-1.RLTS.T32357A2816074.en.

    [2] 陈澜, 常庆瑞, 高一帆, 等. 猕猴桃叶片叶绿素含量高光谱估算模型研究[J]. 西北农林科技大学学报(自然科学版), 2020, 48(6): 79−89.

    Chen L, Chang Q R, Gao Y F, et al. Hyperspectral estimation model of chlorophyll content in kiwifruit leaves[J]. Journal of Northwest A&F University (Natural Science Edition), 2020, 48(6): 79−89.

    [3]

    Qi H X, Wu Z Y, Zhang L, et al. Monitoring of peanut leaves chlorophyll content based on drone-based multispectral image feature extraction[J]. Computers and Electronics in Agriculture, 2021, 187: 106292. doi: 10.1016/j.compag.2021.106292

    [4]

    Sun Q, Chen L P, Xu X B, et al. A new comprehensive index for monitoring maize lodging severity using UAV-based multi-spectral imagery[J]. Computers and Electronics in Agriculture, 2022, 202: 107362. doi: 10.1016/j.compag.2022.107362

    [5] 李永亮, 张怀清, 林辉. 基于红边参数与PCA的GA-BP神经网络估算叶绿素含量模型[J]. 林业科学, 2012, 48(9): 22−29.

    Li Y L, Zhang H Q, Lin H. GA-BP neural network estimation models of chlorophyll content based on red edge parameters and PCA[J]. Scientia Silvae Sinicae, 2012, 48(9): 22−29.

    [6]

    Lin C W, Ding Q, Tu W H, et al. Fourier dense network to conduct plant classification using UAV-based optical images[J]. IEEE Access, 2019, 7: 17736−17749. doi: 10.1109/ACCESS.2019.2895243

    [7]

    Yang B, Wang M, Sha Z, et al. Evaluation of aboveground nitrogen content of winter wheat using digital imagery of unmanned aerial vehicles[J]. Sensors, 2019, 19(20): 4416. doi: 10.3390/s19204416

    [8]

    Zhuo W, Wu N, Shi R, et al. UAV Mapping of the chlorophyll content in a tidal flat wetland using a combination of spectral and frequency indices[J]. Remote Sensing, 2022, 14(4): 827. doi: 10.3390/rs14040827

    [9]

    Bao Q Z, Gao J H, Chen W C. Local adaptive shrinkage threshold denoising using curvelet coefficients[J]. Electronics Letters, 2008, 44(4): 277−278. doi: 10.1049/el:20082831

    [10]

    Wang Y P, Chang Y C, Shen Y. Estimation of nitrogen status of paddy rice at vegetative phase using unmanned aerial vehicle based multispectral imagery[J]. Precision Agriculture, 2022, 23(1): 1−17. doi: 10.1007/s11119-021-09823-w

    [11]

    Suh H K, Hofstee J W, van Henten E J. Investigation on combinations of colour indices and threshold techniques in vegetation segmentation for volunteer potato control in sugar beet[J]. Computers and Electronics in Agriculture, 2020, 179: 105819. doi: 10.1016/j.compag.2020.105819

    [12]

    Huang S Y, Miao Y X, Yuan F, et al. Potential of RapidEye and WorldView-2 satellite data for improving rice nitrogen status monitoring at different growth stages[J]. Remote Sensing, 2017, 9(3): 227. doi: 10.3390/rs9030227

    [13]

    Prey L, Hu Y C, Schmidhalter U. High-throughput field phenotyping traits of grain yield formation and nitrogen use efficiency: optimizing the selection of vegetation indices and growth stages[J]. Frontiers in Plant Science, 2020, 10: 1672. doi: 10.3389/fpls.2019.01672

    [14] 苏伟, 王伟, 刘哲, 等. 无人机影像反演玉米冠层LAI和叶绿素含量的参数确定[J]. 农业工程学报, 2020, 36(19): 58−65.

    Su W, Wang W, Liu Z, et al. Determining the retrieving parameters of corn canopy LAI and chlorophyll content computed using UAV image[J]. Transactions of the Chinese Society of Agricultural Engineering, 2020, 36(19): 58−65.

    [15]

    Huete A R, Liu H Q, Batchily K, et al. A comparison of vegetation indices over a global set of TM images for EOS-MODIS[J]. Remote Sensing of Environment, 1997, 59(3): 440−451. doi: 10.1016/S0034-4257(96)00112-5

    [16]

    Roujean J L, Breon F M. Estimating PAR absorbed by vegetation from bidirectional reflectance measurements[J]. Remote Sensing of Environment, 1995, 51(3): 375−384. doi: 10.1016/0034-4257(94)00114-3

    [17]

    Gitelson A A, Kaufman Y J, Stark R, et al. Novel algorithms for remote estimation of vegetation fraction[J]. Remote Sensing of Environment, 2002, 80(1): 76−87. doi: 10.1016/S0034-4257(01)00289-9

    [18]

    Daughtry C S T, Walthall C L, Kim M S, et al. Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance[J]. Remote Sensing of Environment, 2000, 74(2): 229−239. doi: 10.1016/S0034-4257(00)00113-9

    [19]

    Chen J M. Evaluation of vegetation indices and a modified simple ratio for boreal applications[J]. Canadian Journal of Remote Sensing, 1996, 22(3): 229−242. doi: 10.1080/07038992.1996.10855178

    [20]

    McDaniel K C, Haas R H. Assessing mesquite-grass vegetation condition from Landsat[J]. Photogrammetric Engineering and Remote Sensing, 1982, 48(3): 441−450.

    [21]

    Harwikarya, Ramayanti D. Feature textures extraction of macroscopic image of jatiwood ( Tectona Grandy) based on gray level co-occurence matrix[J]. IOP Conference Series:Materials Science and Engineering, 2018, 453: 012046. doi: 10.1088/1757-899X/453/1/012046

    [22]

    Kauffmann L, Chauvin A, Guyader N, et al. Rapid scene categorization: role of spatial frequency order, accumulation mode and luminance contrast[J]. Vision Research, 2015, 107: 49−57. doi: 10.1016/j.visres.2014.11.013

    [23] 张孟库, 姜立春. 基于机器学习的落叶松树皮厚度预测[J]. 北京林业大学学报, 2022, 44(6): 54−62.

    Zhang M K, Jiang L C. Prediction of bark thickness for Larix gmelinii based on machine learning[J]. Journal of Beijing Forestry University, 2022, 44(6): 54−62.

    [24] 杨灵玉, 高小红, 张威, 等. 基于Hyperion影像植被光谱的土壤重金属含量空间分布反演: 以青海省玉树县为例[J]. 应用生态学报, 2016, 27(6): 1775−1784.

    Yang L Y, Gao X H, Zhang W, et al. Estimating heavy metal concentrations in topsoil from vegetation reflectance spectra of Hyperion images: a case study of Yushu County, Qinghai, China[J]. Chinese Journal of Applied Ecology, 2016, 27(6): 1775−1784.

    [25]

    Breiman L. Random forests[J]. Machine Learning, 2001, 45(1): 5−32. doi: 10.1023/A:1010933404324

    [26]

    Zhang J J, Cheng T, Guo W, et al. Leaf area index estimation model for UAV image hyperspectral data based on wavelength variable selection and machine learning methods[J]. Plant Methods, 2021, 17: 49. doi: 10.1186/s13007-021-00750-5

    [27]

    Gholizadeh H, Robeson S M, Rahman A F. Comparing the performance of multispectral vegetation indices and machine-learning algorithms for remote estimation of chlorophyll content: a case study in the Sundarbans mangrove forest[J]. International Journal of Remote Sensing, 2015, 36(12): 3114−3133. doi: 10.1080/01431161.2015.1054959

    [28]

    Liu Y, Hatou K, Aihara T, et al. A robust vegetation index based on different UAV RGB images to estimate SPAD values of naked barley leaves[J]. Remote Sensing, 2021, 13(4): 686. doi: 10.3390/rs13040686

    [29]

    Qiu Z C, Ma F, Li Z W, et al. Estimation of nitrogen nutrition index in rice from UAV RGB images coupled with machine learning algorithms[J]. Computers and Electronics in Agriculture, 2021, 189: 106421. doi: 10.1016/j.compag.2021.106421

    [30]

    Xie X, Zhang X, Shen J, et al. Poplar’s waterlogging resistance modeling and evaluating: exploring and perfecting the feasibility of machine learning methods in plant science[J]. Frontiers in Plant Science, 2022, 13: 821365. doi: 10.3389/fpls.2022.821365

  • 期刊类型引用(0)

    其他类型引用(3)

图(5)  /  表(9)
计量
  • 文章访问数:  377
  • HTML全文浏览量:  64
  • PDF下载量:  36
  • 被引次数: 3
出版历程
  • 收稿日期:  2023-05-11
  • 修回日期:  2023-10-11
  • 录用日期:  2023-10-11
  • 网络出版日期:  2023-10-15
  • 刊出日期:  2023-11-29

目录

/

返回文章
返回