Airborne small-footprint discrete-return LiDAR data in the assessment of boreal mire surface patterns, vegetation, and habitats

Boreal mires encompass high diversity in species and habitats, many of which are endangered. In Finland, extensive use of peatlands has resulted in habitat fragmentation. A need for accurate and cost-efficient vegetation mapping and monitoring of habitat changes exists in mire conservation, restoration and peatland forestry. LiDAR is an emerging and excellent tool for probing the geometry of vegetation and terrain. Modern systems measure the backscattered signal accurately and also provide radiometric information. Experiments were carried out in a complex minerotrophic–ombrotrophic eccentric raised bog in southern Finland (61°47′N, 24.18′E). First, we tested discrete-return LiDAR for the modeling of mire surface patterns and the detection of hummocks and hollows, as well as the effect of mire plants on the Z accuracy of the surface echoes. Secondly, the response of different mire vegetation samples in LiDAR intensity was examined. Thirdly, we tested area-based geometric and radiometric features in supervised classification of mire habitats to discover the meaningful variables. The vertical accuracy of LiDAR in mire surface modeling was high: 0.05–0.10 m. A binary hummock-hollow model that was estimated from a LiDAR-based elevation model matched flawlessly in aerial images and had moderate explanatory power in habitat classification trials. The intensity of LiDAR in open-mire vegetation was mainly influenced by the surface moisture, and separation of vegetation classes spanning from ombrotrophic to mesotrophic vegetation proved to be difficult. Area-based features that characterize the height distribution of LiDAR points in the canopy were the strongest explanatory variables in the classification of 21 diverse mire site types. Actual qualifying differences in the ground flora were unattainable in the LiDAR data, which resulted in inferior accuracy in the characterization of ecohydrological conditions and nutrient level of open mires and sparsely forested wet sites. Mire habitat classification accuracy with LiDAR surpassed earlier results with optical data. The results suggested that LiDAR constitutes an efficient aid for monitoring applications. We propose the co-use of images and LiDAR for enhanced mapping of open mires and tree species. In situ calibration and validation procedures are required until invariant geometric and radiometric features are discovered.     

SVM算法支持下的耕地面积退化遥感监测——以昆明市呈贡区为例

基于Landsat卫星TM、ETM+和OLI数据,采用SVM算法结合RS和GIS技术,分别提取了呈贡区5个时期(1989、1996、2003、2009、2015年)的耕地面积。结果表明,1989—2015年呈贡区耕地面积由156.99 km2减少到40.02 km2,减少了74.51%,年平均减少4.50 km2;空间分布上,耕地集中在呈贡西部、西南部、东北部和中部地区,其中以中部地区减少最为剧烈和明显。原因主要是政府的宏观调控、城市化速率和城市建设占用了大量耕地。建议政府协调人、地、自然和城市发展建设之间的关系,严守耕地红线,出台耕地保护措施。     

基于近红外光谱的番茄黄化曲叶病抗病性识别研究

使用传统的聚合酶链反应技术检测番茄(Lycopersicon esculentum Mill.)植株是否有黄化曲叶病的抗病基因,进而确定植株是否具有抗病性;采集鉴定后的植株叶片的近红外漫反射光谱,采用多种方法对原始光谱进行预处理,并将不同预处理后的数据以及原始光谱作为输入,利用支持向量机建立抗病性的识别模型。采用标准正态变量变换和去趋势算法预处理后所建立的模型对预测集的识别准确率可以达到96.153 8%。表明通过近红外光谱技术可以识别番茄植株对黄化曲叶病是否具有抗病性。     

支持向量机在地下水水质评价中的应用

【目的】针对地下水水质评价中影响因素的模糊性和各因素与评价等级之间的不确定性等问题,采用支持向量机模型进行地下水水质评价研究,改进地下水水质评价方式。【方法】应用非线性支持向量机模型中的分类支持向量机,选用Gauss核函数,以羊毛湾灌区部分水井的水质资料为研究对象,进行地下水水质评价,并利用综合指数法和BP人工神经网络法对评价结果进行验证。【结果】羊毛湾灌区水质评价结果显示,该区地下水资源已被污染,需要进行保护性开发。3种方法的评价结果较为相似,但相较于综合指数法,支持向量机计算速度较快,易于通过计算机实现;相较于标准BP人工神经网络模型,支持向量机的评价精度较高,收敛速度较快,且所需参数较少。【结论】支持向量机能将复杂的非线性问题转化为线性问题,从而有效地避免过学习问题,并且拥有极大的泛化能力和对小样本问题的处理能力,可有效提高地下水水质评价精度,简化评价过程,为地下水水质评价提供了一条新思路。     

锡林郭勒草原土壤含水量遥感反演模型及干旱监测

土壤含水量是农牧业干旱监测系统的重要因子,对草原植被的生长及其与环境的相互作用具有重要意义。利用MODIS MOD16A2 蒸散发(evapotranspiration,ET)月产品数据及测墒站实测土壤含水量,通过相关分析和回归分析等方法,建立了基于蒸散发亏缺指数(evapotranspiration deficit index,ETDI)的土壤体积含水量(soil volumetric moisture,SVM)反演模型SVM=-48.851×ETDI+54.669,该模型的均方根误差为3.27%,达到了区域土壤含水量快速反演的应用水平。基于国家标准确定该模型的干旱等级阈值,分析近15年(2000-2014年)锡林郭勒草原的干旱动态特征,结果表明,草甸草原的土壤含水量在14%上下波动,基本为无旱,只有2007和2009年发生轻旱;典型草原、沙地植被和总体的土壤含水量在11%上下波动,基本为轻旱,只有2007和2009年发生中旱;荒漠草原的土壤含水量在8%上下波动,基本为重旱等级,只有2002、2003和2012年为中旱。近15年锡林郭勒草原平均约66%的面积存在不同程度的干旱,无旱和重旱面积呈增加趋势,轻旱和中旱面积呈减少趋势,但变化均不显著(P>0.05)。     

基于支持向量机的玉米根茬行图像分割

玉米根茬行的准确识别是实现玉麦轮作机器视觉式小麦自动对行免耕播种技术的前提。针对华北一年两熟区联合收获机玉米留茬收获后根茬行较难准确分割的问题，该研究以直立玉米根茬为研究对象，提出一种基于支持向量机（Support Vector Machine，SVM）的玉米根茬行分割方法。首先，利用主成分分析（Principal Components Analysis，PCA）对提取的目标（直立根茬）与背景（行间秸秆及裸露地表）的颜色和纹理特征进行分析，优选出21个特征，构成特征向量作为训练直立根茬SVM识别模型的输入；然后，根据图像坐标设置图像中间包含完整玉米根茬行的矩形区域为感兴趣区域（Region of Interest，ROI）；最后，使用训练好的直立根茬SVM识别模型以25×25（像素）的窗口在ROI内滑动检测，采用阈值法分割根茬行并通过形态学处理优化得到最终的玉米根茬行二值图像。利用在农业农村部河北北部耕地保育农业科学观测实验站采集的100幅玉米根茬行图像进行试验，结果表明，本文方法对于不同行间秸秆覆盖量和不同光照条件下的根茬行分割表现出较好的准确性和鲁棒性，直立根茬平均识别准确率、平均分割准确率、平均召回率、平均分割准确率和平均召回率的加权调和平均值（F1avr值）分别为93.8%、93.72%、92.35%和93.03%，每幅图像的平均分割时间为0.06 s，具有较好的实时性。基于SVM的分割方法可实现联合收获机玉米留茬收获后根茬行图像的分割，为下一步检测玉米根茬行直线并将其作为导航基准线进行视觉导航的研究提供良好基础。     

小波域马铃薯典型虫害图像特征选择与识别

为准确、快速地识别马铃薯典型虫害,提出了一种基于小波域的马铃薯典型虫害特征提取与识别方法。该方法以自然环境下的马铃薯虫害分割图像为对象,提取小波域高斯空间模型的高频协方差阵特征值与低频低阶矩(HELM)的12个不变纹理特征、空间域Hu不变矩的4个形状特征,进行支持向量机(SVM)的虫害分类识别。通过对8类典型虫害的识别,试验结果表明:在SVM识别方法下,本文HELM特征提取方法,相比传统纹理特征提取方法,在特征计算量不增加的同时,平均识别率至少提高了17个百分点;在HELM特征与Hu矩特征下,本文SVM的运行时间为0.481 s,比人工神经网络快了近2 s,平均识别率为97.5%,比人工神经网络、贝叶斯分类器识别率提高了至少6个百分点,有明显的识别优势。     

Comparison of machine learning algorithms for mapping mango plantations based on Gaofen-1 imagery

Mango is a commercial crop on Hainan Island, China, that is cultivated to develop the tropical rural economy. The development of accurate and up-to-date maps of the spatial distribution of mango plantations is necessary for agricultural monitoring and decision management by the local government. Pixel-based and object-oriented image analysis methods for mapping mango plantations were compared using two machine learning algorithms (support vector machine (SVM) and Random Forest (RF)) based on Chinese high-resolution Gaofen-1 (GF-1) imagery in parts of Hainan Island. To assess the importance of different features on classification accuracy, a combined layer of four original bands, 32 gray-level co-occurrence (GLCM) texture indices, and 10 vegetation indices were used as input features. Then five different sets of variables (5, 10, 20, and 30 input variables and all 46 variables) were classified with the two machine learning algorithms at object-based level. Results of the feature optimization suggested that homogeneity and variance were very important variables for distinguishing mango plantations patches. The object-based classifiers could significantly improve overall accuracy between 2–7% when compared to pixel-based classifiers. When there were 5 and 10 input variables, SVM showed higher classification accuracy than RF, and when the input variables exceeded 20, RF showed better performances. After the accuracy achieved saturation points, there were only slightly classification accuracy improvements along with the numbers of feature increases for both of SVM and RF classifiers. The results indicated that GF-1 imagery can be successfully applied to mango plantation mapping in tropical regions, which would provide a useful framework for accurate tropical agriculture land management.     

深锥浓密机底流浓度预测与外部结构参数优化

针对困扰支持向量机(SVM)模型参数选择问题,结合遗传算法(GA),建立了深锥浓密机底流放砂浓度的GA-SVM预测模型,研究了不同结构参数状态下底流浓度的变化规律,进行了深锥浓密机的外部结构参数优化选择。以司家营铁矿为例,在最优底流放砂浓度为72%的条件下,经外部结构参数优化后的深锥浓密机锥高10 m、锥角为30°,系统稳定可靠、底流连续均匀,动力荷载较同类矿山降低约15%,压耙停机故障降低80%。     

基于支持向量机的机械系统多故障分类方法

提出了一种利用支持向量机(SVM)对机械系统故障进行分类的新方法；以二值分类为基础，开发了基于支持向量机的多值分类器。并以齿轮的多种故障分类为例，进行了实际应用验证。结果表明，该方法具有很好的分类能力和较高的计算效率，不需要对原始数据进行预处理就可达到满意的效果，可以满足在线诊断的要求，适合于机械故障诊断中的多故障分类。该方法的应用，为故障诊断技术向智能化方向发展提供了新的途径。