Parallelizing maximum likelihood classification on computer cluster and graphics processing unit for supervised image classification

ABSTRACT

Supervised image classification has been widely utilized in a variety of remote sensing applications. When large volume of satellite imagery data and aerial photos are increasingly available, high-performance image processing solutions are required to handle large scale of data. This paper introduces how maximum likelihood classification approach is parallelized for implementation on a computer cluster and a graphics processing unit to achieve high performance when processing big imagery data. The solution is scalable and satisfies the need of change detection, object identification, and exploratory analysis on large-scale high-resolution imagery data in remote sensing applications.     

基于Gabor纹理描述的遥感影像分割

本文引入了在不同影像层次上的Gabor纹理特征,采用分裂-合并加智能像素精致的方法,实现遥感影像的非监督分割。实验结果表明采用Gabor滤波器为基础的多分辨率分析来描述高分辨率遥感影像的纹理特征,可以明显地描述影像的高、低频特征,并且基于Gabor纹理特征进行遥感影像的分割是有效的;将本文方法的分割结果与经典的eCognition分割结果进行了对比试验,表明本文方法的分割结果较好。     

FORSAT: a 3D forest monitoring system for cover mapping and volumetric 3D change detection

ABSTRACT

A 3D forest monitoring system, called FORSAT (a satellite very high resolution image processing platform for forest assessment), was developed for the extraction of 3D geometric forest information from very high resolution (VHR) satellite imagery and the automatic 3D change detection. FORSAT is composed of two complementary tasks: (1) the geometric and radiometric processing of satellite optical imagery and digital surface model (DSM) reconstruction by using a precise and robust image matching approach specially designed for VHR satellite imagery, (2) 3D surface comparison for change detection. It allows the users to import DSMs, align them using an advanced 3D surface matching approach and calculate the 3D differences and volume changes (together with precision values) between epochs. FORSAT is a single source and flexible forest information solution, allowing expert and non-expert remote sensing users to monitor forests in three and four (time) dimensions. The geometric resolution and thematic content of VHR optical imagery are sufficient for many forest information needs such as deforestation, clear-cut and fire severity mapping. The capacity and benefits of FORSAT, as a forest information system contributing to the sustainable forest management, have been tested and validated in case studies located in Austria, Switzerland and Spain.     

基于相位一致的高分辨率遥感图像分割方法

基于分水岭变换的图像分割性能在很大程度上依赖于用来计算待分割图像梯度的算法。根据频域相位信息对图像特征的表征能力,引入相位一致的思想计算图像特征,应用Log Gabor小波提取高分辨率遥感图像的多尺度梯度。接着在对相位一致梯度进行分水岭分割时发现,在抑制分水岭算法的过度分割方面,经典的基于前景标记和背景标记的方法并不适合于遥感图像的分割,给出一种基于前景标记和梯度重建的分水岭算法。对IKONOS Pan图像上的农田、厂房和居民楼等地物进行特征提取和图像分割实验,结果表明相位一致方法优于空域特征检测算子,根据相位一致特征得到较好的分水岭分割结果。     

A multi-index learning approach for classification of high-resolution remotely sensed images over urban areas

In recent years, it has been widely agreed that spatial features derived from textural, structural, and object-based methods are important information sources to complement spectral properties for accurate urban classification of high-resolution imagery. However, the spatial features always refer to a series of parameters, such as scales, directions, and statistical measures, leading to high-dimensional feature space. The high-dimensional space is almost impractical to deal with considering the huge storage and computational cost while processing high-resolution images. To this aim, we propose a novel multi-index learning (MIL) method, where a set of low-dimensional information indices is used to represent the complex geospatial scenes in high-resolution images. Specifically, two categories of indices are proposed in the study: (1) Primitive indices (PI): High-resolution urban scenes are represented using a group of primitives (e.g., building/shadow/vegetation) that are calculated automatically and rapidly; (2) Variation indices (VI): A couple of spectral and spatial variation indices are proposed based on the 3D wavelet transformation in order to describe the local variation in the joint spectral-spatial domains. In this way, urban landscapes can be decomposed into a set of low-dimensional and semantic indices replacing the high-dimensional but low-level features (e.g., textures). The information indices are then learned via the multi-kernel support vector machines. The proposed MIL method is evaluated using various high-resolution images including GeoEye-1, QuickBird, WorldView-2, and ZY-3, as well as an elaborate comparison to the state-of-the-art image classification algorithms such as object-based analysis, and spectral-spatial approaches based on textural and morphological features. It is revealed that the MIL method is able to achieve promising results with a low-dimensional feature space, and, provide a practical strategy for processing large-scale high-resolution images.     

Comparisons of two global built area land cover datasets in methods to disaggregate human population in eleven countries from the global South

ABSTRACT

Mapping built land cover at unprecedented detail has been facilitated by increasing availability of global high-resolution imagery and image processing methods. These advances in urban feature extraction and built-area detection can refine the mapping of human population densities, especially in lower income countries where rapid urbanization and changing population is accompanied by frequently out-of-date or inaccurate census data. However, in these contexts it is unclear how best to use built-area data to disaggregate areal, count-based census data. Here we tested two methods using remotely sensed, built-area land cover data to disaggregate population data. These included simple, areal weighting and more complex statistical models with other ancillary information. Outcomes were assessed across eleven countries, representing different world regions varying in population densities, types of built infrastructure, and environmental characteristics. We found that for seven of 11 countries a Random Forest-based, machine learning approach outperforms simple, binary dasymetric disaggregation into remotely-sensed built areas. For these more complex models there was little evidence to support using any single built land cover input over the rest, and in most cases using more than one built-area data product resulted in higher predictive capacity. We discuss these results and implications for future population modeling approaches.     

Detecting and Enumerating New Building Structures Utilizing Very‐High Resolution Imaged Data and Image Processing

Abstract

Information on the number and type of new building structures is required by urban and transportation planners and the real estate industry. The goal of this paper is to explore the potential of high resolution imagery for meeting public and private sector demands for information on new buildings. The value of 1 m, 5 m, and 10 m panchromatic and 1 m color scanned aerial photography images acquired in 1997 and 1998 for a study area within the City of San Diego, California is assessed for general change detection and building enumeration. Both semi‐automated and interactive change‐detection approaches are evaluated. We demonstrate that interactive, visual‐based approaches appear to be the most accurate (within 1% of actual count) and efficient approach for generating information on the number of new buildings associated with single family residential land use. More automated approaches to detecting and enumerating image microfeatures may be useful as enhancements for visual‐based assessments and may be practical in areas composed mostly of large buildings associated with commercial and industrial land use. The highest accuracy for automated approaches was an undercounting of 11% for residential buildings and overcounting of 20% for those associated with commercial and industrial land use.     

多层级二值模式的高光谱影像空-谱分类

利用高光谱遥感影像的空间纹理特征，可以提高高光谱遥感影像的分类精度。提出了一种多层级二值模式的高光谱影像空-谱联合分类方法。该方法将高光谱影像转化为局部二值模式特征图像获取像元微观特征，基于特征图像生成多层级特征向量获取像元宏观特征。为验证该方法的有效性，选取PaviaU、Salinas和Chikusei高光谱影像数据，利用核极限学习机分类器，分别针对光谱、局部二值模式、多层级二值模式等特征开展实验。结果表明，多层级二值模式空-谱分类总体精度分别达到97.31%、98.96%和97.85%，明显优于传统光谱、3Gabor空-谱等分类方法。该方法可为高光谱影像分类提供更加有效的类别判定特征，有助于提高影像分类精度并获取更加平滑的分类结果图。     

基于模板分解与递归式滤波的遥感图像快速Gabor纹理特征提取

设计一种在x、y轴方向上进行2维Gabor滤波器模板分解的可行方法,从而避免模板分解时在倾斜方向上进行重采样所带来的效率、精度损失;接着采用递归方法实现分解后的1维滤波器以进一步提高算法效率.采用高斯滤波对Gabor滤波结果进行校正平滑作为纹理特征输出,并采用k-means算法对其进行聚类以验证方法在提取图像纹理区域时的有效性.和以快速傅里叶变换方式实现的Gabor纹理提取方法进行对比,实验表明,该方法在纹理特征提取上的精度损失很小,但在算法执行效率上则有显著的提高.

Abstract：

A fast remotely sensed image texture feature extracting method is proposed. It firstly decomposes a 2-D Gabor filter along x, y axes Into a set of 1-D filters, which avoids the precision and efficiency losing of re-samplingwhich is necessary when the decomposing is carried out along some inclined orientations of an image plane. Besides, a recursive method is implemented to further improve the efficiency of the decomposed 1-D filtering. A Gaussian filter is used to smooth the filtering outputs, which are then subjected to k-means clustering method for textural image segmentation. A comparison between the method and FFT-based Gabor filtering method is carried out. It demonstrates that our method is o feasible and fast way to extract texture features from remotely sensed imagery,for its higher algorithm efficiency and little precision losing.     

Assessment of different topographic correction methods in ALOS AVNIR-2 data over a forest area

Abstract

Because the removal of topographic effects is one the most important pre-processing steps when extracting information from satellite images in digital Earth applications, the problem of differential terrain illumination on satellite imagery has been investigated for at least 20 years. As there is no superior topographic correction method applicable to all areas and all images, a comparison of topographic normalization methods in different regions and images is necessary. In this study, common topographic correction methods were applied on an ALOS AVNIR-2 image of a rugged forest area, and the results were evaluated through different criteria. The results show that the simple correction methods [Cosine, Sun-Canopy-sensor (SCS), and Minnaert correction] are inefficient in exceptionally rough forests. Among the improved correction methods (SCS+C, modified Minnaert, and pixel-based Minnaert), the best result was achieved using a pixel-based Minnaert approach in which a separate correction factor in various slope angles is used. Thus, this method should be considered for topographic correction, especially in forests with severe topography.     

基于SPOT-5遥感影像的城市信息自动提取技术研究

城市信息的提取是城市动态监测和分析的基础,而城市动态监测对社会发展和人类生活具有重要意义。本文基于三峡地区的SPOT-5遥感影像,以城市绿地和建筑物为研究对象,用ENVI FX影像处理软件,对实验区的绿地和建筑物进行多尺度影像分割信息提取。结果表明,采用多尺度分割技术提取高分辨率影像中地物的提取精度更高,并有效地避免了"椒盐现象"。     

高分辨率遥感影像建筑区域局部几何特征提取

及时准确地获取城市建筑区域的空间分布及其变化信息对于城市规划、空间地理数据库建设及区域社会经济分析具有重要意义。本文提出一种基于多尺度Gabor变换和感知聚类方法即张量投票TV (Tensor Voting)相结合的自适应局部几何不变特征检测方法,并将其应用于高空间分辨率遥感影像建筑区域提取。首先,考虑到高分辨率遥感影像复杂的几何结构特征,使用Gabor滤波器组对影像进行多尺度多方向变换检测奇异性特征。然后,在感知聚类框架下,根据张量投票理论将不同方向子带系数位置编码为相应的二阶对称方向张量,为了突出影像几何特征,对不同尺度、不同方向子带中任意像素位置方向张量使用滤波器响应系数加权并求和完成多尺度特征融合。再次,对张量特征分解得到点结构与线结构显著性图并使用非极大抑制提取相应角点和曲线等局部几何特征,同时生成约束准则筛选角点以确定建筑物坐标。最后,利用概率密度估计结合局部角点特征生成全局概率密度场描述影像中像素从属于建筑目标的概率,并使用最大类间方差法(Otsu)阈值分割自动提取居民地多边形区域。使用分辨率分别为0.49 m、0.98 m的Google Earth及0.8 m的高分二号等影像数据集进行实验,实验结果表明本文方法相对于已有的Harris和HSCD点检测算法,在建筑区域提取质量上(Quality)上分别提高了4.79%,5.96%;1.47%,3.76%和1.91%,4.08%。     

An improved approach for the production of satellite-based geospatial reference imagery

Abstract

An innovative and practical satellite image product is described that is ideal for applications in Northern Canada because of its wide area coverage and mapping-quality features. This product is generated from a new procedure developed at the Canada Centre for Remote Sensing (CCRS) for processing Landsat 7 imagery, and by extension, imagery from other Earth Observation satellites. By working with multiple satellite passes, each containing the equivalent of multiple scenes, the new procedure could dramatically reduce the turn-around time for generating georeferenced image products, and also increase their geometric and radiometric accuracy compared to those produced by the current methods. The objective of the process has been to generate satellite image mosaics covering large areas (e.g. >500 000 km²) with uniformly distributed errors at sub-pixel resolution. The paper discusses the theoretical basis of a photogrammetric adjustment for satellite imagery and the results obtained from several tests. The process is generic, involving a sensor model, a satellite orbit model and ground control information; thus it may be easily adapted to any satellite that allows for repeat coverage with overlapping paths. By performing an adjustment to correct the satellite position and attitude data prior to the production of orthoimage products, it is possible to create a mosaic with a single resampling process which minimises both the radiometric and geometric resampling artifacts. The results from three separate tests are presented, along with a discussion of the procedures that were followed in each case. All three tests have successfully demonstrated that sub-pixel sample size errors may be consistently obtained over large areas. A by-product process developed to support the measurement of ground control point coordinates for the satellite adjustment was the automatic matching of geographic features such as lakes and islands in vector data format. This has been a significant development in that it has eliminated manual intervention in the measurement of these features in the imagery, allowing the ground control for entire passes containing several scenes to be obtained in minutes instead of hours.     

Use of Very High Spatial Resolution Remotely Sensed Imagery for Assessing Land‐Cover Changes in Shrub Habitat Preserves of Southern California

Abstract

The goal of this research was to explore the utility of very high spatial resolution, digital remotely sensed imagery for monitoring land‐cover changes in habitat preserves within southern California coastal shrublands. Changes were assessed for Los Penasquitos Canyon Preserve, a large open space in San Diego County, over the 1996 to 1999 period for which imagery was available.

Multispectral, digital camera imagery from two summer dates, three years apart, was acquired using the Airborne Data Acquisition and Registration (ADAR) digital‐camera system. These very high resolution (VHR) image data (1m), composed of three visible and one near‐infrared wavebands (V/NIR), were the primary image input for assessing land cover change. Image‐derived datasets generated from georeferenced and registered ADAR imagery included multitemporal overlays and multitemporal band differencing with threshold selection. Two different multitemporal image classifications were generated from these datasets and compared. Single‐date imagery was analyzed interactively with image‐derived datasets and with information from field observations in an effort to discern change types. A ground sampling survey conducted soon after the 1999 image acquisition provided concurrent ground reference data.

Most changes occurring within the three‐year interval were associated with transitional phenological states and differential precipitation effects on herbaceous cover. Variations in air temperatures and timing of rainfall contributed to differences that the seven‐week image acquisition offset may have caused. Disturbance factors of mechanical clearing, erosion, potentially invasive plants, and fire were evident and their influence on the presence, absence, and type of vegetation cover were likely sources of change signals.

The multitemporal VHR, V/NIR image data enabled relatively fine‐scale land cover changes to be detected and identified. Band differencing followed by multitemporal classification provided an effective means for detecting vegetation increase or decrease. Detailed information on short‐term disturbance effects and long‐term vegetation type conversions can be extracted if image acquisitions are carefully planned and geometric and radiometric processing steps are implemented.     

Estimation and improvement in the geolocation accuracy of rational polynomial coefficients with minimum GCPs using KOMPSAT-3A

ABSTRACT

In this paper, we propose a method to regenerate Rational Polynomial Coefficients (RPCs) using KOMPSAT-3A imagery and to reduce the geolocation error using minimum ground control points (GCPs). To estimate the new RPCs, the physical sensor model fitted to KOMPSAT-3A imagery was utilized and virtual GCPs over the study area were created. The size of the virtual grid used was 20x20x20. To remove the sensor-related errors in physical sensor model, three different image correction models (image coordinate translation model, shift and drift model, and affine transformation model) were additionally applied. We evaluated our proposed method in two areas within Korea, one in urban (Seoul) and one in rural (Goheung) areas. The results showed that there was a significant improvement after applying the suggested approach in the two areas. The image coordinate translation model is suggested in terms of GCP requirement and expected errors estimated from the error propagation analysis using Gauss–Markov Model (GMM).     

Detecting shadows in multi-temporal aerial imagery to support near-real-time change detection

Multi-temporal aerial imagery captured via an approach called repeat station imaging (RSI) facilitates post-hazard assessment of damage to infrastructure. Spectral-radiometric (SR) variations caused by differences in shadowing may inhibit successful change detection based on image differencing. This study evaluates a novel approach to shadow classification based on bi-temporal imagery, which exploits SR change signatures associated with transient shadows. Changes in intensity (brightness from red–green–blue images) and intensity-normalized blue waveband values provide a basis for classifying transient shadows across a range of material types with unique reflectance properties, using thresholds that proved versatile for very different scenes. We derive classification thresholds for persistent shadows based on hue to intensity ratio (H/I) images, by exploiting statistics obtained from transient shadow areas. We assess shadow classification accuracy based on this procedure, and compare it to the more conventional approach of thresholding individual H/I images based on frequency distributions. Our efficient and semi-automated shadow classification procedure shows improved mean accuracy (93.3%) and versatility with different image sets over the conventional approach (84.7%). For proof-of-concept, we demonstrate that overlaying bi-temporal imagery also facilitates normalization of intensity values in transient shadow areas, as part of an integrated procedure to support near-real-time change detection.     

一种基于相位一致的高分辨率遥感图像特征检测方法

精确检测图像边缘特征是进行高分辨率遥感图像分割和识别的关键。空域特征检测算子以解决阶跃形边缘为主,得到的边缘特征对图像的亮度和对比度敏感。本文引入了一种基于频域相位一致的图像特征检测方法,该方法对遥感图像亮度和对比度具有不变性,同时适用于多种边缘特征的检测。使用Log Gabor小波计算IKONOS Pan图像的相位一致多尺度梯度,对农田、道路和厂房等典型地物进行特征检测的结果表明,相位一致算法对图像局部亮度和对比度不敏感;并且对线形物体产生单线响应,不似空域检测算子产生双线响应。最后考察滤波器尺度和方向参数变化及添加高斯噪声对检测结果的影响,发现相位一致算法无需先使用低通滤波去除噪声,因而具有更稳定的特征定位精度;并且抗噪声干扰的能力强,检测结果不会因为噪声而出现波动。基于相位一致的遥感图像不变特征提取,为高分辨率遥感图像的分割和对象识别提供了基础。     

A METHOD OF AUTOMATING THE PRESENTATION OF INFORMATION DURING COMPLEX AIR PHOTO INTERPRETATION

A method of complex image processing, i.e., the simultaneous use of various kinds of remote sensing imagery in the mapping and study of geographic features, is outlined. It features the use of computerized techniques (a) to identify shots or frames of auxiliary types of imagery containing the same specific geographic features identified on the principal type of imagery [this through the scanning of code lines containing information about the coordinates of imaging, flight direction and altitude of the plane or sensing platform, etc.], and (b) to precisely locate the features of interest within these shots or frames. Translated from: Izvestiya vysshykh uchebnykh zavededeniy, Geodeziya i aerofotos'yemka, 1986, No. 1, pp. 86–91.     

Texture-based classification of sub-Antarctic vegetation communities on Heard Island

This study was the first to use high-resolution IKONOS imagery to classify vegetation communities on sub-Antarctic Heard Island. We focused on the use of texture measures, in addition to standard multispectral information, to improve the classification of sub-Antarctic vegetation communities. Heard Island’s pristine and rapidly changing environment makes it a relevant and exciting location to study the regional effects of climate change. This study uses IKONOS imagery to provide automated, up-to-date, and non-invasive means to map vegetation as an important indicator for environmental change. Three classification techniques were compared: multispectral classification, texture based classification, and a combination of both. Texture features were calculated using the Grey Level Co-occurrence Matrix (GLCM). We investigated the effect of the texture window size on classification accuracy. The combined approach produced a higher accuracy than using multispectral bands alone. It was also found that the selection of GLCM texture features is critical. The highest accuracy (85%) was produced using all original spectral bands and three uncorrelated texture features. Incorporating texture improved classification accuracy by 6%.