利用面向对象变化向量分析(OCVA)检测土地利用变化

为推广国产高分数据在土地利用变化检测方面的应用,以延庆区张山营镇2015和2018年2期高分二号(GF-2)影像为数据源,采用面向对象变化向量分析法(OCVA),先通过不同分割模式获取对象,其次利用引入权重的欧氏距离构建变化向量的模,再通过目标函数确定最佳检测阈值后对研究区进行变化检测,最后对变化区域进行面向对象分类,得到具体的"从…到"变化类型。结果表明:1)多时相组合分割模式下的OCVA法精度要高于多时相分别分割模式;2)利用加权组合方法构建对象变化向量的模更易确定最佳检测阈值,且最后的总体精度要高于传统OCVA法;3)结合面向对象分类的OCVA法能有效减少面向对象分类后变化检测法和单一OCVA法的伪变化,检测精度达到92.50%。     

天然气管道高后果区多尺度融合变化检测算法

针对天然气管道高后果区变化人工调查效率低的问题，提出基于多尺度融合的高后果区变化检测算法。首先对天然气管道沿线高后果区两期影像进行矢量裁剪，然后将两期高后果区影像输入变化检测算法中，计算出整幅影像的建筑变化特征信息，最后根据整幅影像计算结果生成变化图斑矢量。网络中设计的多尺度融合模块融合了高层特征信息和底层特征信息，此外增加的注意力机制模块从通道和空间范围提取具有区分性的特征信息。以武汉地区忠武线天然气管道线路为研究对象，使用2019年和2020年遥感影像开展实验，F1值和召回率分别达到0.849、0.838。结果表明，采用基于多尺度融合的高后果区变化检测算法计算结果能够满足高后果区变化检测的实际应用需求。（图5，表3，参20）     

基于加权Fisher模糊判别准则的真彩色影像分类方法研究

针对Fisher判别分类中,选取样本不平衡导致Fisher线性判别性能下降,提出了加权Fisher判别对类内散布矩阵进行修正的方法.同时,针对Fisher判别准则为阈值分类器,引入了一种新的Fisher模糊判别准则.该方法应用于无人机真彩色影像的道路、植被、裸土地分类,取得了较好的分类结果.并与传统Fisher方法进行了对比研究,实验结果表明,基于加权Fisher模糊判别方法比传统Fisher方法的分类性能明显提高.     

基于WorldView2和GF-2的面向对象多指标综合植被变化分析

目的利用高分辨率卫星影像获取精确的植被变化信息对植被资源合理利用及可持续经营有重要意义。传统的基于像元的直接变化检测法容易产生椒盐噪声,而用面向对象分类法结果又严重依赖于分类精度。本文在分析现有算法优劣势基础上,力图找到一种针对高分辨率遥感数据进行植被变化检测的相对客观算法,并验证其有效性。方法基于现有的多指标综合变化分析算法（MIICA）,提出了面向对象的MIICA。本算法用准确率（P）和查全率（R）分析确定的最优分割参数对前后两期跨传感器影像进行统一分割,利用分割获得的对象影像进行特征参数提取,并用ROC曲线法选择合适的阈值进行变化信息提取并整合,最终获得植被变化位置及方向（植被增多或减少）。结果经与基于像元的MIICA及面向对象分类法的比较,本方法的生产者精度高于基于像元的MIICA,用户精度高于面向对象分类法,并且总体精度和Kappa系数分别达到了0.880和0.805。本方法能更好地反映植被变化的位置及形状,也能较准确地检测出一些面积微小的变化。结论面向对象的MIICA能弥补基于像元的MIICA和面向对象分类的缺点,提高检测精度,对存在高人为影响的森林公园或自然保护区植被变化分析、植被资源合理利用及可持续经营有重要意义。      

基于遗传规划的地表覆盖变化检测方法研究

【目的 】针对地表覆盖变化受干扰因素影响而产生伪变化的问题,提出一种基于遗传规划（Genetic Programming,GP）的变化检测方法,以提高土地管理、灾害评估和生态监测中地表覆盖变化检测的准确性和实时性。【方法 】文章以山东省阳谷县为研究区,基于研究区两幅Landsat-8 OLI影像,采用GP算法进行种群初始化、适应度评估和种群进化,进而进化出最佳个体用于地表覆盖变化检测,并将GP检测结果与变化向量分析、主成分分析和相关系数法这3种传统遥感变化检测方法进行比较。【结果 】基于GP的变化检测方法获得的总体精度为95.55%,Kappa系数为0.91,变化和未变化像元的错分率分别为3.37%和5.50%,漏分率分别为5.58%和3.32%,优于3种传统变化检测方法。【结论 】基于GP的变化检测方法可以自动选择地表覆盖敏感光谱波段构建变化强度树,并自动集成阈值和变化强度树,避免了手动设置阈值带来的主观干扰。     

基于改进模糊聚类的烟草品质集成评价模型

针对烟草化学成分与烟草品质之间难以建立确定的数学模型的问题,提出了一种基于改进模糊聚类的烟草品质评价方法。该方法以烟叶样品的化学成分的差异性为依据,以模型分类结果与专家评吸结果的一致性为目标,利用模拟退火算法对现有的模糊聚类算法进行优化改进,建立基分类器;在此基础上,利用Ada Boost将基分类器对于不同样本集的多个分类结果进行集成,形成最终的烟草品质评价模型。以130组烟叶作为烟草样本,测定了各烟叶样品中总糖、还原糖、总氮、烟碱、氧化钾、氯离子、蛋白质7种化学成分含量,并采用改进的模糊聚类方法与神经网络算法、模糊聚类算法进行对比试验,该方法的误检率为6.7%,具有提升小样本数据的辨识能力,优于所比较的其他2种方法。     

面向对象的高分辨率遥感影像校园绿地快速提取

利用面向对象的影像处理软件——易康(e Cognition)作为软件支持和分类方法,以Quick Bird影像作为数据,以面向对象技术方法为主要技术,分别利用规则提取以及样本多边形对象建立训练区等2种方法,提取云南师范大学呈贡校区西区校园绿地。通过与传统的面向像元的分类方法的比较,体现了该方法信息获取的周期短、精度高、成本低等特点,实现了校园绿地的快速提取。     

应用植被变化与火烧迹地对森林火灾反演

火灾后的植被变化与火烧迹地面积是森林火灾反演的重要指标.以2020年3月28日四川省凉山州木里藏族自治县项脚乡火烧迹地为研究对象,依据Sentinel-2卫星影像数据,采用归一化差分植被指数、差分归一化燃烧比、迭代加权多元变化检测算法和追踪识别提取算法,分析该地区火灾前后的植被变化和受灾面积.结果 表明:实验检测的植被变化判别精度为91.95％;提取火烧迹地像元面积精度为90.77％,Kappa系数为0.87.     

基于连续型条件属性的模糊规则约简算法

针对粗糙集对于连续域属性决策表的处理能力差以及不容易获得模糊集之间关系等问题,提出一种把模糊集和粗糙集结合起来的连续型条件属性模糊规则约简算法。该算法首先引入三角隶属度函数将连续属性值转化为模糊值,并使用离散模糊神经网络方法获得数据集之间关系。实例验证表明,采用该算法,用户可以根据实际决策需要和领域知识更改阈值,从而获得满意的模糊规则结果。     

基于改进的相似度判别的气象传真图底图去除方法

在对气象传真图像进行矢量化识别之前,为了消除底图的影响,需要将底图从图像中去除,本文在研究以往底图去除算法的基础上,提出了一种改进的相似度判别的底图去除方法。该方法从细化后的接收图像中提取出各条分离的曲线,将曲线分别拿到标准的底图数据中相对应位置进行相似度判别运算,当遍历完接收图像中所有曲线后,即完成对整幅图像的判别。整个过程相当于只对整幅标准底图遍历一遍,所有曲线也仅参与运算一次即可得到判别。该方法在准确识别底图的基础上,使得运算的数据量大大减小,改善了原有方法运算量大且算法复杂的缺点。     

抑菌圈边缘检测图像预处理的研究

用Visual C++6.0作为开发工具进行了程序设计,实现了多种图像预处理算法。通过实验比较了不同滤波方法、锐化方法和二值化方法对抑菌圈边缘检测的影响。对来自扫描仪的抑菌圈图像,发现在进行图像预处理时,进行一次中值滤波和梯度锐化之后,再进行一次中值滤波,然后再采用阈值法二值化,可显著地提高抑菌圈边缘检测的成功率,使在扫描分辨力为100、200、300 dpi时的抑菌圈边缘检测率分别达到100%、98%、80%。     

数学形态学在高分辨率图像土地覆盖分类中的应用

高分辨率影像可以清楚地显示树木房屋等地物细节,但是用基于像元的分类很难将这些细碎的地物融合为通常的土地覆盖分类图。首先对高分辨率影像进行有监分类,再根据有监分类的结果对不同地类进行二值化,运用数学形态学对各地类二值图进行处理,从而得到地类的边界图。通过多组实验表明,数学形态学运用于高分辨率图像土地覆盖分类既保留了图像的地类特征,也快速得到了不同地类的边界。     

基于ICA阈值优化耦合信息熵的边缘提取算法

为了解决传统边缘提取算法对噪声敏感和阈值难以选取,边缘清晰度不高以及边缘不平滑等问题,提出了一种基于ICA阈值优化耦合信息熵的边缘提取算法.首先,基于灰度分布模式将图像分成若干子块,并计算每个子块的分段阈值;然后,为了从大量的分段阈值选择合适的阈值,引入了帝国主义竞争(imperialist competitive algorithm,ICA)优化算法,计算图像的最优阈值,根据获得的最优阈值将每个图像子块划分为不同的均匀区域;最后,通过计算每个均匀区域的信息熵,利用信息熵检测所有处于不同均匀区域的边界像素来提取边缘.实验结果表明:与当前常用的边缘提取算法比较,本文算法具有更高的品质因数与边缘连续性,能够抑制过于微小和琐碎的细节,突出有效的边缘信息,边缘定位精度高且平滑连贯,能够准确地提取目标轮廓.     

基于兰氏距离NDVI的土地覆盖变化检测

将兰氏距离变化检测指标引入NDVI变化检测,来实现土地利用的变化检测。在计算阈值时,对原有的局部阈值分割法进行改进,提出基于标准图像地物分类的局部阈值分割法。对检测结果进行精度评价,证实了该方法的可行性和有效性。     

Development of real-time onion disease monitoring system using image acquisition

In this study, real-time disease monitoring was conducted on onion which is the most representative crop in Republic of Korea, using an image acquisition system newly developed for the mobile measurement of phenotype. The purpose of this study was to improve the accuracy of prediction of disease and state variables by processing images acquired from monitoring. The image acquisition system was consisted of two parts, a motorized driving system and a PTZ (pan, tilt and zoom) camera to take images of the plants. The acquired images were processed as follows. Noise was removed through an image filter and RGB (red, green and blue) colors were converted to HSV (hue, saturation and value), which enabled thresholding of areas with different colors and properties for image binarization by comparing the color of onion leaf with ambient areas. Four objects with the most significant browning in the onion leaf to the naked eye were selected as the samples for data acquired. The thresholding method with image processing was found to be superior to the naked eye in identifying accurate disease areas. In addition, it was found that the incidence of disease was different in each disease area ratio. As a result, the use of image acquisition system in image processing analysis will enable more prompt detection of any changes in the onion and monitoring of disease outbreaks during the crop lifecycle.     

奶牛图像识别系统的研究

对奶牛图像识别系统作了研究,分析了奶牛图像的二值化处理,对光照过强、奶牛背部颜色等影响下的直方图作了分析,并提出了相应改进措施,给出了最佳阈值确定流程图。为进一步完善二值化结果,提出了以黑色点为种子点作区域增长的算法,还提出对二值化图像的识别方法。     

Real-time automatic lameness detection based on back posture extraction in dairy cattle: Shape analysis of cow with image processing techniques

In this paper results on utilizing image analysis techniques towards early lameness detection in dairy cattle are presented. Data from two different dairy farms in Belgium were gathered. Preprocessing on raw data is required because of non-predictable behaviours of cows such as stopping for a while in front of the camera or non-uniform walking behaviour during experiments. Prelocalization of cow in each frame has been done based on two steps separation: (1) A coarse estimation of moving objects was obtained through background subtraction, (2) second statistical analysis of intensities in gray-scale image along with binarization was utilized to detect moving object in video. A common problem in on-farm collected videos is the similarity of the background and the cow's body colour since the use of classic algorithms for segmentation purposes does not work. Here a hierarchy background/foreground exaggeration is proposed to segment the cow in each frame and track it in video. The combination of logarithm and exponential, background subtraction as well as statistical filtering are used to find the accurate shape of the cow. Furthermore, the back posture of each cow during standing and walking was extracted automatically. It was done by detecting the arc of back posture and fitting a circle through selected points on the spine line. The average inverse radius of four frames displaying the hind hoofs in contact with the ground (two frames for each hoof in a row) was assigned to the cow. Based on this curvature value, a score representing the status of lameness in the individual cow was given automatically. Experimental results from two different databases show promising results in automatic lameness detection based on back posture information.     

基于变差函数的中高分辨率SAR影像农村建筑区提取

目的 基于中高分辨率影像进行大范围的农村建筑区提取时,由于影像分辨率的限制以及农村建筑区规划自身规划特点等因素,造成了传统变差函数方法的高错分误差。为了准确提取农村建筑区,为后续获取建筑密度和人口密度等工作建立基础,提出了一种基于迭代P参数法的阈值确定方法。方法 通过设定亮度阈值,在变差函数纹理计算中为满足条件的像元点赋以权值。本文方法确保在4个方向都满足条件的像元点(认为是建筑区)获得较大的变差函数值加成,而仅在一个方向或者没有方向满足条件的像元点(认为是非建筑区)获得较小加成或不变,以此改进传统变差函数方法,抑制了农村建筑区与周边非建筑区的混淆。结果 以Radarsat-2的多个极化波段影像为数据源进行了实验,改进变差函数方法在实验区1与实验区2的各个波段平均检测率分别为91.58%和90.11%,平均错分误差分别为19.83%和31.87%。结论 与传统变差函数方法以及最小距离法相比,既保证了较高的检测率,同时显著降低了错分误差,不足之处是在建筑区与非建筑区的边缘处以及与建筑区具有相似纹理特征的非建筑区处出现错分,需要进一步的研究和完善。     

Analysis and processing of decayed log CT image based on multifractal theory

Nondestructive log testing technology is a new subject that has recently seen rapid development. X-ray computed tomography (CT) scanning technology has been applied to the detection of internal defects in logs for the purpose of obtaining prior information that can be used to arrive at better log sawing decisions. Because sawyers currently cannot see the inside of a log until the log faces are revealed by sawing, there is little perceived need to obtain scanned images as detailed as those obtained by CT imaging system. Thus, the recognition of internal defects has become increasingly important. The traditional means of describing objects, based on the well-known Euclidean geometry, is not capable of describing different natural objects and phenomena. On the contrary, fractal geometry and its multifractal extension are new tools which can be used for describing, modeling, analyzing and processing different complex shapes and images. A method in log CT image edge detection based on multifractal theory was applied in this paper. The Hölder exponent α (x,y) of image pixels was computed first, then its multifractal spectrum f (α) was estimated and different image pixels were classified, with f(α) = 1 corresponding to smoothing edge point and 1 ≤ f(α) < 1.5 to a singular edge point. Based on multifractal theory, the set of both singular edge points and smoothing edge points is the set of image edge points. Experimental results showed that the method of log CT image in the edge detection based on multifractal theory was a more effective and more local method than classical method of edge testing.