Estimation of above ground biomass of drone Diarthron tianschanicum based on multi feature fusion

doi:10.6048/j.issn.1001-4330.2024.10.020

Abstract

Abstract:

【Objective】 This project aims to explore the ability of UAV multi feature to construct D. tianschanicum aboveground biomass (AGB) estimation model.The finding has provided a reference for the grading basis for the classification of grassland degradation degree. 【Methods】 Diarthron tianschanicum is one of the degradation indicator plants, and its growth status can reflect the degree of grassland degradation. and to extract the spectral features, texture features and D. tianschanicum coverage from visible high spatial resolution remote sensing images, and the three were used as inputs to establish a univariate linear model. The three types of features were fused to construct multiple stepwise regression and artificial neural network models, and the effect of multi feature fusion to estimate AGB was analyzed. 【Results】 (1) The best coverage extraction window period of D. tianschanicum was in full bloom, and the effect of D. tianschanicum extraction model constructed by RF algorithm was ideal, and the overall accuracy was more than 81%. (2) Spectral features, texture features and coverage were all correlated with AGB, and the texture feature G had the highest correlation, which was 0.784. (3) Compared with single vegetation index, texture feature, coverage and any two feature combinations as input amount, the accuracy of AGB was the highest, with R² and RMSE of 0.870 and 15.383, respectively. 【Conclusion】 It is verified by artificial neural network mode that the fusion of spectral features, texture index and coverage can effectively improve the accuracy of AGB estimation.

Key words: drone; Diarthron tianschanicum biomass; visible light; coverage; texture features

摘要：

【目的】 研究无人机多特征构建天山假狼毒地上生物量(AGB)估算模型的能力,为草地退化程度的分级提供参考依据。【方法】 天山假狼毒(Diarthron tianschanicum)作为退化指示植物之一,其生长状况可反映草地退化程度。通过可见光高空间分辨率遥感影像提取光谱特征、纹理特征和天山假狼毒覆盖度,将三者分别作为输入量建立一元线性模型,三类特征融合构建多元逐步回归与人工神经网络模型,分析多特征融合估算天山假狼毒AGB的效果。【结果】 (1)盛花期为天山假狼毒最佳覆盖度提取窗口期,利用RF算法构建的天山假狼毒提取模型效果较为理想,总体精度在81%以上。(2)光谱特征、纹理特征、覆盖度均与天山假狼毒AGB具有相关性,其中纹理特征cor_G相关性最高,达0.784。(3)对比单一植被指数、纹理特征、覆盖度及任意两种特征组合作为输入量,多特征融合估算天山假狼毒AGB的精度最高,R²、RMSE为0.870、15.383,并利用人工神经网络模型对此结论进行验证。【结论】 融合光谱特征、纹理指数、覆盖度能有效提高天山假狼毒AGB估算精度。

关键词: 无人机, 天山假狼毒生物量, 可见光, 覆盖度, 纹理特征

CLC Number:

S812

HOU Zhengqing, YAN An, XIE Kaiyun, YUAN Yilin, XIA Wenqiu, XIAO Shuting, ZHANG Zhenfei, SUN Zhe. Estimation of above ground biomass of drone Diarthron tianschanicum based on multi feature fusion[J]. Xinjiang Agricultural Sciences, 2024, 61(10): 2527-2536.

侯正清, 颜安, 谢开云, 袁以琳, 夏雯秋, 肖淑婷, 张振飞, 孙哲. 基于多特征融合的无人机天山假狼毒地上生物量估算[J]. 新疆农业科学, 2024, 61(10): 2527-2536.

Figures/Tables 10

Tab.1 Quantitative characteristics of main plants in the study area community

物种 Species	高度 Height(cm)	盖度 Coverage(%)	密度 Density/plant(m²)	重要值 Important value
天山假狼毒 Stelleropsis tianschanica	30.188±5.522	27.999±14.224	3.125±0.976	0.205
紫苞鸢尾 Iris ruthenica	18.086±2.854	28.046±10.126	16.556±5.577	0.159
无芒雀麦 Smooth bromegrass	25.233±14.769	15.569±8.164	48.183±52.220	0.159
夏至草 Horehound	13.511±3.483	15.008±8.780	70.348±54.406	0.157
百里香 Thyme	11.387±1.697	10.889±5.271	15.048±13.048	0.157
车前 Snokeweed	14.237±13.342	13.567±4.375	15.500±8.635	0.141

Tab.2 Visible light vegetation index and calculation formula

植被指数 Vegetation Indexes	计算公式 Formula
归一化绿红差异指数 Normalized green-red difference index,NGRDI	(G-R)/(G+R)^[13]
归一化绿蓝差异指数 Normalized green and blue disparity index,NGBDI	(G-B)/(G+B)^[13]
绿蓝比值指数Green-blue ratio index,GBRI	G/B^[13]
红光标准化值Normalized redness intensity,NRI	R/(R+B+G)^[19]
绿光标准化值Normalized greenness intensity,NGI	G/(R+B+G)^[19]
蓝光标准化值Normalized blueness intensity,NBI	B/(R+B+G)^[19]
红绿比值指数Red-green ratio index,RGRI	R/G^[19]
过红指数Excess red index,ExR	1.4R-G^[19]

Tab.3 Texture features and its formulas

纹理特征 Texture feature	公式 Formula
均值 Mean	$m e a n = \sum_{i, j}^{N - 1} i P_{i, j}$
方差 Variance	$v a r = \sum_{i, j = 0}^{N - 1} i P_{i, j} {(i - m e a n)}^{2}$
同质性 Homogeneity	$h o m = \sum_{i, j = 0}^{N - 1} i \frac{P_{i, j}}{1 + {(i - j)}^{2}}$
对比度 Contrast	$c o n = \sum_{i, j = 0}^{N - 1} i P_{i, j} {(i - j)}^{2}$
差异性 Dissimilarity	$d i s = \sum_{i, j = 0}^{N - 1} i P_{i, j} \|i - j\|$
熵 Entropy	$e n t = \sum_{i, j = 0}^{N - 1} i P_{i, j} (- l n P_{i, j})$
二阶距 Second Moment	$s m = \sum_{i, j = 0}^{N - 1} i {P_{i, j}}^{2}$
相关性 Correlation	$c o r = \sum_{i, j = 0}^{N - 1} i P_{i, j} [\frac{(i - m e a n) (j - m e a n)}{\sqrt{v a r_{i} \times v a r_{j}}}]$

Tab.3 Texture features and its formulas

纹理特征 Texture feature	公式 Formula
均值 Mean	$m e a n = \sum_{i, j}^{N - 1} i P_{i, j}$
方差 Variance	$v a r = \sum_{i, j = 0}^{N - 1} i P_{i, j} {(i - m e a n)}^{2}$
同质性 Homogeneity	$h o m = \sum_{i, j = 0}^{N - 1} i \frac{P_{i, j}}{1 + {(i - j)}^{2}}$
对比度 Contrast	$c o n = \sum_{i, j = 0}^{N - 1} i P_{i, j} {(i - j)}^{2}$
差异性 Dissimilarity	$d i s = \sum_{i, j = 0}^{N - 1} i P_{i, j} \|i - j\|$
熵 Entropy	$e n t = \sum_{i, j = 0}^{N - 1} i P_{i, j} (- l n P_{i, j})$
二阶距 Second Moment	$s m = \sum_{i, j = 0}^{N - 1} i {P_{i, j}}^{2}$
相关性 Correlation	$c o r = \sum_{i, j = 0}^{N - 1} i P_{i, j} [\frac{(i - m e a n) (j - m e a n)}{\sqrt{v a r_{i} \times v a r_{j}}}]$

Fig. 1 Correlation between spectral, texture features, and coverage with AGB of Diarthron tianschanicum

Fig. 2 Spectral differences between Diarthron tianschanicum and other vegetation in late June

Fig.3 Orthorectified image of the experimental area captured by the UAV

Fig. 4 Distribution map of Diarthron tianschanicum (after merging small patches)

Tab.4 Univariate linear estimation models and accuracy evaluation for AGB of Diarthron tianschanicum

模型输入 Model input	模型公式 Model formula	决定系数 R²	均方根误差 RMSE (g/m²)
ExR	Y=26.173-2.028X	0.544	29.315
NRI	Y=-188.86+919.344X	0.400	33.623
NBI	Y=628.877-1 971.515X	0.381	34.166
NGRDI	Y=-65.082+582.556X	0.318	35.860
NGBDI	Y=--166.357+1 080.686X	0.467	35.741
NGI	Y=-843.908+2 060.782X	0.308	36.122
GBRI	Y=-24.727+59.308X	0.137	40.334
RGRI	Y=485.473-664.829X	0.492	30.934
mean_B	Y=280.073-8.103 624X	0.345	35.144
hom_B	Y=229.254-947.174X	0.220	38.294
cor_B	Y=145.009-217.139X	0.193	38.999
mean_G	Y=-52.137+3.867X	0.511	30.360
hom_G	Y=93.821-0.692X	0.186	39.164
cor_G	Y=-68.339+369.227X	0.606	27.237
hom_R	Y=-225.691-914.770 713X	0.195	38.940
cor_R	Y=-62.148+355.199 683X	0.576	28.251
FVC	Y=-15.636+161.079X	0.501	30.664

Tab.5 Multivariate linear stepwise regression estimation models and accuracy evaluation for AGB of Diarthron tianschanicum

模型输入量 Model input	模型公式 Model formula	因子 Factor	决定系数 R²	均方根误差 RMSE (g/m²)
VIs+FVC	Y=-0.888-1.907exr+93.623FVC	exr,FVC	0.712	23.281
TIs+FVC	Y=-160.891+7.335mean_G-9.285mean_B+1.439hom_G+994.878hom_B+60.704FVC	mean_G,mean_B,hom_G,hom_B,FVC	0.830	17.398
VIs+TIs	Y=-400.961+6.644mean_G+1 487.090hom_R-7.485mean_B+1.138hom_G+728.831ngbdi	mean_G,hom_R,mean_B,hom_G,ngbdi	0.850	16.739
VIS+TIS+FVC	Y=778.539-1.141exr+53.348FVC+1445.883hom_B+641.677nri+2.519mean_G-2152.134nbi	Exr,FVC,hom_B,nri,mean_G,nbi,ngi	0.870	15.383

Tab.6 Artificial neural network forecast model and accuracy evaluation of AGB

模型输入 Model input	决定系数 R²	均方根误差 RMSE (g/m²)
VIs	0.648	24.060
TIs	0.680	18.249
FVC	0.562	17.261
VIs+TIs	0.785	20.345
VIs+FVC	0.726	23.987
TIs+FVC	0.787	22.511
VIs+TIs+FVC	0.806	22.685

References 32

[1]	江惠, 王明利, 励汀郁, 等. 新疆草原畜牧业转型升级:发展现状、现实困境与实现路径[J]. 华中农业大学学报, 2023, 42(5): 42-52.
	JIANG Hui, WANG Mingli, LI Tingyu, et al. Transformation and upgrading of grassland animal husbandry in Xinjiang: development status, realistic dilemma and realization path[J]. Journal of Huazhong Agricultural University, 2023, 42(5): 42-52.
[2]	闫俊杰, 刘海军, 崔东, 等. 近15年新疆伊犁河谷草地退化时空变化特征[J]. 草业科学, 2018, 35(3): 508-520.
	YAN Junjie, LIU Haijun, CUI Dong, et al. Spatiotemporal dynamics of grassland degradation in Yili Valley of Xinjiang over the last 15 years[J]. Pratacultural Science, 2018, 35(3): 508-520.
[3]	李宏, 陈卫民, 陈翔, 等. 新疆伊犁草原毒害草种类及其发生与危害[J]. 草业科学, 2010, 27(11): 171-173.
	LI Hong, CHEN Weimin, CHEN Xiang, et al. Poisonous weed species and their harmfulness in the grassland of Yili regions, Xinjiang Uygur Autonomous Region of China[J]. Pratacultural Science, 2010, 27(11): 171-173.
[4]	李宏, 陈卫民, 王华, 等. 四种吡啶类除草剂防除白喉乌头的筛选试验[J]. 现代农药, 2010, 9(3): 48-50.
	LI Hong, CHEN Weimin, WANG Hua, et al. Screening experiment of four pyridine herbicides on the Aconitum leucostomum[J]. Modern Agrochemicals, 2010, 9(3): 48-50.
[5]	范云豹, 宫兆宁, 赵文吉, 等. 基于高光谱遥感的植被生物量反演方法研究[J]. 河北师范大学学报(自然科学版), 2016, 40(3): 267-271.
	FAN Yunbao, GONG Zhaoning, ZHAO Wenji, et al. Study on vegetation biomass inversion method based on hyperspectral remote sensing[J]. Journal of Hebei Normal University (Natural Science Edition), 2016, 40(3): 267-271.
[6]	刘建刚, 赵春江, 杨贵军, 等. 无人机遥感解析田间作物表型信息研究进展[J]. 农业工程学报, 2016, 32(24): 98-106.
	LIU Jiangang, ZHAO Chunjiang, YANG Guijun, et al. Review of field-based phenotyping by unmanned aerial vehicle remote sensing platform[J]. Transactions of the Chinese Society of Agricultural Engineering, 2016, 32(24): 98-106.
[7]	Candiago S, Remondino F, De Giglio M, et al. Evaluating multispectral images and vegetation indices for precision farming applications from UAV images[J]. Remote Sensing, 2015, 7(4): 4026-4047.
[8]	Potgieter A B, George-Jaeggli B, Chapman S C, et al. Multi-spectral imaging from an unmanned aerial vehicle enables the assessment of seasonal leaf area dynamics of Sorghum breeding lines[J]. Frontiers in Plant Science, 2017, 8: 1532. DOI PMID
[9]	Yang G J, Li C C, Wang Y J, et al. The DOM generation and precise radiometric calibration of a UAV-mounted miniature snapshot hyperspectral imager[J]. Remote Sensing, 2017, 9(7): 642.
[10]	Nie S, Wang C, Dong P L, et al. Estimating leaf area index of maize using airborne discrete-return LiDAR data[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016, 9(7): 3259-3266.
[11]	Guo Q H, Su Y J, Hu T Y, et al. An integrated UAV-borne lidar system for 3D habitat mapping in three forest ecosystems across China[J]. International Journal of Remote Sensing, 2017, 38(8/9/10): 2954-2972.
[12]	陶惠林, 徐良骥, 冯海宽, 等. 基于无人机数码影像的冬小麦株高和生物量估算[J]. 农业工程学报, 2019, 35(19): 107-116.
	TAO Huilin, XU Liangji, FENG Haikuan, et al. Estimation of plant height and biomass of winter wheat based on UAV digital image[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(19): 107-116.
[13]	张正健, 李爱农, 边金虎, 等. 基于无人机影像可见光植被指数的若尔盖草地地上生物量估算研究[J]. 遥感技术与应用, 2016, 31(1): 51-62.
	ZHANG Zhengjian, LI Ainong, BIAN Jinhu, et al. Estimating aboveground biomass of grassland in zoige by visible vegetation index derived from unmanned aerial vehicle image[J]. Remote Sensing Technology and Application, 2016, 31(1): 51-62.
[14]	陈鹏飞, Nicolas Tremblay, 王纪华, 等. 估测作物冠层生物量的新植被指数的研究[J]. 光谱学与光谱分析, 2010, 30(2): 512-517.
	CHEN Pengfei, Nicolas Tremblay, WANG Jihua, et al. New index for crop canopy fresh biomass estimation[J]. Spectroscopy and Spectral Analysis, 2010, 30(2): 512-517. PMID
[15]	刘欣谊, 仲晓春, 陈晨, 等. 利用无人机图像颜色与纹理特征数据在小麦生育前期对产量进行预测[J]. 麦类作物学报, 2020, 40(8): 1002-1007.
	LIU Xinyi, ZHONG Xiaochun, CHEN Chen, et al. Prediction of wheat yield using color and texture feature data of UAV image at early growth stage[J]. Journal of Triticeae Crops, 2020, 40(8): 1002-1007.
[16]	刘杨, 冯海宽, 孙乾, 等. 不同分辨率无人机数码影像的马铃薯地上生物量估算研究[J]. 光谱学与光谱分析, 2021, 41(5): 1470-1476.
	LIU Yang, FENG Haikuan, SUN Qian, et al. Estimation study of above ground biomass in potato based on UAV digital images with different resolutions[J]. Spectroscopy and Spectral Analysis, 2021, 41(5): 1470-1476.
[17]	刘艳慧, 蔡宗磊, 包妮沙, 等. 基于无人机大样方草地植被覆盖度及生物量估算方法研究[J]. 生态环境学报, 2018, 27(11): 2023-2032. DOI
	LIU Yanhui, CAI Zonglei, BAO Nisha, et al. Research of grassland vegetation coverage and biomass estimation method based on major quadrat from UAV photogrammetry[J]. Ecology and Environmental Sciences, 2018, 27(11): 2023-2032.
[18]	杭艳红, 苏欢, 于滋洋, 等. 结合无人机光谱与纹理特征和覆盖度的水稻叶面积指数估算[J]. 农业工程学报, 2021, 37(9): 64-71.
	HANG Yanhong, SU Huan, YU Ziyang, et al. Estimation of rice leaf area index combining UAV spectrum, texture features and vegetation coverage[J]. Transactions of the Chinese Society of Agricultural Engineering, 2021, 37(9): 64-71.
[19]	牛庆林, 冯海宽, 杨贵军, 等. 基于无人机数码影像的玉米育种材料株高和LAI监测[J]. 农业工程学报, 2018, 34(5): 73-82.
	NIU Qinglin, FENG Haikuan, YANG Guijun, et al. Monitoring plant height and leaf area index of maize breeding material based on UAV digital images[J]. Transactions of the Chinese Society of Agricultural Engineering, 2018, 34(5): 73-82.
[20]	Jiang B R, Wang P, Zhuang S, et al. Detection of maize drought based on texture and morphological features[J]. Computers and Electronics in Agriculture, 2018, 151: 50-60.
[21]	刘咏梅, 胡念钊, 龙永清, 等. 无人机RGB影像在高寒草地狼毒入侵监测及盖度估算中的应用[J]. 中国草地学报, 2023, 45(2): 1-12.
	LIU Yongmei, HU Nianzhao, LONG Yongqing, et al. Application of UAV RGB image in monitoring and coverage estimation of Stellera chamaejasme invasion in alpine grasslands, Qinghai-Tibet Plateau[J]. Chinese Journal of Grassland, 2023, 45(2): 1-12.
[22]	贾坤, 姚云军, 魏香琴, 等. 植被覆盖度遥感估算研究进展[J]. 地球科学进展, 2013, 28(7): 774-782. DOI
	JIA Kun, YAO Yunjun, WEI Xiangqin, et al. A review on fractional vegetation cover estimation using remote sensing[J]. Advances in Earth Science, 2013, 28(7): 774-782. DOI
[23]	刘智彪. 基于无人机遥感的白喉乌头防效监测方法的研究[D]. 乌鲁木齐: 新疆农业大学, 2022.
	LIU Zhibiao. Study on Monitoring Method of Aconitum Leucostomum Prevention Effect Based on Unmanned Aerial Vehicl Remote Sensing[D]. Urumqi: Xinjiang Agricultural University, 2022.
[24]	Haralick R M, Shanmugam K, Dinstein I. Textural features for image classification[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1973, SMC-3(6): 610-621.
[25]	范宏. 基于UAV影像的草原毒害草识别检测方法研究——以白喉乌头为例[D]. 乌鲁木齐: 新疆大学, 2020.
	FAN Hong. Research on Identification and Detection Method of Grassland Poisonous Grass Based on UAV Image[D]. Urumqi: Xinjiang University, 2020.
[26]	陈鹏, 冯海宽, 李长春, 等. 无人机影像光谱和纹理融合信息估算马铃薯叶片叶绿素含量[J]. 农业工程学报, 2019, 35(11): 63-74.
	CHEN Peng, FENG Haikuan, LI Changchun, et al. Estimation of chlorophyll content in potato using fusion of texture and spectral features derived from UAV multispectral image[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(11): 63-74.
[27]	Karal Ö. Performance comparison of different kernel functions in SVM for different k value in k-fold cross-validation[C]// 2020 Innovations in Intelligent Systems and Applications Conference (ASYU). Istanbul, Turkey. IEEE, 2020: 1-5.
[28]	Chanda M, Biswas M. Plant disease identification and classification using Back-Propagation Neural Network with Particle Swarm Optimization[C]// 2019 3rd International Conference on Trends in Electronics and Informatics (ICOEI). Tirunelveli, India. IEEE, 2019: 1029-1036.
[29]	王来刚, 贺佳, 郑国清, 等. 基于无人机多光谱遥感的玉米FPAR估算[J]. 农业机械学报, 2022, 53(10): 202-210.
	WANG Laigang, HE Jia, ZHENG Guoqing, et al. Estimation of maize FPAR based on UAV multispectral remote sensing[J]. Transactions of the Chinese Society for Agricultural Machinery, 2022, 53(10): 202-210.
[30]	杨福芹, 冯海宽, 肖天豪, 等. 融合无人机影像光谱与纹理特征的冬小麦氮营养指数估算[J]. 农业现代化研究, 2020, 41(4): 718-726.
	YANG Fuqin, FENG Haikuan, XIAO Tianhao, et al. Nitrogen nutrition index estimation in winter wheat by UAV spectral information and texture feature fusion[J]. Research of Agricultural Modernization, 2020, 41(4): 718-726.
[31]	汪小钦, 江洪, 傅银贞. 森林叶面积指数遥感研究进展[J]. 福州大学学报(自然科学版), 2009, 37(6): 822-828.
	WANG Xiaoqin, JIANG Hong, FU Yinzhen. The progress of the forest LAI estimation from remote sensing data[J]. Journal of Fuzhou University (Natural Science Edition), 2009, 37(6): 822-828.
[32]	Yu Z Y, Ustin S L, Zhang Z C, et al. Estimation of a new canopy structure parameter for rice using smartphone photography[J]. Sensors, 2020, 20(14): 4011.