Early detection and classification of plant diseases with Support Vector Machines based on hyperspectral reflectance

Automatic methods for an early detection of plant diseases are vital for precision crop protection. The main contribution of this paper is a procedure for the early detection and differentiation of sugar beet diseases based on Support Vector Machines and spectral vegetation indices. The aim was (I) to discriminate diseased from non-diseased sugar beet leaves, (II) to differentiate between the diseases Cercospora leaf spot, leaf rust and powdery mildew, and (III) to identify diseases even before specific symptoms became visible. Hyperspectral data were recorded from healthy leaves and leaves inoculated with the pathogens Cercospora beticola, Uromyces betae or Erysiphe betae causing Cercospora leaf spot, sugar beet rust and powdery mildew, respectively for a period of 21 days after inoculation. Nine spectral vegetation indices, related to physiological parameters were used as features for an automatic classification. Early differentiation between healthy and inoculated plants as well as among specific diseases can be achieved by a Support Vector Machine with a radial basis function as kernel.The discrimination between healthy sugar beet leaves and diseased leaves resulted in classification accuracies up to 97%. The multiple classification between healthy leaves and leaves with symptoms of the three diseases still achieved an accuracy higher than 86%. Furthermore the potential of presymptomatic detection of the plant diseases was demonstrated. Depending on the type and stage of disease the classification accuracy was between 65% and 90%.     

Spectral signatures of sugar beet leaves for the detection and differentiation of diseases

This study examines the potential of hyperspectral sensor systems for the non-destructive detection and differentiation of plant diseases. In particular, a comparison of three fungal leaf diseases of sugar beet was conducted in order to facilitate a simplified and reproducible data analysis method for hyperspectral vegetation data. Reflectance spectra (400–1050 nm) of leaves infected with the fungal pathogens Cercospora beticola, Erysiphe betae, and Uromyces betae causing Cercospora leaf spot, powdery mildew and rust, respectively, were recorded repeatedly during pathogenesis with a spectro-radiometer and analyzed for disease-specific spectral signatures. Calculating the spectral difference and reflectance sensitivity for each wavelength emphasized regions of high interest in the visible and near infrared region of the spectral signatures. The best correlating spectral bands differed depending on the diseases. Spectral vegetation indices related to physiological parameters were calculated and correlated to the severity of diseases. The spectral vegetation indices Normalised Difference Vegetation Index (NDVI), Anthocyanin Reflectance Index (ARI) and modified Chlorophyll Absorption Integral (mCAI) differed in their ability to assess the different diseases at an early stage of disease development, or even before first symptoms became visible. Results suggested that a distinctive differentiation of the three sugar beet diseases using spectral vegetation indices is possible using two or more indices in combination.     

基于深度学习和支持向量机的4种苜蓿叶部病害图像识别

为实现苜蓿叶部病害的快速准确诊断和鉴别,基于图像处理技术,对常见的4种苜蓿叶部病害(苜蓿褐斑病、锈病、小光壳叶斑病和尾孢菌叶斑病)的识别方法进行探索。对采集获得的899张苜蓿叶部病害图像,利用人工裁剪方法从每张原始图像中获得1张子图像,然后利用结合K中值聚类算法和线性判别分析的分割方法进行病斑图像分割,得到4种病害的典型病斑图像(每张典型病斑图像中仅含有1个病斑)共1 651张。基于卷积神经网络提取病斑图像特征,建立病害识别支持向量机(Support vector machine,SVM)模型。结果表明:当病斑图像尺寸归一化为32×32像素,利用归一化的特征HSV(即特征H、特征S和特征V归一化后的组合特征)构建的病害识别SVM模型最优,其训练集识别正确率为94.91%,测试集识别正确率为87.48%。本研究基于深度学习和SVM所建立的病害识别模型可用于识别上述4种苜蓿叶部病害。     

基于特征融合的随机森林模型茶鲜叶分类

目的 解决机采茶鲜叶中混有不同等级的茶叶,且混杂度高、物理特征分类精确度低的问题。方法 利用随机森林分类模型,提出一种基于颜色和边缘特征融合的方法。试验采集3种不同等级的茶鲜叶,对原始图像进行裁剪、尺寸归一化和去噪等处理,再进行颜色特征和边缘特征提取。通过参数的修改和测试,构建最优的随机森林分类模型,并且同K最近邻、SVM分类器进行对比试验。结果 特征融合之后随机森林模型的分类准确率达到99.45%,比单一颜色特征和边缘特征的分类准确率分别高7.14和9.34个百分点;比K最近邻和SVM分类器准确率分别高15.38和5.49个百分点。结论 所建立的方法能够对茶鲜叶单芽、一芽一叶、一芽二叶进行精确的分类。     

Plant species identification using Elliptic Fourier leaf shape analysis

Elliptic Fourier (EF) and discriminant analyses were used to identify young soybean (Glycine max (L.) merrill), sunflower (Helianthus pumilus), redroot pigweed (Amaranthus retroflexus) and velvetleaf (Abutilon theophrasti Medicus) plants, based on leaf shape. Chain encoded, Elliptic Fourier harmonic functions were generated based on leaf boundary. A complexity index of the leaf shape was computed using the variation between consecutive EF functions. Principle component analysis was used to select the Fourier coefficients with the best discriminatory power. Canonical discriminant analysis was used to develop species identification models based on leaf shapes extracted from plant color images during the second and third weeks after germination. The classification results showed that plant species during the third week were successfully identified with an average of correct classification rate of 89.4%. The discriminant model correctly classified on average: 77.9% of redroot pigweed, 93.8% of sunflower, 89.4% of velvetleaf and 96.5% of soybean. Using all of the leaves extracted from the second and the third weeks, the overall classification accuracy was 89.2%. The discriminant model correctly classified 76.4% of redroot pigweed, 93.6% of sunflower, 81.6% of velvetleaf, 91.5% of soybean leaf extracted from trifoliolate and 90.9% of soybean unifoliolate leaves. The Elliptic Fourier shape feature analysis could be an important and accurate tool for weed species identification and mapping.     

基于改进VGG卷积神经网络的棉花病害识别模型

为实现自然条件下棉花病害图像准确分类,提出基于改进VGG-16卷积神经网络的病害识别模型。该模型在VGG-16网络模型基础上,优化全连接层层数,并用6标签SoftMax分类器替换原有VGG-16网络中的SoftMax分类器,优化了模型结构和参数,通过微型迁移学习共享预训练模型中卷积层与池化层的权值参数。从构建的棉花病害图像库中随机抽取病害图像样本作为训练集和测试集,用以测试该方法的性能。试验结果表明:该模型能有效提取出棉花病害叶片图像的多层特征图像,并通过Relu激活函数的处理更能凸显棉花病害的边缘信息与纹理信息,分辨率为512像素×512像素图像在样本训练与验证试验效果最好。在平均识别准确率方面,本研究模型较BP神经网络、支持向量机、AlexNET、GoogleNET、VGG-16NET效果最好,达到89.51%,实现对棉花的褐斑病、炭疽病、黄萎病、枯萎病、轮纹病、正常叶片的准确区分。该模型在棉花病害识别领域具备良好的分类性能,可实现自然条件下棉花病害的准确识别。     

玄参叶的形态解剖学研究

目的 为玄参叶的开发利用和质量控制提供科学依据。方法 按生药学常规方法对玄参叶进行形态解剖学研究,其中组织切片采用徒手切片和石蜡切片相结合的方法,叶表面制片采用撕取表皮的方法,叶表面扫描电镜观察采用脱水、干燥、喷金后常规扫描的方法,组织、叶表面、粉末图均采用显微拍摄。结果 找出了玄参叶的性状和显微鉴别特征。结论 此研究可作为鉴定玄参叶的依据。     

深度学习在大豆叶片图像数据管理中的识别与分类研究

[目的/意义]为了提高大豆叶片图像的分类精度与效率,进一步对大豆叶片图像进行存储与管理。[方法/过程]本文利用深度学习方法,针对肉眼观察准确率较低且不同人群分类结果差异较大的大豆叶片图像数据提出了一种自动分类方法。本研究首先对大豆叶片进行ROI感兴趣区域划分,进而利用分水岭分割方法对大豆叶片进行提取,最后通过深度学习高效精确的实现了大豆叶片的分类识别。[结果/结论]通过分析大豆叶片形态图像特点后,基于深度学习开展了对大豆叶片形态的分类识别的研究,达到了较高的识别准确率。     

Use of imaging spectroscopy to discriminate symptoms caused by Heterodera schachtii and Rhizoctonia solani on sugar beet

Diseases caused by nematodes and non-sporulating soil-borne fungi have low mobility and are likely to be suitable targets for precision agriculture applications. Sensors which assess the reflectance of plant leaves may be useful tools to detect soil-borne pathogens. The development of symptoms caused by the plant parasitic nematode Heterodera schachtii and the fungal pathogen Rhizoctonia solani anastomosis group 2-2IIIB alone or in combination was studied by leaf reflectance recorded with a hyperspectral imaging system (range 400–1000 nm) for 9 weeks twice per week. Three image processing methods were tested for their suitability to generate the most sensitive spectral information for disease detection. Nine spectral vegetation indices were calculated from spectra to correlate them to leaf symptom recordings. Supervised classification by spectral angle mapper was tested for the discrimination of leaf symptoms caused by the diseases. The symptoms of Rhizoctonia crown and root rot caused by R. solani and symptoms caused by H. schachtii induced modifications that could be detected by hyperspectral image analysis. Rhizoctonia crown and root rot symptom development in mixed inoculations was faster and more severe than in single inoculations, indicating complex interactions among fungus, nematode and plant. The results from this study under controlled conditions are currently used to transfer the sensor technology to the field.     

土沉香对黄野螟的抗性研究

【目的】黄野螟Heortia vitessoides是珍贵树种土沉香Aquilaria sinensis的重要食叶害虫,通过大面积林间调查土沉香受害情况,筛选可能存在的抗虫植株,为黄野螟的科学预防和土沉香抗虫品种的选育奠定基础。【方法】在黄野螟危害盛期,对野外土沉香林定期进行大面积调查,在严重受害的土沉香林中,观察不同受害水平土沉香的外观形态和叶片物理结构,同时采集具有不同抗虫性植株的叶片饲养黄野螟幼虫,观察初孵幼虫对不同抗性植株叶片的选择和拒食情况,取食不同抗性土沉香叶片后,测定黄野螟幼虫存活率、生长发育、化蛹和羽化的差异。【结果】在土沉香严重受害的林分中发现了2株未受害的土沉香植株,其表现出较好的抗虫性(抗1和抗2)。抗性植株(抗1和抗2)与感虫植株在叶片长度和厚度上差异显著(P0.05),而叶片长宽比无显著差异。在叶片物理结构上,抗2土沉香叶片的上表皮角质层厚度显著高于感虫叶片。抗2土沉香植株对幼虫取食抑制率高于抗1土沉香植株,两者均达44.81%以上。强迫取食抗性土沉香试验的幼虫存活率、成虫羽化率、蛹质量、成虫寿命均显著低于取食感虫土沉香叶片幼虫的相应指标,而取食抗性土沉香的幼虫、蛹发育历期均显著长于取食感虫土沉香叶片的害虫。【结论】叶片嫩绿的土沉香植株较易受黄野螟的为害,而叶片厚的或叶片颜色偏黄、墨绿的土沉香植株对黄野螟具有较强的抗性。抗性土沉香植株对黄野螟幼虫取食活性具有较强的抑制作用,对幼虫的发育有阻碍作用。     

广州附近的草莓病害

 草莓在广州附近的主要病害为轮斑病(Phomopsis obscurans)、灰霉病(Botrytis cinerea)、灰斑病(Cercospora fagarina sp.nov.)、叶霉病(Cladosporium cladosporioides)、基腐病(Rhizoctonia solani)、软腐病(Rhizopus stolonifer,R.sexualis)和褐斑病(Pestalotiopsis adusta)等7种,尤以第一、二种危害严重,Cercospora fragarina是个新种,Rhizopus sexualis是国内新纪录,草莓在广州附近虽试种成功,今后扩大栽培,不可不注意病害的防治。     

Development,evaluation and application of a spectral knowledge base to detect yellow rust in winter wheat

In most cases, statistical models for monitoring the disease severity of yellow rust are based on hyperspectral information. The high cost and limited cover of airborne hyperspectral data make it impossible to apply it to large scale monitoring. Furthermore, the established models of disease detection cannot be used for most satellite images either because of the wide range of wavelengths in multispectral images. To resolve this dilemma, this paper presents a novel approach by constructing a spectral knowledge base (SKB) of diseased winter wheat plants, which takes the airborne images as a medium and links the disease severity with band reflectance from environment and disaster reduction small satellite images (HJ-CCD) accordingly. Through a matching process with a SKB, we estimated the disease severity with a disease index (DI) and degrees of disease severity. The proposed approach was validated against both simulated data and field surveyed data. Estimates of DI (%) from simulated data were more accurate, with a coefficient of determination (R ²) of 0.9 and normalized root mean square error (NRMSE) of 0.2. The overall accuracy of classification reached 0.8, with a kappa coefficient of 0.7. Validation of the estimates against field measurements showed that there were some errors in the DI value with the NRMSE close to 0.5. The result of the classification was more encouraging with an overall accuracy of 0.77 and a kappa coefficient of 0.58. For the matching process, Mahalanobis distance performed better than the spectral angle (SA) in all analyses in this study. The potential of SKB for monitoring the incidence and severity of yellow rust is illustrated in this study.     

基于图像处理技术的四种苜蓿叶部病害的识别

基于图像处理技术,对4种苜蓿叶部病害进行识别研究。利用结合K中值聚类算法和线性判别分析的分割方法对病斑图像作分割,获得了较好的分割效果。结果表明:该分割方法在由4种病害图像数据集整合成的汇总图像数据集上综合得分的平均值和中值分别为0.877 1和0.899 7;召回率的平均值和中值分别为0.829 4和0.851 4;准确率的平均值和中值分别为0.924 9和0.942 4。进一步提取病斑图像的颜色特征、形状特征和纹理特征共计129个,利用朴素贝叶斯方法和线性判别分析方法建立病害识别模型,并结合顺序前向选择方法实现特征筛选,分别获得最优特征子集;同时利用这2个最优特征子集,结合支持向量机(Support vector machine,SVM)建立病害识别模型。比较各模型的识别效果,发现利用所建线性判别分析模型下的最优特征子集,结合SVM建立的病害识别模型识别效果最好,训练集识别正确率为96.18%,测试集识别正确率为93.10%。由此可见,本研究所建基于图像处理技术的病害识别模型可用于识别上述4种苜蓿叶部病害,为苜蓿病害的诊断和鉴别提供了一定依据。     

桑叶叶际和家蚕肠道细菌群落结构及功能分析

家蚕是我国重要的经济昆虫,主要以桑叶为食,了解家蚕肠道及所食桑叶叶际微生物对减少家蚕病害具有重要意义。以5龄第3天的家蚕幼虫和所食桑叶为研究对象,利用高通量测序技术对家蚕肠道和叶际细菌的16S rDNA V3-V4区序列进行测序分析。结果表明：桑叶叶际和家蚕肠道各部位细菌的多样性指数（Shannon指数和Simpson指数）无显著差异（P > 0.05）;NMDS分析显示桑叶叶际细菌与家蚕肠道细菌各自聚类,家蚕肠道各部分的细菌菌群无显著差异。桑叶叶际和家蚕肠道中共注释到的主要分类阶元有29 个门、268 个属：前4个优势门为变形菌门（Proteobacteria）、厚壁菌门（Firmicutes）、拟杆菌门（Bacteroidetes）和放线菌门（Actinobacteria）;芽孢杆菌属Bacillus为桑叶叶际细菌的最优势属,泛菌属Pantoea为家蚕前肠细菌的最优势属,鞘氨醇单胞菌属Sphingomonas为家蚕中肠和后肠细菌的最优势属。通过功能预测可知,消化系统、免疫系统、能量代谢、代谢性疾病和传染性疾病在家蚕的前肠、中肠中占比较高,氨基酸代谢、碳水化合物代谢、神经退行性疾病、免疫系统疾病、脂类代谢、其他氨基酸代谢、心血管疾病和核苷酸代谢在家蚕的后肠中占比较高。家蚕肠道细菌群落与叶际细菌群落前4个优势菌门一致;家蚕肠道的前肠和中肠功能菌群相似,与后肠差异较大。     

Tectona grandis canopy cover predicted by remote sensing

The phytosanitary status of Tectona grandis plantations are monitored conventionally with periodic data collection in the field, which is often costly and has low efficiency. The objective of this research was to develop a methodology to predict the canopy cover of T. grandis plantations using multispectral images of the Sentinel-2 (S2) satellite and photographic imagery. The study was carried out in a T. grandis plantation of seminal origin, in Cáceres, Mato Grosso state, Brazil. Hemispherical photographic (HP) images of the plant canopy were obtained with a digital camera coupled to a “fisheye” lens fixed at 1.3 m high at two dates in the rainy and the dry season. Cloudless and no shadow images of the S2 satellite bands were concurrently obtained with the field images. Multivariate permutative analysis of variance (PERMANOVA) and partial least squares regression (PLSR) were used to predict canopy cover percentage. The accuracy of the predicted T. grandis canopy cover (%) by the PLSR model approach was 77.8?±?0.09%. The results indicate that a PLS model calibrated with 28 HP sample images can accurately estimate the percentage canopy cover for a continuous area of T. grandis plantations and facilitate mapping of canopy heterogeneity to monitor threats of diseases, mortality, fires, pests and other disturbances.

     

Cloning and structure analysis of an NBS-LRR disease-resistant gene from Setaria italica Beauv

Degenerate PCR primers targeting conserved motifs of most NBS-LRR disease-resistant genes in plants were tested in Setaria italica Beauv. cultivar Shilixiang, which is resistant to Uromyces setariae-italicae. A sequence with a length of 2673 bp has been obtained by using Genomic Walking technology. The nucleotide sequence contained an open reading frame that encoded 891 amino acid residues with a calculated molecular mass of 101.44 kDa. It was named RUS1 (Resistance against Uromyces setariae-italicae, GenBank No. FJ467296). It contained an NB-ARC domain and three conserved motifs P-loop, kinase 2, and kinase 3, which had the characteristics of NBS-LRR type resistant gene of plant. Phylogenetic analysis indicated that it was similar to RPM1 and might belong to LZ-NBS-LRR type disease resistance gene. Southern blotting result displayed that there were at least three copies of RUS1 in the foxtail millet genome.     

复杂背景与天气条件下的棉花叶片图像分割方法

为实现自然条件下棉花叶片的精准分割,提出一种粒子群(Particle swarm optimization,PSO)优化算法和K-means聚类算法混合的棉花叶片图像分割方法。本算法将棉花叶片图像在RGB颜色空间模式下采用二维卷积滤波进行去噪预处理,并将预处理后的彩色图像从RGB转换到目标与背景差异性最大的Q分量、超G分量、a*分量;随后在K均值聚类的一维数据空间中,利用PSO算法向全局像素解的子空间搜寻,通过迭代搜寻得到全局最优解,确定最佳聚类中心点,改善K均值聚类的收敛效果;最后,对像素进行聚类划分,从而得到棉花叶片分割结果。按照不同天气条件和不同背景采集了1 200幅棉花叶片样本图像,对本研究算法进行测试。试验结果表明:该算法对于晴天、阴天和雨天图像中目标(棉花叶片)分割准确率分别达到92.39%、93.55%、88.09%,总体平均分割精度为91.34%,并与传统K均值算法比较,总体平均分割精度提高了5.41%。分割结果表明,本研究算法能够对3种天气条件(晴天、阴天、雨天)与4种复杂背景(白地膜、黑地膜、秸秆、土壤)特征混合的棉花叶片图像实现准确分割,为棉花叶片的特征提取与病虫害识别等后续处理提供支持。     

Bacterial populations on the phyllosphere of Mediterranean plants: influence of leaf age and leaf surface

In the present study, we estimated the size of phyllosphere bacterial populations in young and mature leaves from the same plants and also assessed the population abundance on adaxial and abaxial leaf surfaces. We examined eight perennial species naturally occurring in the same area, in Halkidiki (northern Greece). They are Arbutus unedo, Quercus coccifera, Pistacia lentiscus, and Myrtus communis (evergreen sclerophyllous species), Lavandula stoechas and Cistus incanus (drought semideciduous species), and Calamintha nepeta and Melissa officinalis (nonwoody perennial species). Young and mature leaves were examined from the four sclerophyllous evergreen species for their epiphytic bacterial colonization, and it was found that mature leaves were highly populated compared to the younger ones except in M. communis. As regards the bacterial colonization of the two leaf surfaces, no differences were found in most species except for the drought semideciduous type where the two leaf surfaces behaved differently.     

融合无人机多光谱和纹理特征的马铃薯LAI估算

目的 研究融合无人机遥感影像多光谱信息和纹理特征估算马铃薯Solanum tuberosum叶面积指数(Leaf area index,LAI)方法,提高马铃薯LAI反演精度。方法 利用大疆P4M无人机采集2021年2－4月南方冬种马铃薯幼苗期、现蕾期、块茎膨大期多光谱影像,用LAI-2000冠层分析仪实测LAI数据。提取影像光谱、纹理等信息,分析植被指数、纹理特征与LAI的相关性,基于R²_adj的全子集分析优选特征变量。采用主成分分析,融合光谱和纹理特征,用PCA-MLR(Principal component analysis-multiple linear regression)模型估算马铃薯LAI。结果 从幼苗期到块茎膨大期,PCA-MLR估算模型优于T-MLR(Texture multiple linear regression)和VI-MLR(Vegetation index multiple linear regression)模型,R²分别为0.73、0.59和0.66。结论 本研究提出一种估算马铃薯LAI的PCA-MLR方法,为马铃薯的长势监测和田间管理提供数据支持。     

Detection of target spot and bacterial spot diseases in tomato using UAV-based and benchtop-based hyperspectral imaging techniques

Early and accurate diagnosis is a critical first step in mitigating losses caused by plant diseases. An incorrect diagnosis can lead to improper management decisions, such as selection of the wrong chemical application that could potentially result in further reduced crop health and yield. In tomato, initial disease symptoms may be similar even if caused by different pathogens, for example early lesions of target spot (TS) caused by the fungus Corynespora cassicola and bacterial spot (BS) caused by Xanthomonas perforans. In this study, hyperspectral imaging (380–1020 nm) was utilized in laboratory and field (collected by an unmanned aerial vehicle; UAV) settings to detect both diseases. Tomato leaves were classified into four categories: healthy, asymptomatic, early and late disease development stages. Thirty-five spectral vegetation indices (VIs) were calculated to select an optimum set of indices for disease detection and identification. Two classification methods were utilized: (i) multilayer perceptron neural network (MLP), and (ii) stepwise discriminant analysis (STDA). Best wavebands selection was considered in blue (408–420 nm), red (630–650 nm) and red edge (730–750 nm). The most significant VIs that could distinguish between healthy leaves and diseased leaves were the photochemical reflectance index (PRI) for both diseases, the normalized difference vegetation index (NDVI850) for BS in all stages, and the triangular vegetation index (TVI), NDVI850 and chlorophyll index green (Chl green) for TS asymptomatic, TS early and TS late disease stage respectively. The MLP classification method had an accuracy of 99%, for both BS and TS, under field (UAV-based) and laboratory conditions.