Estimating maize seedling number with UAV RGB images and advanced image processing methods

Precision Agriculture - Accurately identifying the quantity of maize seedlings is useful in improving maize varieties with high seedling emergence rates in a breeding program. The traditional...     

Predicting cover crop biomass by lightweight UAS-based RGB and NIR photography: an applied photogrammetric approach

Easy-to-capture and robust plant status indicators are important factors when implementing precision agriculture techniques on fields. In this study, aerial red, green and blue color space (RGB) photography and near-infrared (NIR) photography was performed on an experimental field site with nine different cover crops. A lightweight unmanned aerial system (UAS) served as platform, consumer cameras as sensors. Photos were photogrammetrically processed to orthophotos and digital surface models (DSMs). In a first validation step, the spatial precision of RGB orthophotos (x and y, ± 0.1 m) and DSMs (z, ± 0.1 m) was determined. Then, canopy cover (CC), plant height (PH), normalized differenced vegetation index (NDVI), red edge inflection point (REIP), and green red vegetation index (GRVI) were extracted. In a second validation step, the PHs derived from the DSMs were compared with ground truth ruler measurements. A strong linear relationship was observed (R ² = 0.80?0.84). Finally, destructive biomass samples were taken and compared with the remotely-sensed characteristics. Biomass correlated best with plant height (PH), and good approximations with linear regressions were found (R ² = 0.74 for four selected species, R ² = 0.58 for all nine species). CC and the vegetation indices (VIs) showed less significant and less strong overall correlations, but performed well for certain species. It is therefore evident that the use of DSM-based PHs provides a feasible approach to a species-independent non-destructive biomass determination, where the performance of VIs is more species-dependent.     

应用混合线性模型评估蛋白高度遗传参数

为分析蛋白高度动态变化规律,估计该性状不同周龄的遗传力和遗传相关系数,以高产蛋鸡白来航鸡与地方鸡鸡种绿壳蛋鸡构建F2资源群体,采集资源群体蛋白高度数据18 624条,应用约束最大似然算法分析表型方差组分。结果表明,混合线性模型的固定效应宜包含批次和亲代品种效应;资源群体蛋白高度遗传力为0.17~0.30;不同周龄蛋白高度表型相关系数为0.22~0.43,遗传相关系数为0.58~0.97,相邻周龄遗传相关系数较高。结果显示资源群体蛋白高度遗传力属于中等偏下,通过中长期选择才能有效获取遗传进展;蛋鸡分离群体蛋白高度不同周龄之间表型值差异不大、遗传相关系数较高,选择1~2个时间点足以代表全期蛋白高度水平。     

无人机高光谱成像技术在作物生长信息监测中的应用

随着高效精准无损检测技术发展与应用,精准农业发展趋势明显,高光谱遥感作为该领域前沿技术,应用广泛。高光谱成像（Hyperspectral imaging,HSI）技术具有波段多、光谱窄、数据量大优点,利用光谱特征研究作物生长信息、各种光谱模型及精细分类和识别。文章重点论述近5年国内外HSI技术在无人机（Unmanned aerial system,UAS）遥感系统中对作物生长信息快速无损检测应用的研究进展,通过数据集扩充与注释、最佳波段组合和利用,配合反演方案开发与训练,以及精确图像处理技术评估作物氮素营养、生物量和产量、病害或胁迫、表型及植物功能特征等生长信息,该技术精准快速反映农业生产中许多具体问题,为现代化精细化农业发展提供有力保障。     

Weed detection by UAV: simulation of the impact of spectral mixing in multispectral images

This study aimed to assess the spectral information potential of images captured with an unmanned aerial vehicle, in the context of crop–weed discrimination. A model is proposed in which the entire image acquisition chain is simulated in order to compute the digital values of image pixels according to several parameters (light, plant characteristics, optical filters, sensors…) to reproduce in-field acquisition conditions. The spectral mixings in the pixels are modeled, based on an image with a 60 mm spatial resolution, to estimate the impact of the resolution on the ability to discriminate small plants. The classification potential (i.e. the ability to separate two classes) in soil and vegetation and in monocotyledon and dicotyledon classes is studied using simulations for different vegetation rates (defined as the proportion of vegetation covering the surface projected in the considered pixel). The classification is unsupervised and based on the Mahalanobis distance computation. The results of soil-vegetation discrimination show that pixels with low vegetation rates can be classified as vegetation: pixels with vegetation rate greater than 0.5 had a probability to be correctly classified between 80 and 100%. Classification between monocotyledonous and dicotyledonous plants requires pixels with a high vegetation rate: to obtain a probability to be correctly classified better than 80%, vegetation rates in the pixels have to be over 0.9. To compare the results with data from real images, the same classification was tested on multispectral images of a weed infested field. The comparison confirmed the ability of the model to assess vegetation–soil and crop–weed discrimination potential for specific sensors (such as the multiSPEC 4C sensor, AIRINOV, Paris, France), where the acquisition chain parameters can be tested.     

Optimization of multi-source UAV RS agro-monitoring schemes designed for field-scale crop phenotyping

Precision Agriculture - Unmanned aerial vehicle (UAV) system is an emerging remote sensing tool for profiling crop phenotypic characteristics, as it distinctly captures crop real-time information...     

Quantifying effect of tassels on near-ground maize canopy RGB images using deep learning segmentation algorithm

Precision Agriculture - Vegetation indexes (VIs) are a key variable for monitoring the crop growth and estimating crop productivity. The spectral traits of the tassels were significantly different...     

Curved and straight crop row detection by accumulation of green pixels from images in maize fields

This study proposes a new method for detecting curved and straight crop rows in images captured in maize fields during the initial growth stages of crop and weed plants. The images were obtained under perspective projection with a camera installed onboard and conveniently arranged at the front of a tractor. The final goal was the identification of the crop rows which are crucial for precise autonomous guidance and site-specific treatments, including weed removal based on the identification of plants outside the crop rows. Image quality is affected by uncontrolled lighting conditions in outdoor agricultural environments and by gaps in the crop rows (due to lack of germination or defects during planting). Also, different plants heights and volumes occur due to different growth stages affecting the crop row detection process. The proposed method was designed with the required robustness to cope with the above undesirable situations and it consists of three sequentially linked phases: (i) image segmentation, (ii) identification of starting points and (iii) crop row detection. The main contribution is the ability of the method to detect curved crop rows as well as straights rows even with irregular inter-row spaces. The method performance has been tested in terms of accuracy and time processing.     

结合无人机可见光和激光雷达数据的杉木树冠信息提取

以年珠实验林场为研究区,以无人机可见光正射影像和激光雷达数据为数据源,采用分水岭分割与面向对象结合的方法提取不同郁闭度下杉木单木树冠信息,并对提取精度进行验证首先采用面向对象法基于无人机可见光影像提取树冠区域,然后基于构建的CHM进行分水岭分割获取单木树冠初步分割结果,最后基于初步分割结果对树冠区域进行二次分割,提取单木树冠信息。结果表明：不同郁闭度林分条件下单木树冠信息提取效果较好,其中单木树冠提取F测度分别为88.07%～95.08%和78.57%～88.29%;提取的树冠面积与实测面积建立的线性回归模型,R²分别为0.8591和0.7367,RMSE分别为2.49 m²和3.29 m²;提取的冠幅与实测冠幅建立的线性回归模型,R²分别为0.8306和0.7246,RMSE分别为0.46 m和0.57 m。基于无人机可见光影像采用面向对象多尺度分割法提取树冠区域很好的消除了样地内裸地及林下灌木等因素的影响;同时,无人机LiDAR数据能够更加精确的区分单木信息,2种数据源结合发挥了二者的优势,提高了单木树冠的提取精度。本研究可为快速获取不同郁闭度林分下单木树冠信息提供参考。     

应用多光谱数字图像识别苗期作物与杂草

通过对多光谱成像仪获得的数字图片,采用一定的目标分割与形态学处理,对豆苗和杂草进行识别判断.为解决识别速度与正确率的矛盾,以豆苗和杂草图像的识别为例,提出一种基于多光谱图像算法的杂草识别新方法.应用3CCD多光谱成像仪获取豆苗与杂草图像,以多光谱图像的近红外IR通道图像为基础,利用图像分割和形态学方法,将所有豆苗叶子影像提取出来.对于剩下的2种杂草(牛筋草,空心莲子草)图像,先利用图像分析工具统计出图像块的长度、宽度、面积等基本特征参数,并根据它们形状的不同,总结出两条简单的判别规则,进行进一步的识别.本试验对147个目标进行判断,其中误判14个,正确率为90.5%,表明该方法算法简单、计算量小、速度快,能够有效识别这2种杂草,为田间杂草的快速识别提供了一种新方法.     

豆类杂粮作物种子物性参数的试验研究

为了获得豆类杂粮作物相关播种装备的设计依据，选取大豆、绿豆、豌豆、红豆和蚕豆5种典型豆类杂粮作物为研究对象，采用称重法、量筒法、低温烘干法等进行三轴尺寸、千粒重、含水率、密度、休止角、堆积角、摩擦系数、碰撞恢复系数、刚度系数、破碎力、弹性模量和剪切模量等相应物理特性的试验研究。结果表明：所测杂豆种子三轴尺寸的长、宽、高和球度范围分别为4.78～21.11 mm、3.58～15.63 mm、3.61～8.33 mm和0.66～ 0.97；千粒重、含水率和密度范围分别为51.9～1 548.4 g、11.5%～14.4%和1.067～1.323 g·cm^-3；所测杂豆种子休止角和堆积角的范围分别为21.840～35.199°和22.579～33.530°；杂豆种子在PVC板上的静摩擦系数最小，除蚕豆外，其他种子与种子间的静摩擦系数和动摩擦系数均最大。5种杂豆种子中，蚕豆的抗挤压能力最大，绿豆的抗挤压能力最小；豌豆的刚度系数、弹性模量及剪切模量均最大，大豆均最小。杂豆种子物性参数的测量可为杂豆排种器的研发提供理论依据。     

基于热红外和RGB图像的番茄苗期高温胁迫检测方法

【目的】针对实际生产场景中番茄苗期生长遇到的高温胁迫问题,提出一种基于热红外和RGB图像的番茄苗期高温胁迫检测方法。【方法】首先,通过番茄苗期热红外图像反演获取番茄冠层温度参数,采用偏最小二乘(Partial least squares, PLS)模型提取冠层温度特征指标;然后,建立采用3种不同主干特征提取网络的MaskRCNN模型,通过迁移学习的方式将番茄苗期RGB图像输入Mask-RCNN模型,进行高温胁迫症状实例分割,得到番茄苗期胁迫症状特征指标;最后,利用提取的温度和胁迫症状特征指标构建分级数据集,输入高温胁迫分级模型,得到高温胁迫等级。【结果】基于PLS模型提取的冠层温度特征指标累计贡献率达95.45%;基于ResNet101+Mask-RCNN的高温胁迫症状分割网络对番茄苗期轻度和重度胁迫的分割精度最高,均值平均查准率(Mean average precision, mAP)分别为77.3%和73.8%;基于温度和胁迫症状特征指标构建的4种高温胁迫分级模型中,反向传播神经网络(Back propagation neural network, BPNN)获得最好的高温胁迫分级...     

新疆冰糖心红富士苹果RGB图像多指标分析

[目的]采用机器视觉技术对新疆冰糖心红富士苹果进行重量、糖度预测和分级.[方法]分析提取苹果RGB图像中单色、波长差、HSV转换后分量等多类型图像,对比图像分割效果确定后续处理图像.采用形态学处理剔除二值化图像果梗区域,提取目标区域几何、灰度和色调频度等特征.采用多元线性和偏最小二乘回归预测苹果重量和糖度,判别分析分类苹果,结合全组合实验方法和特征优选,获得较佳特征集合.[结果]多元线性回归方法建立苹果糖度的预测模型结果最佳,使用几何和灰度的特征集合,建模集和验证集糖度预测相关系数分别为0.623和0.570;使用面积、周长、长轴长度和短轴长度特征集和,或体积、周长、长轴长度和短轴长度四个特征时,多元线性回归预测苹果重量,验证集预测相关系数r为0.992,预测均方根误差为3.88 g,相对分析误差为8.1;采用基于特征优选方法确定41个主要特征,二次判别函数的判别分析分级苹果,验证集分级准确率达到98.7;.[结论]RGB图像能够准确预测新疆冰糖心红富士苹果重量,并能精确分级,但糖度预测效果不佳.     

基于无人机影像的高寒草地鼠害信息提取研究

【目的】草原鼠害是影响草原生态平衡的重要因素,基于低空遥感影像探索提取鼠害 信息的最佳方案和分辨率对解决草原鼠害意义重大。【方法】文章基于高分辨率无人机正射 影像,使用CART 决策树、支持向量机、最邻近、贝叶斯4 种监督分类方法对高原鼠兔和高原 鼢鼠两种鼠害进行分类并比较其精度,再使用不同飞行高度下获取的遥感影像提取鼠害信息。 【结果】在鼠兔鼠害信息提取中,基于决策树分类法的总体精度为89.00%,kappa 系数为0.79; 支持向量机分类方法的总体分类精度为92.00%,Kappa 系数为0.83;最邻近分类法的总体分类 精度为94.00%,Kappa 系数为0.87;基于贝叶斯分类法的混淆矩阵中得到的鼠洞的分类精度最 差,鼠洞的生产者精度与用户精度都在78.00% 以下。在鼢鼠鼠害信息提取中,基于决策树分 类结果的总精度为93%,Kappa 系数为0.86;支持向量机分类结果的总精度达到95%,Kappa 系数为0.90;最邻近法的分类结果的总精度达到97.00%,Kappa 系数为0.95;Bayes 分类法的总 体分类精度为98.00%,Kappa 系数达到了0.95。【结论】基于面向对象的最邻近分类法是高原鼠 兔鼠害信息提取的精度最优方法,基于面向对象的贝叶斯分类法是高原鼢鼠鼠害信息提取的最 佳方法。对于飞行相对高度分别为100 m、120 m 和200 m 的无人机遥感影像数据,随着飞行高 度的增大,影像的空间分辨率越低,其分类所需要的时间、分类精度和斑块数量均呈下降趋势。     

基于无人机影像的银杏单木胸径预估方法

胸径是立木测定的基本因子,自动获取胸径数据是准确高效计算森林蓄积量和生物量的关键。以银杏Ginkgo biloba为研究对象,通过无人机获得影像数据,利用运动恢复结构（SFM）方法生成数字表面模型和正射影像图,进而提取单株银杏的树冠面积（A_c）,冠幅（W_c）及树高（H）。3个参数分别与胸径（D_BH）建立一元回归模型（A_c-D_BH,W_c-D_BH,H-D_BH）,二元回归模型（A_c&W_c-D_BH,A_c&H-D_BH,W_c&H-D_BH）和三元回归模型（A_c&W_c&H-D_BH）。52组拟合样本的结果显示:A_c&W_c&H-D_BH模型的决定系数（R2）最高为0.825 0,均方根误差（E_RMS）最小为0.959 1。19组检测样本的结果显示:A_c&W_c&H-D_BH模型反演的胸径值误差率为4.20%,小于A类森林资源胸径因子允许的误差值（5%）。研究结果表明:通过无人机采集树冠面积、冠幅和树高3个参数,可计算得到较高精度的胸径值。     

Spectral indices from aerial images and their relationship with properties of a corn crop

Identification of areas with similar restrictions to crop productivity could improve the efficiency to manage agricultural systems, guarantee stable yields, and reduce the effect of droughts in rainfed systems. The ability of any vegetation index to discriminate N and moisture-related changes in leaf reflectance would present an important advantage over the present diagnostic system which involves soil-testing for moisture and available N. The purpose of the study was to calibrate different vegetation indices regarding their capacity to identify water and nitrogen availability for rainfed corn crops in the semiarid Pampas of Argentina. A field experiment with corn with a control without fertilization (N0), and fertilized with 120 kg ha^?1 of nitrogen (N120) was used. Two sites, Low (L) and High (H), were identified within the field, according to their altimetry, a multi-spectral aerial photography was taken from a manned airplane during flowering stage of the corn crop, and four spectral indices were calculated (NDVI, green NDVI, NGRDI, (NIR/GREEN)-1). At six georeferenced points at each site soil texture, organic matter, available phosphorus, nitrogen and moisture contents as well as corn aerial biomass and grain yield were determined. The two sites differed in most of the evaluated soil properties, crop biomass and grain yield. The spectral information obtained at crop flowering showed clear differences between sites H and L for all four indices, indicating that any of these would be able to detect the differences in soil moisture and fertility among these environments. Both (NIR/GREEN)-1 and green NDVI had the best correlation with crop yield determined in the field, and therefore could be considered most appropriate for estimating corn yields from images taken at flowering. For estimation of N requirements, green NDVI differentiated best between fertilized and non-fertilized crop in the moisture limited environment (H), while (NIR/GREEN)-1 performed better in the site where soil moisture was non-limiting (L).     

基于四旋翼无人机快速获取大田植株图像的方法及其应用

利用低空无人机获取农田信息,具有实时以及灵活性高、成本低等优势。为快速、精确监测大田规模化种植作物的生长发育状况,以四旋翼无人机为平台,结合数字图像技术,建立快速获取大田烟株中前期图像的方法。结果表明,在天空辐射条件较稳定的条件下,采用较低的飞行高度(如20m)航拍获取田块图像,能够得到清晰的拼接图像和三维重建效果;采用基于决策树的植被分割算法将烟草和非植被部分分割后,得到较高精度的大田植株图像。在此基础上进行大田烟草缺苗数估测,所估算的缺苗数与实测值吻合较好。     

咖啡种质资源遗传多样性的ISSR分析

[目的]对我国收集保存的咖啡种质资源进行遗传多样性分析,为咖啡种质资源的收集、保存、鉴定、创新及有效利用提供理论参考.[方法]从哥伦比亚大学(UBC Primer Set#9)公布的100个ISSR引物序列中筛选出多态性好、扩增条带清晰且重复性好的引物,利用其对72份咖啡种质材料进行扩增,对电泳图谱进行统计分析,并利用NT-SYSpc 2.1计算遗传相似系数和遗传距离.根据非加权算术平均法(UPGMA)进行聚类分析,并进行主坐标分析.[结果]筛选出的19条引物共扩增出153条条带,其中多态性条带数为128条,占总条带数的83.7%;72份咖啡种质材料间的遗传相似系数为0.4531~0.9609,平均为0.6567;遗传距离为0.0351~1.0951,平均为0.4081.其中,大粒种种内遗传相似系数最大(0.9297~0.9531),遗传距离最小(0.0445~0.0669);中粒种种内遗传相似系数最小(0.5781~0.8750),遗传距离最大(0.1210~0.5363).主坐标分析与聚类分析结果一致,均显示72份咖啡种质材料可分为三大类,其中3份大粒种、3份查理种及1份中粒种巴布亚新几内亚-2聚为Ⅰ类群;32份小粒种聚为聚为Ⅱ类群;除巴布亚新几内亚-2外的其他31份中粒种及1份中小粒杂交种Arabusta和1份福建咖啡共33份聚为Ⅲ类群.聚类分析结果与种质地理来源无明显相关性.[结论]咖啡种质资源各种间存在较大的遗传差异,以中粒种种内遗传多样性较丰富,以大粒种种内遗传多样性较低.利用ISSR分子标记可准确分析咖啡资源遗传多样性.     

Modelling agronomic images for weed detection and comparison of crop/weed discrimination algorithm performance

A new method for weed detection based on modelling agronomic images taken from a virtual camera placed in a virtual field is proposed. The aim was to measure and compare the effectiveness of the developed algorithms. Two sets of images with and without perspective effects were simulated. For images with no perspective, based on Gabor filtering and on the Hough transform, the performance of two crop/inter-row weed discrimination algorithms were tested and compared. The method based on the Hough transform is, in any case, better than the one based on Gabor filtering. For images with perspective effects only, an algorithm based on the Hough transform was tested and an extension to real images is discussed. These tests were done by a comparison between the weed infestation rate detected by these algorithms and the true one. This evaluation was completed with a crop/weed pixel classification and it demonstrated that the algorithm based on a Hough transform gave the best results (up to 90%).     

基于无人机遥感影像的沙糖橘果树提取方法研究

目的 通过无人机获取沙糖橘果园的遥感图像,快速提取果树分布位置,为果树的长势监测和产量预估提供参考。方法 以无人机拍摄的可见光遥感图像为研究对象,计算超红指数、超绿指数、超蓝指数、可见光波段差异植被指数、红绿比指数和蓝绿比指数6种可见光植被指数,使用双峰阈值法选取阈值进行果树的提取。在使用光谱指数进行识别的基础上,结合数字表面模型作为识别模型的输入变量,进行对比试验。结果 相比使用单一光谱指数,结合数字表面模型提高了果树和非果树像元的提取精度,6次波段融合后的总体精度均大于97%。超红指数与数字表面模型结合后的总体精度最高,为98.77%,Kappa系数为0.956 7,植被信息提取精度优于其他5种可见光植被指数与数字表面模型结合后的提取精度。结论 数字表面模型结合可见光植被指数的提取方法能够更深层次地挖掘遥感数据蕴含的信息量,为影像中色调相似地物的提取提供参考。