共查询到20条相似文献，搜索用时 31 毫秒
1.
Water depth estimation using optical remote sensing offers a reliable and efficient means of mapping coastal zones. Here, we aim to find a suitable model for fast and practical bathymetry of an estuary using Indian Remote Sensing Satellite (IRS) Linear Imaging Self Scanning Sensor (LISS-3) images. The study examines three different models; (1) least square regression model, (2) spectral band-ratio method and (3) multi-tidal bathymetry model. The findings are supported with in situ observed depth values and statistical estimates. Although the least square regression model has provided best results with root mean square error (RMSE) of 0.4 m, it requires a large number of observed data points for absolute depth estimation. Spectral band-ratio and multi-tidal model provides results with RMSEs 2.1 and 0.9 m, respectively. The present investigation demonstrates that multi-date imagery exploitation at disparate tide levels is the best estimation technique for recursive shallow water bathymetry where in situ observation is not possible. 相似文献
2.
This paper presents a spatially distributed support vector machine (SVM) system for estimating shallow water bathymetry from optical satellite images. Unlike the traditional global models that make predictions from a unified global model for the entire study area, our system uses locally trained SVMs and spatially weighted votes to make predictions. By using IKONOS-2 multi-spectral image and airborne bathymetric LiDAR water depth samples, we developed a spatially distributed SVM system for bathymetry estimates. The distributed model outperformed the global SVM model in predicting bathymetry from optical satellite images, and it worked well at the scenarios with a low number of training data samples. The experiments showed the localized model reduced the bathymetry estimation error by 60% from RMSE of 1.23 m to 0.48 m. Different from the traditional global model that underestimates water depth near shore and overestimates water depth offshore, the spatially distributed SVM system did not produce regional prediction bias and its prediction residual exhibited a random pattern. Our model worked well even if the sample density was much lower: The model trained with 10% of the samples was still able to obtain similar prediction accuracy as the global SVM model with the full training set. 相似文献
3.
The Sentinel-2 Multi-Spectral Imager (MSI) has three spectral bands centered at 705, 740, and 783 nm wavelengths that exploit the red-edge information useful for quantifying plant biochemical traits. This sensor configuration is expected to improve the prediction accuracy of vegetation chlorophyll content. In this work, we assessed the performance of several statistical and physical-based methods in retrieving canopy chlorophyll content (CCC) from Sentinel-2 in a heterogeneous mixed mountain forest. Amongst the algorithms presented in the literature, 13 different vegetation indices (VIs), a non-parametric statistical approach, and two radiative transfer models (RTM) were used to assess the CCC prediction accuracy. A field campaign was conducted in July 2017 to collect in situ measurements of CCC in Bavarian forest national park, and the cloud-free Sentinel-2 image was acquired on 13 July 2017. The leave-one-out cross-validation technique was used to compare the VIs and the non-parametric approach. Whereas physical-based methods were calibrated using simulated data and validated using the in situ reference dataset. The statistical-based approaches, such as the modified simple ratio (mSR) vegetation index and the partial least square regression (PLSR) outperformed all other techniques. As such the modified simple ratio (mSR3) (665, 865) gave the lowest cross-validated RMSE of 0.21 g/m2 (R2 = 0.75). The PLSR resulted in the highest R2 of 0.78, and slightly higher RMSE =0.22 g/m2 than mSR3. The physical-based approach-INFORM inversion using look-up table resulted in an RMSE =0.31 g/m2, and R2 = 0.67. Although mapping CCC using these methods revealed similar spatial distribution patterns, over and underestimation of low and high CCC values were observed mainly in the statistical approaches. Further validation using in situ data from different terrestrial ecosystems is imperative for both the statistical and physical-based approaches' effectiveness to quantify CCC before selecting the best operational algorithm to map CCC from Sentinel-2 for long-term terrestrial ecosystems monitoring across the globe. 相似文献
4.
Forests play a vital role in biological cycles and environmental regulation. To understand the key processes of forest canopies (e.g., photosynthesis, respiration and transpiration), reliable and accurate information on spatial variability of Leaf Area Index (LAI), and its seasonal dynamics is essential. In the present study, we assessed the performance of biophysical parameter (LAI) retrieval methods viz. Look-Up Table (LUT)-inversion, MLRA-GPR (Machine Learning Regression Algorithm- Gaussian Processes Regression) and empirical models, for estimating the LAI of tropical deciduous plantation using ARTMO (Automated Radiative Transfer Models Operator) tool and Sentinel-2 satellite images. The study was conducted in Central Tarai Forest Division, Haldwani, located in the Uttarakhand state, India. A total of 49 ESUs (Elementary Sampling Unit) of 30 m × 30 m size were established based on variability in composition and age of plantation stands. In-situ LAI was recorded using plant canopy imager during the leaf growing, peak and senescence seasons. The PROSAIL model was calibrated with site-specific biophysical and biochemical parameters before used to the predicted LAI. The plantation LAI was also predicted by an empirical approach using optimally chosen Sentinel-2 vegetation indices. In addition, Sentinel-2 and MODIS LAI products were evaluated with respect to LAI measurements. MLRA-GPR offered best results for predicting LAI of leaf growing (R2 = 0.9, RMSE = 0.14), peak (R2 = 0.87, RMSE = 0.21) and senescence (R2 = 0.86, RMSE = 0.31) seasons while LUT inverted model outperformed VI’s based parametric regression model. Vegetation indices (VIs) derived from 740 nm, 783 nm and 2190 nm band combinations of Sentinel-2 offered the best prediction of LAI. 相似文献
5.
积雪深度是大量气候、水文、农业及生态模型的重要输入变量。选用欧空局Sentinel-1主动微波数据,利用合成孔径雷达SAR(Synthetic Aperture Radar)差分干涉测量技术的二轨法,根据积雪相位与雪深之间的转换关系,反演新疆天山中段的巴音布鲁克盆地典型区的积雪雪深分布,提出了基于InSAR二轨差分的雪深估计方法,反演得到2016年12月18日的空间分辨率为13.89 m的雪深分布。研究表明:(1)对Sentinel-1数据进行正确的预处理以后,可以应用SAR差分干涉测量技术的二轨法反演区域雪深分布。但由于像对相干性和积雪状态的差异,积雪深度超过10 cm,可以获取较准确的雪深反演结果,R=0.65,反演误差的均方根误差RMSE=4.52 cm,平均相对误差为22.42%,反演雪深结果均比实测结果略偏低;而当雪深小于10 cm时,雪深反演值较实测值存在较大的误差,相对误差均高于34.52%,且反演雪深值均比实测值偏高。(2)雪深反演精度受高程及实际雪深的差异影响显著,另外雪深反演精度也受限于干涉像对相干性。结果表明,对于获取区域积雪雪深,InSAR技术较光学及被动微波遥感具有非常广阔的应用前景。 相似文献
6.
C.D. Lloyd P.M. Atkinson 《International Journal of Applied Earth Observation and Geoinformation》2004,5(4):293-305
Five techniques were used to map nitrogen dioxide (NO2) concentrations in the United Kingdom. The methods used to predict from point data, collected as part of the UK NO2 diffusion tube network, were local linear regression (LR), inverse distance weighting (IDW), ordinary kriging (OK), simple kriging with a locally varying mean (SKlm) and kriging with an external drift (KED). These techniques may be divided into two groups: (i) those that use only a single variable in the prediction process (IDW, OK) and (ii) those that make use of additional variables as a part of prediction (LR, SKlm and KED). Nitrous oxides emission data were used as secondary data with LR, SKlm and KED. It was concluded that SKlm provided the most accurate predictions based on the summary statistics of prediction errors from cross-validation. 相似文献
7.
在多光谱遥感水深反演研究中,由于影响反演精度的因素较多,传统的水深反演模型具有一定局限性。机器学习算法在解决非线性高复杂问题上较有优势,将其应用在某些特定区域水深反演可提高反演精度。本文利用Sentinel-2多光谱遥感影像和LiDAR测深数据,以瓦胡岛为研究区域,构建CatBoost水深反演模型,与传统水深反演模型及Boosting中的XGBoost和LightGBM模型的反演精度进行比较。试验结果表明,经过参数优化后的CatBoost水深反演模型的决定系数、均方根误差、平均绝对误差和平均相对误差分别为96.19%、1.09 m、0.77 m和9.61%,准确性最高,效果更佳。 相似文献
8.
Sentinel-2数据的冬小麦地上干生物量估算及评价 总被引：3，自引：0，他引：3
作物生物量快速精确的监测对于农业资源的合理利用与农田的精准管理具有重要意义。近年来,遥感技术因其独特的优势已被广泛用于作物生物量的估算中。本文主要针对不同宽波段植被指数在冬小麦生物量(文中的生物量均是指地上干生物量)估算方面的表现进行探索。首先利用欧洲空间局最新的Sentinel-2A卫星数据提取出17种常见的植被指数,之后分别构建其与相应时期内采集的冬小麦地上生物量间的最优估算模型,通过分析两者间的相关性与敏感性,获取适宜进行生物量估算的指数。最后,绘制了研究区的生物量空间分布图。结果表明,所选的植被指数均与生物量显著相关。其中,红边叶绿素指数(CI_(re))与生物量的估算精度最高(决定性系数R~2为0.83;均方根误差RMSE为180.29 g·m~(–2))。虽然相关性较高,但部分指数,如归一化差值植被指数(NDVI)等在生物量较高时会出现饱和现象,从而导致生物量的低估。而加入红边波段的指数不仅能够延缓指数的饱和趋势,而且能够提高反演精度。此外,通过敏感性分析发现,归一化差值指数和比值指数分别在作物生长的早期和中后期对生物量的变化保持较高的敏感性。由于红边比值指数(SR_(re))和MERIS叶绿素敏感指数(MTCI)在冬小麦全生长季内一直对生物量的变化保持高灵敏性,二者是生物量估算中最为稳定的指数。 相似文献
9.
《International Journal of Digital Earth》2013,6(2):187-206
Abstract Multi-frequency C and L bands in the TOPSAR data have been utilized to reconstruct three-dimensional (3-D) bathymetry pattern. The main objective of this study is to utilize fuzzy arithmetic to reduce the errors arising from speckle in synthetic aperture radar (SAR) data when constructing ocean bathymetry from polarized SAR data. In doing so, two 3-D surface models, the Volterra algorithm and a fuzzy B-spline (FBS) algorithm, which construct a global topological structure between the data points, were used to support an approximation to the real surface. Volterra algorithm was used to express the non-linearity of TOPSAR data intensity gradient based on the action balance equation (ABC). In this context, a first-order kernel of Volterra algorithm was used to express ABC equation. The inverse of Volterra algorithm then performed to simulate 2-D current velocities from CVV and LHH band. Furthermore, the 2-D continuity equation then used to estimate the water depth. In order to reconstruct 3-D bathymetry pattern, the FBS has been performed to water depth information which was estimated from 2-D continuity equation. The best reconstruction of coastal bathymetry of the test site in Kuala Terengganu, Malaysia, was obtained with polarized L and C bands SAR acquired with HH and VV polarizations, respectively. With 10 m spatial resolution of TOPSAR data, bias of –0.004 m, the standard error mean of 0.023 m, r 2 value of 0.95, and 90% confidence intervals in depth determination was obtained with LHH band. 相似文献
10.
《International Journal of Digital Earth》2013,6(12):1253-1269
ABSTRACTThis study investigates misregistration issues between Landsat-8/ Operational Land Imager and Sentinel-2A/ Multi-Spectral Instrument at 30?m resolution, and between multi-temporal Sentinel-2A images at 10?m resolution using a phase-correlation approach and multiple transformation functions. Co-registration of 45 Landsat-8 to Sentinel-2A pairs and 37 Sentinel-2A to Sentinel-2A pairs were analyzed. Phase correlation proved to be a robust approach that allowed us to identify hundreds and thousands of control points on images acquired more than 100 days apart. Overall, misregistration of up to 1.6 pixels at 30?m resolution between Landsat-8 and Sentinel-2A images, and 1.2 pixels and 2.8 pixels at 10?m resolution between multi-temporal Sentinel-2A images from the same and different orbits, respectively, were observed. The non-linear random forest regression used for constructing the mapping function showed best results in terms of root mean square error (RMSE), yielding an average RMSE error of 0.07?±?0.02 pixels at 30?m resolution, and 0.09?±?0.05 and 0.15?±?0.06 pixels at 10?m resolution for the same and adjacent Sentinel-2A orbits, respectively, for multiple tiles and multiple conditions. A simpler 1st order polynomial function (affine transformation) yielded RMSE of 0.08?±?0.02 pixels at 30?m resolution and 0.12?±?0.06 (same Sentinel-2A orbits) and 0.20?±?0.09 (adjacent orbits) pixels at 10?m resolution. 相似文献
11.
Forest canopy height is an important indicator of forest carbon storage, productivity, and biodiversity. The present study showed the first attempt to develop a machine-learning workflow to map the spatial pattern of the forest canopy height in a mountainous region in the northeast China by coupling the recently available canopy height (Hcanopy) footprint product from ICESat-2 with the Sentinel-1 and Sentinel-2 satellite data. The ICESat-2 Hcanopy was initially validated by the high-resolution canopy height from airborne LiDAR data at different spatial scales. Performance comparisons were conducted between two machine-learning models – deep learning (DL) model and random forest (RF) model, and between the Sentinel and Landsat-8 satellites. Results showed that the ICESat-2 Hcanopy showed the highest correlation with the airborne LiDAR canopy height at a spatial scale of 250 m with a Pearson’s correlation coefficient (R) of 0.82 and a mean bias of -1.46 m, providing important evidence on the reliability of the ICESat-2 vegetation height product from the case in China’s forest. Both DL and RF models obtained satisfactory accuracy on the upscaling of ICESat-2 Hcanopy assisted by Sentinel satellite co-variables with an R-value between the observed and predicted Hcanopy equalling 0.78 and 0.68, respectively. Compared to Sentinel satellites, Landsat-8 showed relatively weaker performance in Hcanopy prediction, suggesting that the addition of the backscattering coefficients from Sentinel-1 and the red-edge related variables from Sentinel-2 could positively contribute to the prediction of forest canopy height. To our knowledge, few studies have demonstrated large-scale vegetation height mapping in a resolution ≤ 250 m based on the newly available satellites (ICESat-2, Sentinel-1 and Sentinel-2) and DL regression model, particularly in the forest areas in China. Thus, the present work provided a timely and important supplementary to the applications of these new earth observation tools. 相似文献
12.
13.
Satellite Remote Sensing, with both optical and SAR instruments, can provide distributed observations of snow cover over extended and inaccessible areas. Both instruments are complementary, but there have been limited attempts at combining their measurements. We describe a novel approach to produce monthly maps of dry and wet snow areas through application of data fusion techniques to MODIS fractional snow cover and Sentinel-1 wet snow mask, facilitated by Google Earth Engine. The method is demonstrated in a 55,000 km2 river basin in the Indian Himalayan region over a period of ∼2.5 years, although it can be applied to any areas of the world where Sentinel-1 data are routinely available. The typical underestimation of wet snow area by SAR is corrected using a digital elevation model to estimate the average melting altitude. We also present an empirical model to derive the fractional cover of wet snow from Sentinel-1. Finally, we demonstrate that Sentinel-1 effectively complements MODIS as it highlights a snowmelt phase which occurs with a decrease in snow depth but no/little decrease in snowpack area. Further developments are now needed to incorporate these high resolution observations of snow areas as inputs to hydrological models for better runoff analysis and improved management of water resources and flood risk. 相似文献
14.
氮素是植被整个生命周期的必要元素,红树林冠层氮素含量（CNC）遥感估算对红树林健康监测具有重要意义。以广东湛江高桥红树林保护区为研究区,本文旨在基于Sentinel-2影像超分辨率重建技术进行红树林CNC估算和空间制图。研究首先基于三次卷积重采样、Sen2Res和SupReMe算法实现Sentinel-2影像从20 m分辨率到10 m的重建;然后以重建后的影像和原始20 m影像为数据源构建40个相关植被指数,采用递归特征消除法（SVM-RFE）确定CNC估算的最优变量组合,进而构建CNC反演的核岭回归（KRR）模型;最后选取最优模型实现CNC制图。研究结果表明：基于Sen2Res和SupReMe超分辨率算法的重建影像不仅与原始影像具有很高的光谱一致性,且明显提高了影像的清晰度和空间细节。红树林CNC反演波段主要集中在红（B4）、红边（B5）、近红外波段（B8a）以及短波红外波段（B11和B12）,与“红边波段”相关的植被指数（RSSI和TCARIre1/OSAVI）也是红树林CNC反演的有效变量。基于3种方法重建后10 m的影像构建的模型反演精度（R2val>0.579）均优于原始20 m的影像（R2val=0.504）;基于Sen2Res算法重建影像构建的反演模型拟合精度（R2val=0.630,RMSE_val=5.133,RE_val=0.179）与基于三次卷积重采样重建影像的模型拟合精度（R2val=0.640,RMSE_val=5.064,RE_val=0.179）基本相当,前者模型验证精度（R2cv=0.497,RMSE_cv=5.985,RE_cv=0.214）较高且模型变量选择数量最为合理。综合重建影像光谱细节及模型精度,基于Sen2Res算法重建的Sentinel-2影像在红树林CNC估算中具有良好的应用潜力,能为区域尺度红树林冠层健康状况的精细监测提供有效的方法借鉴和数据支撑。 相似文献
15.
In this study, we tested whether the inclusion of the red-edge band as a covariate to vegetation indices improves the predictive accuracy in forest carbon estimation and mapping in savanna dry forests of Zimbabwe. Initially, we tested whether and to what extent vegetation indices (simple ratio SR, soil-adjusted vegetation index and normalized difference vegetation index) derived from high spatial resolution satellite imagery (WorldView-2) predict forest carbon stocks. Next, we tested whether inclusion of reflectance in the red-edge band as a covariate to vegetation indices improve the model's accuracy in forest carbon prediction. We used simple regression analysis to determine the nature and the strength of the relationship between forest carbon stocks and remotely sensed vegetation indices. We then used multiple regression analysis to determine whether integrating vegetation indices and reflection in the red-edge band improve forest carbon prediction. Next, we mapped the spatial variation in forest carbon stocks using the best regression model relating forest carbon stocks to remotely sensed vegetation indices and reflection in the red-edge band. Our results showed that vegetation indices alone as an explanatory variable significantly (p < 0.05) predicted forest carbon stocks with R2 ranging between 45 and 63% and RMSE ranging from 10.3 to 12.9%. However, when the reflectance in the red-edge band was included in the regression models the explained variance increased to between 68 and 70% with the RMSE ranging between 9.56 and 10.1%. A combination of SR and reflectance in the red edge produced the best predictor of forest carbon stocks. We concluded that integrating vegetation indices and reflectance in the red-edge band derived from high spatial resolution can be successfully used to estimate forest carbon in dry forests with minimal error. 相似文献
16.
Lamin R. Mansaray Adam Sheka Kanu Lingbo Yang Jingfeng Huang Fumin Wang 《地理信息系统科学与遥感》2020,57(6):785-796
ABSTRACT Several machine learning regression models have been advanced for the estimation of crop biophysical parameters with optical satellite imagery. However, literature on the comparative performances of such models is still limited in range and scope, especially under multiple data sources, despite the potential of multi-source imagery to improving crop monitoring in cloudy areas. To fill in this knowledge gap, this study explored the synergistic use of Landsat-8, Sentinel-2A, China’s environment and disaster monitoring and forecasting satellites (HJ-1 A and B) and Gaofen-1 (GF-1) data to evaluate four machine learning regression models that include Random Forest (RF), Support Vector Machine (SVM), k-Nearest Neighbor (k-NN), and Gradient Boosting Decision Tree (GBDT), for rice dry biomass estimation and mapping. Taking a major rice cultivation area in southeast China as case study during the 2016 and 2017 growing seasons, a cross-calibrated time series of the Enhanced Vegetation Index (EVI) was obtained from the quad-source optical imagery and on which the aforementioned models were applied, respectively. Results indicate that in the before rice heading scenario, the most accurate dry biomass estimates were obtained by the GBDT model (R2 of 0.82 and RMSE of 191.8 g/m2) followed by the RF model (R2 of 0.79 and RMSE of 197.8 g/m2). After heading, the k-NN model performed best (R2 of 0.43 and RMSE of 452.1 g/m2) followed by the RF model (R2 of 0.42 and RMSE of 464.7 g/m2). Whist the k-NN model performed least in the before heading scenario, SVM performed least in the after heading scenario. These findings may suggest that machine learning regression models based on an ensemble of decision trees (RF and GBDT) are more suitable for the estimation of rice dry biomass, at least with optical satellite imagery. Studies that would extend the evaluation of these machine learning models, to other parameters like leaf area index, and to microwave imagery, are hereby recommended. 相似文献
17.
Roshanak Darvishzadeh Clement Atzberger Andrew Skidmore Martin Schlerf 《ISPRS Journal of Photogrammetry and Remote Sensing》2011,66(6):894-906
Statistical and physical models have seldom been compared in studying grasslands. In this paper, both modeling approaches are investigated for mapping leaf area index (LAI) in a Mediterranean grassland (Majella National Park, Italy) using HyMap airborne hyperspectral images. We compared inversion of the PROSAIL radiative transfer model with narrow band vegetation indices (NDVI-like and SAVI2-like) and partial least squares regression (PLS). To assess the performance of the investigated models, the normalized RMSE (nRMSE) and R2 between in situ measurements of leaf area index and estimated parameter values are reported. The results of the study demonstrate that LAI can be estimated through PROSAIL inversion with accuracies comparable to those of statistical approaches (R2 = 0.89, nRMSE = 0.22). The accuracy of the radiative transfer model inversion was further increased by using only a spectral subset of the data (R2 = 0.91, nRMSE = 0.18). For the feature selection wavebands not well simulated by PROSAIL were sequentially discarded until all bands fulfilled the imposed accuracy requirements. 相似文献
18.
Non-destructive and accurate estimation of crop biomass is crucial for the quantitative diagnosis of growth status and timely prediction of grain yield. As an active remote sensing technique, terrestrial laser scanning (TLS) has become increasingly available in crop monitoring for its advantages in recording structural properties. Some researchers have attempted to use TLS data in the estimation of crop aboveground biomass, but only for part of the growing season. Previous studies rarely investigated the estimation of biomass for individual organs, such as the panicles in rice canopies, which led to the poor understanding of TLS technology in monitoring biomass partitioning among organs. The objective of this study was to investigate the potential of TLS in estimating the biomass for individual organs and aboveground biomass of rice and to examine the feasibility of developing universal models for the entire growing season. The field plots experiments were conducted in 2017 and 2018 and involved different nitrogen (N) rates, planting techniques and rice varieties. Three regression approaches, stepwise multiple linear regression (SMLR), random forest regression (RF) and linear mixed-effects (LME) modeling, were evaluated in estimating biomass with extensive TLS and biomass data collected at multiple phenological stages of rice growth across the entire season. The models were calibrated with the 2017 dataset and validated independently with the 2018 dataset.The results demonstrated that growth stage in LME modeling was selected as the most significant random effect on rice growth among the three candidates, which were rice variety, growth stage and planting technique. The LME models grouped by growth stage exhibited higher validation accuracies for all biomass variables over the entire season to varying degrees than SMLR models and RF models. The most pronounced improvement with a LME model was obtained for panicle biomass, with an increase of 0.74 in R2 (LME: R2 = 0.90, SMLR: R2 = 0.16) and a decrease of 1.15 t/ha in RMSE (LME: RMSE =0.79 t/ha, SMLR: RMSE =2.94 t/ha). Compared to SMLR and RF, LME modeling yielded similar estimation accuracies of aboveground biomass for pre-heading stages, but significantly higher accuracies for post-heading stages (LME: R2 = 0.63, RMSE =2.27 t/ha; SMLR: R2 = 0.42, RMSE =2.42 t/ha; RF: R2 = 0.57, RMSE =2.80 t/ha). These findings implied that SMLR was only suitable for the estimation of biomass at pre-heading stages and LME modeling performed remarkably well across all growth stages, especially for post-heading. The results suggest coupling TLS with LME modeling is a promising approach to monitoring rice biomass at post-heading stages at high accuracy and to overcoming the saturation of canopy reflectance signals encountered in optical remote sensing. It also has great potential in the monitoring of other crops in cloud-cover conditions and the instantaneous prediction of grain yield any time before harvest. 相似文献
19.
The invasion by Striga in most cereal crop fields in Africa has posed a significant threat to food security and has caused substantial socioeconomic losses. Hyperspectral remote sensing is an effective means to discriminate plant species, providing possibilities to track such weed invasions and improve precision agriculture. However, essential baseline information using remotely sensed data is missing, specifically for the Striga weed in Africa. In this study, we investigated the spectral uniqueness of Striga compared to other co-occurring maize crops and weeds. We used the in-situ FieldSpec® Handheld 2™ analytical spectral device (ASD), hyperspectral data and their respective narrow-band indices in the visible and near infrared (VNIR) region of the electromagnetic spectrum (EMS) and four machine learning discriminant algorithms (i.e. random forest: RF, linear discriminant analysis: LDA, gradient boosting: GB and support vector machines: SVM) to discriminate among different levels of Striga (Striga hermonthica) infestations in maize fields in western Kenya. We also tested the utility of Sentinel-2 waveband configurations to map and discriminate Striga infestation in heterogenous cereal crop fields. The in-situ hyperspectral reflectance data were resampled to the spectral waveband configurations of Sentinel-2 using published spectral response functions. We sampled and detected seven Striga infestation classes based on three flowering Striga classes (low, moderate and high) against two background endmembers (soil and a mixture of maize and other co-occurring weeds). A guided regularized random forest (GRRF) algorithm was used to select the most relevant hyperspectral wavebands and vegetation indices (VIs) as well as for the resampled Sentinel-2 multispectral wavebands for Striga infestation discrimination. The performance of the four discriminant algorithms was compared using classification accuracy assessment metrics. We were able to positively discriminate Striga from the two background endmembers i.e. soil and co-occurring vegetation (maize and co-occurring weeds) based on the few GRRF selected hyperspectral vegetation indices and the GRRF selected resampled Sentinel-2 multispectral bands. RF outperformed all the other discriminant methods and produced the highest overall accuracy of 91% and 85%, using the hyperspectral and resampled Sentinel-2 multispectral wavebands, respectively, across the four different discriminant models tested in this study. The class with the highest detection accuracy across all the four discriminant algorithms, was the “exclusively maize and other co-occurring weeds” (>70%). The GRRF reduced the dimensionality of the hyperspectral data and selected only 9 most relevant wavebands out of 750 wavebands, 6 VIs out of 15 and 6 out of 10 resampled Sentinel-2 multispectral wavebands for discriminating among the Striga and co-occurring classes. Resampled Sentinel-2 multispectral wavebands 3 (green) and 4 (red) were the most crucial for Striga detection. The use of the most relevant hyperspectral features (i.e. wavebands and VIs) significantly (p ≤ 0.05) increased the overall classification accuracy and Kappa scores (±5% and ±0.2, respectively) in all the machine learning discriminant models. Our results show the potential of hyperspectral, resampled Sentinel-2 multispectral datasets and machine learning discriminant algorithms as a tool to accurately discern Striga in heterogenous maize agro-ecological systems. 相似文献
20.
海底底质分类对于海洋资源开发与利用、海洋科学研究等多方面具有重要意义。目前,多波束探测是实现大范围海底底质分类的有效手段之一,通常基于多波束反向散射强度提取角度响应(AR)特征及反向散射图像特征进行底质分类。由于特征来源较单一,分类器结构简单,往往分类精度不高。为此,本文提出了一种基于深层卷积神经网络(CNN)的多波束海底底质分类方法。除反向散射强度特征外,还利用地形特征,将特征向量转换为波形图,再输入卷积神经网络进行训练和分类。试验对比不同特征组合以及BP网络、支持向量机(SVM)、K近邻(KNN)、随机森林(RF)4种常规分类器,本文模型算法总体分类精度达到94.86%,Kappa系数为0.93,精度具有明显优势,效率也比较高。表明该方法有效利用两种数据类型所蕴含的海底底质信息,充分发挥卷积神经网络权值共享、高效率等特点,实现高分辨率海底底质分类,可对海底底质分类研究提供参考。 相似文献