天山云杉林无人机可见光影像树冠信息提取方法研究

doi:10.13466/j.cnki.lyzygl.2020.01.016

摘要/Abstract

摘要：

高精度轻小型无人机森林树冠参数信息提取方法是森林资源监测和生态功能评估的重要基础。以新疆山地森林优势树种天山云杉为研究对象,在南山实习林场采集积雪背景下无人机遥感影像并进行预处理,对比分析3种方法(以光谱为特征空间的面向对象法、以光谱+纹理为特征空间的面向对象法、随机森林法)提取天山云杉林树冠参数信息的精度。结果表明:3种方法提取天山云杉林分郁闭度精度均高于93%,其中以光谱为特征空间的面向对象法最优,精度可达93.73%;3种方法自动提取单木树冠面积与目视解译结果的 R ² 均大于0.91,虽然随机森林法的统计指标最优,但面向对象法可获得完整的单木树冠闭合曲线,故以光谱+纹理为特征空间的面向对象法效果较优;面向对象法的最优分割尺度为29,纹理特征的加入会导致郁闭度提取精度降低0.66%,但会优化单木树冠面积提取效果。

关键词: 无人机遥感, 天山云杉, 面向对象, 随机森林, 树冠参数

Abstract:

The extraction method of forest tree crown parameter information obtained by high-precision light and miniature unmanned aerial vehicle(UAV)is an important basis for forest resource monitoring and ecological function evaluation.Taking the Piceaschrenkiana var.tianschanica,the dominant tree species in the mountainous forests of Xinjiang,as the research object.Collecting and preprocessing remote sensing images of UAV with the background of snow in the Nanshan Internship Forest Farm.The three methods(object-based method with spectral feature space,object-based method with spectral+texture as feature space and random forest method)were compared and analyzed to extract the accuracy of forest tree crown parameter information of Piceaschrenkiana var.tianschanica forest.The results show:the accuracy of the canopy density to be extracted by the three methods is higher than 93%,and object-based method with spectral+texture as feature space is the best,its accuracy reaches 93.73%.The R ² of the visual interpretation result and the single tree crown area to be extracted by the three methods aremore than 0.91.Although the results to be extracted by random forest method are closest to the visual interpretation value,the object-based method obtains the complete single tree crown closure curve,so the object-based method with spectral+texture as the feature space is the best method for extracting the single tree crown area.The optimal segmentation scale for extracting canopy density and single tree crown area using object-based method is 29,and adding texture features reduces the accuracy of canopy density extraction by 0.66%,but optimizes the extraction effect of single tree crown area.Therefore,the UAV image is an effective way to extracting theforest tree crown parameter information accurately,quickly and automatically,and optimal extraction method can be selected according to different crown parameter information.

Key words: UAV remote sensing, Piceaschrenkiana var.tianschanica, object-based, random forests, crown parameters

中图分类号:

S758
TP751

金忠明, 曹姗姗, 王蕾, 孙伟. 天山云杉林无人机可见光影像树冠信息提取方法研究[J]. 林业资源管理, 2020,(1): 125-135.

Zhongming JIN, Shanshan CAO, Lei WANG, Wei SUN. Study on Extraction of Tree Crown Information from UAV Visible Light Image of Piceaschrenkiana var.tianschanica Forest[J]. FOREST RESOURCES WANAGEMENT, 2020,(1): 125-135.

图/表 12

图1

图2

图3

图4

图5

表1

图6

图7

表2

图8

图9

图10

参考文献 27

[1]	Poulio t D A, King D J, Bell F W , et al. Automated Tree Crown Detection and Delineation in High-resolution Digital Camera Imagery of Coniferous Forest Regeneration[J]. Remote Sensing of Environment, 2002(82):322-334.
[2]	刘文萍, 仲亭玉, 宋以宁 . 基于无人机图像分析的树木胸径预测[J]. 农业工程学报, 2017,33(21):99-104.
[3]	Panagiotidis D, Abdollahnejad A, Surov P , et al. Determining treeheight and crown diameter from highresolution UAV imagery[J]. International Journal of Remote Sensing, 2017,38(8):2392-2410. doi: 10.1080/01431161.2016.1264028
[4]	李擎, 王振锡, 王雅佩 , 等. 基于GF-2号遥感影像的天山云杉林郁闭度估测研究[J]. 中南林业科技大学学报, 2019(8):48-54.
[5]	何游云 . 无人机遥感估测林木地上生物量的研究[D]. 北京:北京林业大学, 2016.
[6]	耿林, 李明泽, 范文义 , 等. 基于机载LiDAR的单木结构参数及林分有效冠的提取[J]. 林业科学, 2018,54(7):62-72.
[7]	张凝, 冯跃文, 张晓丽 , 等. 结合航空影像纹理和光谱特征的单木冠幅提取[J]. 北京林业大学学报, 2015,37(3):13-19.
[8]	井然, 邓磊, 赵文吉 , 等. 基于可见光植被指数的面向对象湿地水生植被提取方法[J]. 应用生态学报, 2016,27(5):1427-1436.
[9]	李赟 . 基于UAV高分影像的林木冠幅提取与蓄积量估测研究[D]. 南京:南京林业大学, 2017.
[10]	毛学刚, 邢秀丽, 李佳蕊 , 等. 基于航空正射影像的面向对象林隙识别[J]. 林业科学, 2019,55(2):87-96.
[11]	张雷, 王琳琳, 张旭东 , 等. 随机森林算法基本思想及其在生态学中的应用—以云南松分布模拟为例[J]. 生态学报, 2014,34(3):650-659. doi: 10.5846/stxb201306031292
[12]	郭昱杉, 刘庆生, 刘高焕 , 等. 基于标记控制分水岭分割方法的高分辨率遥感影像单木树冠提取方法[J]. 地球信息科学学报, 2016,18(9):1259-1266. doi: 10.3724/SP.J.1047.2016.01259
[13]	Mads Jeppe Tarp-Johansen . Automatic Stem Mapping in Three Dimensions by Template Matching from Aerial Photographs[J]. Scandinavian Journal of Forest Research, 2002,17(4):359-368. doi: 10.1080/02827580260138107
[14]	刘贵峰, 臧润国, 丁易 , 等. 天山云杉种群结构的研究[J]. 林业科学研究, 2011,24(5):659-662.
[15]	李路, 常亚鹏, 许仲林 . 乌鲁木齐南山森林生物量和碳储量空间分布特征[J]. 水土保持研究, 2018,25(5):72-77.
[16]	肖中琪, 黄建文, 凌成星 . 新疆天山西部天然林保护工程区林地及森林生物量空间变化分析[J]. 林业资源管理, 2018(1):57-62.
[17]	Dragut L, Tiede D, Levick S R . ESP:A tool to estimate scale parameter for multiresolution image segmentation of remotely sensed data[J]. International Journal of Geographical Information Science, 2010,24(6):859-871. doi: 10.1080/13658810903174803
[18]	马浩然 . 基于多层次分割的遥感影像面向对象森林分类[D]. 北京:北京林业大学, 2014.
[19]	Meyer G E, NetoJ C. Verification of color vegetation indices for automated crop imaging applications[J]. Computers and Electronics in Agriculture, 2008,63(2):282-293. doi: 10.1016/j.compag.2008.03.009
[20]	Breiman, L. Random forests[J]. Machine Learning, 2001,45(1):5-32. doi: 10.1023/A:1010933404324
[21]	Kulkarni V Y, Sinha P K. Pruning of random forest classifiers:A survey and future directions[C]. Cochin:International Conference on Data Science & Engineering(ICDSE), 2012: 64-68.
[22]	Gan Haiyan, Li Jun, Chapman M , et al. Integration of orthoimagery and lidar data for object-based urban thematic mapping using random forests[J]. International Journal of Remote Sensing, 2013,34(14):5166-5186. doi: 10.1080/01431161.2013.788261
[23]	Lawrence R, Wood S, Sheley R . Mapping invasive plants using hyperspectral imagery and breiman cutler classifications(randomforest)[J]. Remote Sensing of Environment, 2006,100(3):356-362. doi: 10.1016/j.rse.2005.10.014
[24]	Colditz R . An evaluation of different training sample allocation schemes for discrete and continuous land cover classification using decision tree-based algorithms[J]. Remote Sensing, 2015,7(8):9655-9681. doi: 10.3390/rs70809655
[25]	Gislason P, Benediktsson J, Sveinsson J . Random forests for land cover classification[J]. Pattern Recognition Letters, 2006,27(4):294-300. doi: 10.1016/j.patrec.2005.08.011
[26]	Van der Linden S, Rabe A, Held M , et al. The EnMAP-Box—A Toolbox and Application Programming Interface for EnMAP Data Processing[J]. Remote Sens. 2015,7(9):11249-11266. doi: 10.3390/rs70911249
[27]	姚国慧 . 基于雷达数据的天山云杉林郁闭度信息提取[D]. 乌鲁木齐:新疆农业大学, 2016.

特征分类	特征参数
光谱特征	Mean Red,Mean Green,Mean Blue,Ratio Red,Ratio Green,Ratio Blue,StdDev Red,StdDev Green,StdDev Blue,Brightness,ExG-ExR
纹理特征	GLCM Mean(all dir.),GLCM Dissimilarity(all dir.),GLCM Correlation(all dir.),GLCM StdDev(all dir.),GLCM Contrast(all dir.),GLCM Homogeneity(all dir.),GLDV Mean(all dir.),GLDV Contrast(all dir.)

样地编号	目视解译法计算郁闭度值	以光谱为特征空间的面向对象法		以光谱+纹理为特征空间的面向对象法		随机森林法
样地编号	目视解译法计算郁闭度值	郁闭度值	相对误差/%	郁闭度值	相对误差/%	郁闭度值	相对误差/%
1	0.384	0.388	1.04	0.382	0.52	0.397	3.39
2	0.503	0.479	4.77	0.473	5.96	0.480	4.57
3	0.621	0.803	29.31	0.832	33.98	0.842	35.59
4	0.553	0.565	2.17	0.603	9.04	0.558	0.86
5	0.405	0.430	6.17	0.441	8.89	0.427	5.43
6	0.401	0.411	2.49	0.428	6.73	0.392	2.24
7	0.427	0.480	12.41	0.464	8.67	0.426	0.23
8	0.463	0.486	4.97	0.478	3.24	0.430	7.12
9	0.470	0.525	11.70	0.523	11.28	0.528	12.34
10	0.326	0.313	3.99	0.324	0.61	0.311	4.60
11	0.432	0.449	3.94	0.424	1.85	0.404	6.48
12	0.413	0.414	0.24	0.395	4.36	0.378	8.47
13	0.547	0.603	10.24	0.574	4.94	0.540	1.28
14	0.291	0.300	3.09	0.304	4.47	0.271	6.87
15	0.349	0.340	2.58	0.355	1.72	0.341	2.29
16	0.343	0.347	1.17	0.327	4.66	0.332	3.21