Use of airborne multispectral imagery to discriminate and map weed infestations in a citrus orchard

Reliable information on weed abundance and distribution within fields is essential for weed management in agricultural systems. Such information is necessary to adopt localized and variable rates of herbicide spraying, thus reducing chemical waste, crop damage, and environmental pollution. This paper examined the potential of airborne multispectral imagery to discriminate and map weed infestations in an experimental citrus orchard in Japan. Using an airborne digital sensor, multispectral imagery was acquired over the study site on 10 April 2003. The obtained reflectance imagery was analyzed using an image object-based approach in eCognition. After creating image objects on the image, the spectral information for weeds and citrus, represented by corresponding selected sample image objects, was extracted. Significant differences in the spectral characteristics between weeds and citrus were observed in each of the red, green, and blue wavebands. The simple average values of these wavebands were used to classify image objects with the nearest neighbor algorithm. Maps were generated with different classes or levels of class groups. A subsequent accuracy assessment demonstrated that the weeds were successfully discriminated from other image objects with a classification accuracy of 99.07%. Therefore, maps generated based on the classification result could provide valuable information for developing a site-specific weed management program for the study orchard.     

Evaluating the accuracy of mapping weeds in fallow fields using airborne digital imaging: Panicumeffusum in oilseed rape stubble

The potential accuracy of using airborne multispectral imaging to map weed patches rapidly in a fallow field has been evaluated. An image of a field of oilseed rape ( Brassica napus L.) stubble interspersed with Panicum effusum R. Br. was acquired using a four-camera airborne digital imaging system; recording in the infrared, red, green and blue wavebands. The image was converted into georectified weed maps using supervised and unsupervised classification procedures. Comparison of the airborne-derived maps with an accurate weed map compiled from a detailed ground survey demonstrated that weed:non-weed classification and mapping accuracies of better than 87% are possible. The limitations of assessing the accuracy of classified imagery using ground-truth data of similar spatial resolution are discussed.     

Automated identification of sugar beet diseases using smartphones

Cercospora leaf spot (CLS) poses a high economic risk to sugar beet production due to its potential to greatly reduce yield and quality. For successful integrated management of CLS, rapid and accurate identification of the disease is essential. Diagnosis on the basis of typical visual symptoms is often compromised by the inability to differentiate CLS symptoms from similar symptoms caused by other foliar pathogens of varying significance, or from abiotic stress. An automated detection and classification of CLS and other leaf diseases, enabling a reliable basis for decisions in disease control, would be an alternative to visual as well as molecular and serological methods. This paper presents an algorithm based on a RGB‐image database captured with smartphone cameras for the identification of sugar beet leaf diseases. This tool combines image acquisition and segmentation on the smartphone and advanced image data processing on a server, based on texture features using colour, intensity and gradient values. The diseases are classified using a support vector machine with radial basis function kernel. The algorithm is suitable for binary‐class and multi‐class classification approaches, i.e. the separation between diseased and non‐diseased, and the differentiation among leaf diseases and non‐infected tissue. The classification accuracy for the differentiation of CLS, ramularia leaf spot, phoma leaf spot, beet rust and bacterial blight was 82%, better than that of sugar beet experts classifying diseases from images. However, the technology has not been tested by practitioners. This tool can be adapted to other crops and their diseases and may contribute to improved decision‐making in integrated disease control.     

Evaluating the potential of Unmanned Aerial Systems for mapping weeds at field scales: a case study with Alopecurus myosuroides

Mapping weed densities within crops has conventionally been achieved either by detailed ecological monitoring or by field walking, both of which are time‐consuming and expensive. Recent advances have resulted in increased interest in using Unmanned Aerial Systems (UAS ) to map fields, aiming to reduce labour costs and increase the spatial extent of coverage. However, adoption of this technology ideally requires that mapping can be undertaken automatically and without the need for extensive ground‐truthing. This approach has not been validated at large scale using UAS ‐derived imagery in combination with extensive ground‐truth data. We tested the capability of UAS for mapping a grass weed, Alopecurus myosuroides , in wheat crops. We addressed two questions: (i) can imagery accurately measure densities of weeds within fields and (ii) can aerial imagery of a field be used to estimate the densities of weeds based on statistical models developed in other locations? We recorded aerial imagery from 26 fields using a UAS . Images were generated using both RGB and R_mod (R_mod 670–750 nm) spectral bands. Ground‐truth data on weed densities were collected simultaneously with the aerial imagery. We combined these data to produce statistical models that (i) correlated ground‐truth weed densities with image intensity and (ii) forecast weed densities in other fields. We show that weed densities correlated with image intensity, particularly R_mod image data. However, results were mixed in terms of out of sample prediction from field‐to‐field. We highlight the difficulties with transferring models and we discuss the challenges for automated weed mapping using UAS technology.     

Predicting light leaf spot (Pyrenopeziza brassicae) risk on winter oilseed rape (Brassica napus) in England and Wales, using survey, weather and crop information

Data from surveys of winter oilseed rape crops in England and Wales in growing seasons with harvests in 1987–99 were used to construct statistical models to predict, in autumn (October), the incidence of light leaf spot caused by Pyrenopeziza brassicae on winter oilseed rape crops the following spring (March/April), at both regional and individual crop scales. Regions (groups of counties) with similar seasonal patterns of incidence (percentage of plants affected) of light leaf spot were defined by using principal coordinates analysis on the survey data. At the regional scale, explanatory variables for the statistical models were regional weather (mean summer temperature and mean monthly winter rainfall) and survey data for regional light leaf spot incidence (percentage of plants with affected pods) in July of the previous season. At the crop scale, further explanatory variables were crop cultivar (light leaf spot resistance rating), sowing date (number of weeks before/after 1 September), autumn fungicide use and light leaf spot incidence in autumn. Risk of severe light leaf spot (> 25% plants affected) in a crop in spring was also predicted, and uncertainty in predictions was assessed. The models were validated using data from spring surveys of winter oilseed rape crops in England and Wales from 2000 to 2003, and reasons for uncertainty in predictions for individual crops are discussed.     

Mapping Ridolfia segetum patches in sunflower crop using remote sensing

Ridolfia segetum is a frequent umbelliferous weed in sunflower crops in the Mediterranean basin. Field and remote sensing research was conducted in 2003 and 2004 over two naturally infested fields to determine the potential of multispectral imagery for discrimination and mapping of R. segetum patches in sunflower crops. The efficiency of the four wavebands blue (B), green (G), red (R) and near‐infrared (NIR), selected vegetation indices and the spectral angle mapper (SAM) classification method were studied using aerial photographs taken in the late vegetative (mid‐May), flowering (mid‐June) and senescence (mid‐July) crop growth stages. Discrimination efficiency of R. segetum patches in sunflower crops is consistently affected by their phenological stages, in this order: flowering > senescence > vegetative. In both fields, R. segetum patches were efficiently discriminated in mid‐June, corresponding to the flowering phase, by using the waveband G, the ratio R/B or SAM with overall accuracies ranging from 85% to 98%. The application of the median‐filtering algorithm to any of the classified images improved the accuracy. Our results suggest that mapping R. segetum weed patches in sunflower to implement site‐specific weed management techniques is feasible with aerial photography when images are taken from 8 to 10 weeks before harvesting.     

Multispectral classification of grass weeds and wheat (Triticum durum) using linear and nonparametric functional discriminant analysis and neural networks

Field studies were conducted to determine the potential of multispectral classification of late‐season grass weeds in wheat. Several classification techniques have been used to discriminate differences in reflectance between wheat and Avena sterilis, Phalaris brachystachys, Lolium rigidum and Polypogon monspeliensis in the 400–900 nm spectrum, and to evaluate the accuracy of performance for a spectral signature classification into the plant species or group to which it belongs. Fisher’s linear discriminant analysis, nonparametric functional discriminant analysis and several neural networks have been applied, either with a preliminary principal component analysis (PCA) or not and in different scenarios. Fisher’s linear discriminant analysis, feedforward neural networks and one‐layer neural network, all showed classification percentages between 90% and 100% with PCA. Generally, a preliminary computation of the most relevant principal components considerably improves the correct classification percentage. These results are promising because A. sterilis and L. rigidum, two of the most problematic, clearly patchy and expensive‐to‐control weeds in wheat, could be successfully discriminated from wheat in the 400–900 nm range. Our results suggest that mapping grass weed patches in wheat could be feasible with analysis of real‐time and high‐resolution satellite imagery acquired in mid‐May under these conditions.     

基于Sentinel-2影像的河套灌区作物种植结构提取

作物识别是提取作物种植结构的基础,利用遥感技术对作物进行监测识别,对优化生产布局、调整农业生产模式有着重要意义。文中选取河套灌区杭锦后旗为研究区域,基于2019年覆盖生长周期的Sentinel-2号卫星影像数据,构建NDVI时间序列数据集,利用Savitzky-Golay(S-G)滤波对NDVI时间序列数据集进行平滑,分析不同作物不同发育期的光谱曲线特征,计算各主要作物识别关键期的光谱阈值,构建基于决策树分层分类的农作物种植面积提取模型,并用验证样本对分类结果进行精度验证。结果表明:利用整个生育期内的NDVI最大合成影像确定植被地表覆盖,NDVI曲线变化区别林地与耕地,逐层提取地物,简便易行;采用S-G滤波重构高质量的NDVI时间序列曲线,研究证明重构后曲线更加平滑符合作物生长趋势;基于Sentinel-2号遥感数据和整个生育期NDVI时序数据,构建分层分类决策树模型,作物分类总体精度达92.1%,Kapppa系数精度达0.857。本研究采用的方法满足遥感观测应用化需求,也为县级区域农作物分类提供重要参考价值。     

Spectral discrimination of Ridolfia segetum and sunflower as affected by phenological stage

Ridolfia segetum is an umbelliferous weed frequent and abundant in sunflower crops in the Mediterranean basin. Field research was conducted to evaluate the potential of hyperspectral and multispectral reflectance and five vegetation indices in the visible to near infrared spectral range, for discriminating bare soil, sunflower and R. segetum at different phenological stages. This was a preliminary step for mapping R. segetum patches in sunflower using remote sensing for herbicide application decisions. Reflectance data were collected at three sampling dates (mid‐May, mid‐June and mid‐July, corresponding to vegetative‐early reproductive, flowering and senescent phenological stages respectively) using a handheld field spectroradiometer. Differences observed in hyperspectral reflectance curves were statistically significant within and between crop and weed phenological stages depending on sampling date, which facilitates their discrimination. Statistically significant differences in the multispectral and vegetation indices analysis showed that it is also possible to distinguish any of the classes studied. Our study provides some information for constructing the spectral libraries of sunflower and R. segetum in which the different phenological stages co‐existing in the field were considered. Hyperspectral and multispectral results suggest that mapping R. segetum patches in sunflower is feasible using airborne hyperspectral sensors, and high‐resolution satellite imagery or aerial photography, respectively, taking into account specific timeframes.     

Discrimination of Ambrosia artemisiifolia and Artemisia vulgaris by hyperspectral image analysis during the growing season

Ambrosia artemisiifolia (ragweed) is an invasive plant in Europe. An optical detection system for effective monitoring requires differences in the spectral reflectance properties compared with other plant species. A. artemisiifolia often occurs together with Artemisia vulgaris (mugwort). Both plant species are of the Asteraceae family and they are almost indistinguishable in appearance. With the help of hyperspectral image analysis, a method was developed to determine characteristic wavelengths for their classification. High‐resolution hyperspectral images (400–1000 nm) were generated indoors. The factors measured were two weed species, two tissue classes (leaf and stem) and three growth stages (rosette growth, inflorescence emergence and fruit development). Only the stems of A. artemisiifolia in the fruit development stage showed different reflectance behaviour compared with its leaves and with the stems and leaves of A. vulgaris. At wavelengths ranging from 550 to 650 nm, the reflectance increased, and then at wavelengths up to 680 nm, the reflectance decreased. The other tissue classes showed constantly decreasing spectral reflectance from 550 to 680 nm. In the two other early growth stages, the reflectance of all four tissue classes decreased similarly. Thus, using two wavelengths of 550 and 650 nm, classification between A. artemisiifolia and A. vulgaris at fruit development was achieved. The findings could be a first step to develop an optical outdoor detection system to identify hot spots of A. artemisiifolia.     

基于地表高光谱与OLI影像的土壤含盐量和pH值估测

针对宁夏银北地区大面积土壤盐碱化监测的需要,利用实测植被冠层光谱与Landsat 8 OLI影像相结合进行土壤含盐量和pH值估测研究.对实测植被冠层高光谱与影像多光谱反射率进行倒数、对数、三角函数及其一阶微分等一系列变换,确定最佳光谱变换形式,筛选敏感植被指数和敏感波段,分别建立基于实测植被光谱与Landsat 8 O...     

不同氮水平下冬小麦农学参数与光谱植被指数的相关性

利用光谱仪通过大田试验测量不同氮素水平及不同生育期冬小麦冠层的光谱反射率,测算叶面积指数(LAI)、叶绿素含量(CHL)、叶绿素密度(CHL.D)、地上鲜生物量和地上干生物量等农学参数;在此基础上分析了不同氮素水平冬小麦生育期内的光谱植被指数的变化,并分析了农学参数与植被指数之间的相关性。结果表明:小麦叶面积指数、叶绿素密度与比值植被指数(RVI)和归一化差值植被指数(NDVI)在各生育期呈显著相关,小麦叶片的叶绿素含量与RVI、NDVI在抽穗期呈极显著相关,而地上鲜生物量、地上干生物量与RVI和NDVI从起身到孕穗期呈显著相关。     

高分辨率影像的干旱区植被信息提取方法

基于黑河下游额济纳旗地区的Quickbird影像,采用决策树(Decision Tree)、人工神经网络(Artifi-cial neural net,ANN)及支持向量机(Support Vector Machine,SVM)方法对干旱区植被信息进行提取。对三种方法的精度进行评价,结果显示:决策树分类得到的结果零碎,总体分类精度为84.87%;ANN法较决策树方法适宜度高,总体分类精度为91.87%;纹理信息辅助的SVM法取得效果最好,总体分类精度可达96.53%。试验中发现使用高分辨率影像提取干旱区植被种类信息时,大窗口的纹理特征辅助效果较好,但是分类结果的边界出现失常,随着纹理窗口越大,失常的范围也越大。     

Assessment of leaf cover and crop soil cover in weed harrowing research using digital images

Objective assessment of crop soil cover, defined as the percentage of leaf cover that has been buried in soil because of weed harrowing, is crucial to further progress in post‐emergence weed harrowing research. Up to now, crop soil cover has been assessed by visual scores, which are biased and context‐dependent. The aim of this study was to investigate whether digital image analysis is a feasible method to estimate crop soil cover in the early growth stages of cereals. Two main questions were examined: (i) how to capture suitable digital images under field conditions with a standard high‐resolution digital camera and (ii) how to analyse the images with an automated digital image analysis procedure. The importance of light conditions, camera angle, size of recorded area, growth stage and direction of harrowing were investigated, in order to establish a standard for image capture and an automated image analysis procedure based on the excess green colour index was developed. The study shows that the automated digital image analysis procedure provided reliable estimations of leaf cover, defined as the proportion of pixels in digital images determined to be green, which were used to estimate crop soil cover. A standard for image capture is suggested and it is recommended that digital image analysis be used to estimate crop soil cover in future research. The prospects of using digital image analysis in future weed harrowing research are discussed.     

民勤绿洲物候季节划分及景观季相特征

本文根据甘肃民勤沙生植物园 1 974- 2 0 0 2年的物候观测资料 ,应用“植物物候频率分布型法”对民勤绿洲物候季节和季相特征进行了研究 ,通过分析 ,民勤绿洲物候季节可分为 4个大季段和 1 2个小季段 ,春、夏、秋、冬四大季段中冬长秋短 ,春夏居中 ,每个季段都具有明显的指示物候、典型的植物物候形态组合以及独特的季相特征。这一研究结果客观地反映了该地区自然季节更替的规律和景观季相的演变特点。     

Plant classification for weed detection using hyperspectral imaging with wavelet analysis

The goal of this study is to develop a new weed detection method that can be applied for automatic mechanical weed control. For successful weed detection, plants must be classified into crops and weeds according to their species. In this study, we employed a portable hyperspectral imaging system. The hyperspectral camera can capture landscape images that include crops, weeds, and the soil surface, and can provide more extensive information than conventional red, green, and blue (RGB) images. Although RGB images consist of red, green, and blue wavebands, the obtained hyperspectral images consist of 240 wavebands of spectral information. Hyperspectral imaging is expected to provide powerful technology for agricultural sensing. In the initial step of this study, the image pixels of the plants (crop or weeds) were segmented from the background soil surface using Euclidean distance as the discriminant function. In the next step, the image pixels of the crop (sugarbeet) and weeds (four species) were classified using the difference in the spectral characteristics of the plant species. In this process, classification variables were generated using wavelet transformation for data compression, noise reduction, and feature extraction, and then stepwise linear discriminant analysis was applied. The validation results indicate that the developed classification method has potential for practical use.     

Assessment of Dothistroma Needle Blight of Pinus radiata Using Airborne Hyperspectral Imagery

ABSTRACT Dothistroma needle blight is a serious foliar disease in Australian Pinus radiata plantations causing defoliation, decreased productivity and, in extreme cases, tree death. Conventional methods of monitoring forest health such as aerial survey and ground assessments are labor intensive, time consuming, and subjective. Remote sensing provides a synoptic view of the canopy and can indicate areas affected by damaging agents such as pests and pathogens. Hyperspectral airborne remote sensing imagery (CASI-2) was acquired over pine stands in southern New South Wales, Australia which had been ground assessed and ranked on an individual tree basis, according to the extent of Dothistroma needle blight. A series of spectral indices were tested using two different approaches for extracting crown-scale reflectance measurements and relating these to ground-based estimates of severity. Dothistroma needle blight is most severe in the lower crown and statistically significant relationships were found between crown reflectance values and ground estimates using a 'halo' approach (which ignored each tree crown's brightest central pixels). Independent accuracy assessment of the method indicated that the technique could successfully detect three levels of Dothistroma needle blight infection with an accuracy of over 70%.     

基于同化叶面积指数和条件植被温度指数的冬小麦单产估测

应用数据同化方法将遥感信息与作物生长模型融合,是估测区域作物产量的重要方法之一。以2008—2014年越冬后的冬小麦为研究对象,选择与作物长势、产量及水分胁迫信息密切相关的叶面积指数(LAI)和条件植被温度指数(VTCI),采用粒子滤波算法对CERES-Wheat模型模拟和遥感数据观测的LAI和VTCI实施同化,分别基于观测LAI和VTCI、同化LAI和VTCI构建冬小麦单产估测模型。结果表明,同化LAI变化趋势更加符合关中平原冬小麦的实际生长状况,同化VTCI能更好地反映冬小麦的水分胁迫程度。应用观测LAI和VTCI构建的估产模型决定系数为0.402,而单独应用LAI或VTCI单变量构建的估产模型决定系数分别为0.279和0.339,说明应用LAI和VTCI双变量构建的估产模型的精度优于单独应用LAI或VTCI单变量的精度。相比于观测LAI和VTCI构建的估产模型,基于同化LAI和VTCI构建的估产模型的决定系数从0.402提高到0.547。表明基于同化LAI和VTCI构建的估产模型的精度明显提高。     

Measuring lesion attributes and analysing their spatial patterns at the leaf scale using digital image analysis

Digital image analysis was used to quantify size, shape and relative positions of individual plant disease lesions to determine their spatial distribution pattern at the leaf scale. Rice brown spot was used as a necrotrophic pathogen causing numerous discrete lesions. A 50‐leaf subsample was selected from an existing data set of 350 images of leaves taken from the field, and analysed for disease severity using image analysis. Further measurements included size, shape and the relative positions of lesions for all leaves with severity > 8% (n = 25) and an additional 25‐leaf sample with severity <8%. A total of 3964 necrotic and/or halo areas were selected using a manually defined threshold in the computer program Assess . There were significant and positive associations (Pearson's r > 0.81; P < 0.001) between the size‐related measurements (lesion area, longest and shortest axis). Coalesced areas, formed by interconnection of lesions and associated haloes, and a high number of small lesions were found with an increase in severity, suggesting a secondary cycle and autoinfection process. Results from quadrat‐based (Poisson distribution and Spatial Analysis by Distance IndicEs) and distance‐based (point‐process Poisson) spatial methods were in good agreement and, together with a Taylor power law model, suggested a shift from random to predominantly aggregated patterns of lesions at severities approaching 10%. This framework, which is applicable to other foliar diseases, proved useful in providing quantitative knowledge of epidemic processes at the leaf scale. Finally, these results may be useful in improving simulation models and disease assessment methods.     

基于无人机多光谱影像的完熟期玉米倒伏信息提取

以吉林省梨树县的玉米试验田为研究区，按受灾后完熟期玉米的状态将研究区分为倒伏、半倒伏和未倒伏3种类型。基于无人机采集的多光谱影像提取15种光谱指数和8种纹理特征，采用面向对象法、最大似然法和多元Logistic回归模型进行玉米倒伏信息的提取；而后通过目视方法选取400个样本点进行玉米倒伏信息提取结果的精度验证。结果表明：面向对象法精度最高，对玉米3种倒伏状态信息识别的总体精度为88.13%,Kappa系数为0.83。研究用于区分倒伏与未倒伏玉米的最佳光谱指数是归一化差异植被指数，对区分倒伏与半倒伏、半倒伏与未倒伏玉米贡献最大的特征均为对比度纹理特征。研究表明基于无人机多光谱影像的面向对象方法在对田块尺度玉米倒伏信息的精准识别中具有较大潜力。