Accuracy assessment of high resolution satellite imagery orientation by leave-one-out method

Interest in high-resolution satellite imagery (HRSI) is spreading in several application fields, at both scientific and commercial levels. Fundamental and critical goals for the geometric use of this kind of imagery are their orientation and orthorectification, processes able to georeference the imagery and correct the geometric deformations they undergo during acquisition. In order to exploit the actual potentialities of orthorectified imagery in Geomatics applications, the definition of a methodology to assess the spatial accuracy achievable from oriented imagery is a crucial topic.In this paper we want to propose a new method for accuracy assessment based on the Leave-One-Out Cross-Validation (LOOCV), a model validation method already applied in different fields such as machine learning, bioinformatics and generally in any other field requiring an evaluation of the performance of a learning algorithm (e.g. in geostatistics), but never applied to HRSI orientation accuracy assessment.The proposed method exhibits interesting features which are able to overcome the most remarkable drawbacks involved by the commonly used method (Hold-Out Validation — HOV), based on the partitioning of the known ground points in two sets: the first is used in the orientation–orthorectification model (GCPs — Ground Control Points) and the second is used to validate the model itself (CPs — Check Points). In fact the HOV is generally not reliable and it is not applicable when a low number of ground points is available.To test the proposed method we implemented a new routine that performs the LOOCV in the software SISAR, developed by the Geodesy and Geomatics Team at the Sapienza University of Rome to perform the rigorous orientation of HRSI; this routine was tested on some EROS-A and QuickBird images. Moreover, these images were also oriented using the world recognized commercial software OrthoEngine v. 10 (included in the Geomatica suite by PCI), manually performing the LOOCV since only the HOV is implemented.The software comparison guaranteed about the overall correctness and good performances of the SISAR model, whereas the results showed the good features of the LOOCV method.     

高分辨率卫星影像车辆检测的抗体网络

将车辆检测的过程视为一种“抗体”检测“危险抗原”的过程, 其中车辆是“危险抗原”, 车辆检测模板是“抗体”。利用一些车辆图像作为训练样本, 建立一种抗体网络学习并获取一组优化的“抗体”。这些“抗体”经过与待测影像的匹配, 实现对道路车辆目标的有效提取。采用0.6m分辨率的QuickBird全色数据进行实验, 实验结果验证了该方法的有效性和可行性。     

附加视线向量修正的卫星影像区域网平差

针对传统有理函数模型(RFM)区域网平差方法局限于姿态和轨道测量误差小、相机视场角小及影像交会角良好的情况,提出了附加视线向量修正的卫星影像区域网平差方法。首先利用影像附带的有理多项式系数(RPC)计算出像元视线向量,其次根据该视线向量恢复成像时刻虚拟位置和姿态信息,然后对恢复的虚拟位置和姿态构建误差补偿模型,最后通过最小二乘方法整体解算模型参数和连接点物方坐标。该方法从系统误差产生的原因构建补偿模型,可以规避传统区域网平差方法的近似假设和条件限制。通过对模拟数据以及多套测绘卫星和非测绘卫星数据进行试验的结果表明,该方法处理大姿态角误差、大视场角以及弱交会角等各种严苛条件下的卫星影像能达到比传统方法更好的效果。     

基于中间件的高分辨率卫星影像正射纠正的分布式处理

当前摄影测量系统软件正向分布式并行处理方向发展,目前大家研究比较多的是在高性能计算机系统(如刀片机)下进行并行数据处理,而研究普通PC机集群分布式并行处理数据则相对比较少。因此本文在多PC机环境下按照分布式处理思想提出了一种对高分辨率卫星影像进行正射纠正的基于中间件模式的分布式并行处理系统,着重叙述了该系统的组成以及任务调度中的相关处理等问题,并在实践中验证了它的可行性。     

多条带WorldView卫星图像几何定位精度分析

本文针对多个条带WorldView卫星图像,研究了基于有理函数模型的附加参数区域网平差方法,对高分辨率卫星图像定位精度进行了详细分析。青海地区WorldView卫星图像试验结果表明:无地面控制点时,多条带整体区域网定位结果优于各个条带单独平差结果,区域网平差方法结果优于直接前方交会方法;有地面控制点时,多条带区域网平差结果与各个条带单独平差结果相当,且沿路线不大于10km布设一个地面控制点,其精度可满足1∶2000比例尺加密精度要求,可用于公路初步设计。     

山地地区高分卫星影像正射纠正

随着国内外对地观测卫星技术的不断进步,山地地区高分卫星影像获取能力大幅提升。针对山地区域高分卫星影像云雾多、成像阴影多、纠正位置精度控制难等问题,该文提出山地区域高分卫星正射纠正生产方案。试验证明,经过大气校正、可视化的地形修复等处理后可解决薄雾去除和影像扭曲变形等问题;同时,GPU并行计算可实现影像快速融合、控制点自动选择等操作,提高生产效率,可为山地区域高分卫星影像快速处理提供参照。     

Mapping post-fire forest regeneration and vegetation recovery using a combination of very high spatial resolution and hyperspectral satellite imagery

Careful evaluation of forest regeneration and vegetation recovery after a fire event provides vital information useful in land management. The use of remotely sensed data is considered to be especially suitable for monitoring ecosystem dynamics after fire. The aim of this work was to map post-fire forest regeneration and vegetation recovery on the Mediterranean island of Thasos by using a combination of very high spatial (VHS) resolution (QuickBird) and hyperspectral (EO-1 Hyperion) imagery and by employing object-based image analysis. More specifically, the work focused on (1) the separation and mapping of three major post-fire classes (forest regeneration, other vegetation recovery, unburned vegetation) existing within the fire perimeter, and (2) the differentiation and mapping of the two main forest regeneration classes, namely, Pinus brutia regeneration, and Pinus nigra regeneration. The data used in this study consisted of satellite images and field observations of homogeneous regenerated and revegetated areas. The methodology followed two main steps: a three-level image segmentation, and, a classification of the segmented images. The process resulted in the separation of classes related to the aforementioned objectives. The overall accuracy assessment revealed very promising results (approximately 83.7% overall accuracy, with a Kappa Index of Agreement of 0.79). The achieved accuracy was 8% higher when compared to the results reported in a previous work in which only the EO-1 Hyperion image was employed in order to map the same classes. Some classification confusions involving the classes of P. brutia regeneration and P. nigra regeneration were observed. This could be attributed to the absence of large and dense homogeneous areas of regenerated pine trees in the study area.     

高分辨率立体测绘卫星技术研究

本文针对高精度立体测绘卫星设计和实现中的关键技术难点,在充分分析国内外测绘卫星的发展历程和技术特点的基础上,结合测绘卫星的设计关键——高图像定位精度技术实现,对高分辨率立体测绘卫星的设计约束条件、测绘体制选取、卫星载荷和平台关键产品的设计重点难点等进行了分析研究。分析指出了三线阵测绘体制、两线阵测绘体制和单线阵测绘体制的技术特点、实现约束和在测绘卫星不同发展阶段的工程实现优势;明确了基于目前工程技术水平,两线阵测绘体制在大范围、高分辨率、高精度测绘卫星中应用的特有优势。提出了测绘卫星高定位精度关键技术的设计要素和解决途径。结合国内首颗亚米级高精度立体测绘卫星——高分七号（GF-7）卫星的设计状态,说明卫星在保证测绘任务要求方面所提出的多项技术创新,并给出卫星用户对卫星在国土测绘及其他扩展应用中的测试结果。在轨数据表明,依照本论文提出的高分辨率立体测绘卫星系统设计方法,高分七号卫星在轨性能全面满足且部分优于设计指标,达到了世界的领先水平。论文的研究成果为后续更大比例尺的立体测绘卫星设计提供了有力参考。     

An improved geopositioning model of QuickBird high resolution satellite imagery by compensating spatial correlated errors

A lot of studies have been done for correcting the systematic biases of high resolution satellite images (HRSI), which is a fundamental work in the geometric orientation and the geopositioning of HRSI. All the existing bias-corrected models eliminate the biases in the images by expressing the biases as a function of some deterministic parameters (i.e. shift, drift, or affine transformation models), which is indeed effective for most of the commercial high resolution satellite imagery (i.e. IKONOS, GeoEye-1, WorldView-1/2) except for QuickBird. Studies found that QuickBird is the only one that needs more than a simple shift model to absorb the strong residual systematic errors. To further improve the image geopositioning of QuickBird image, in this paper, we introduce space correlated errors (SCEs) and model them as signals in the bias-corrected rational function model (RFM) and estimate the SCEs at the ground control points (GCPs) together with the bias-corrected parameters using least squares collocation. With these estimated SCEs at GCPs, we then predict the SCEs at the unknown points according to their stochastic correlation with SCEs at the GCPs. Finally, we carry out geopositioning for these unknown points after compensating both the biases and the SCEs. The performance of our improved geopositioning model is demonstrated with a stereo pair of QuickBird cross-track images in the Shanghai urban area. The results show that the SCEs exist in HRSI and the presented geopositioning model exhibits a significant improvement, larger than 20% in both latitude and height directions and about 2.8% in longitude direction, in geopositioning accuracy compared to the common used affine transformation model (ATM), which is not taking SCEs into account. The statistical results also show that our improved geopositioning model is superior to the ATM and the second polynomial model (SPM) in both accuracy and reliability for the geopositioning of HRSI.     

Improving georeferencing accuracy of Very High Resolution satellite imagery using freely available ancillary data at global coverage

ABSTRACT

While impressive direct geolocation accuracies better than 5.0?m CE90 (90% of circular error) can be achieved from the last DigitalGlobe’s Very High Resolution (VHR) satellites (i.e. GeoEye-1 and WorldView-1/2/3/4), it is insufficient for many precise geodetic applications. For these sensors, the best horizontal geopositioning accuracies (around 0.55?m CE90) can be attained by using third-order 3D rational functions with vendor’s rational polynomial coefficients data refined by a zero-order polynomial adjustment obtained from a small number of very accurate ground control points (GCPs). However, these high-quality GCPs are not always available. In this work, two different approaches for improving the initial direct geolocation accuracy of VHR satellite imagery are proposed. Both of them are based on the extraction of three-dimensional GCPs from freely available ancillary data at global coverage such as multi-temporal information of Google Earth and the Shuttle Radar Topography Mission 30?m digital elevation model. The application of these approaches on WorldView-2 and GeoEye-1 stereo pairs over two different study sites proved to improve the horizontal direct geolocation accuracy values around of 75%.     

Leaf area index retrieval using gap fractions obtained from high resolution satellite data: Comparisons of approaches,scales and atmospheric effects

This study is aimed at demonstrating the feasibility of the large scale LAI inversion algorithms using red and near infrared reflectance obtained from high resolution satellite imagery. Radiances in digital counts were obtained in 10 m resolution acquired on cloud free day of August 23, 2007, by the SPOT 5 high resolution geometric (HRG) instrument on mostly temperate hardwood forest located in the Great Lakes – St. Lawrence forest in Southern Quebec. Normalized difference vegetation index (NDVI), scaled difference vegetation index (SDVI) and modified soil-adjusted vegetation index (MSAVI) were applied to calculate gap fractions. LAI was inverted from the gap fraction using the common Beer–Lambert's law of light extinction under forest canopy. The robustness of the algorithm was evaluated using the ground-based LAI measurements and by applying the methods for the independently simulated reflectance data using PROSPECT + SAIL coupled radiative transfer models. Furthermore, the high resolution LAI was compared with MODIS LAI product. The effects of atmospheric corrections and scales were investigated for all of the LAI retrieval methods. NDVI was found to be not suitable index for large scale LAI inversion due to the sensitivity to scale and atmospheric effects. SDVI was virtually scale and atmospheric correction invariant. MSAVI was also scale invariant. Considering all sensitivity analysis, MSAVI performed best followed by SDVI for robust LAI inversion from high resolution imagery.     

Automated mapping of impervious surfaces in urban and suburban areas: Linear spectral unmixing of high spatial resolution imagery

Quantifying impervious surfaces in urban and suburban areas is a key step toward a sustainable urban planning and management strategy. With the availability of fine-scale remote sensing imagery, automated mapping of impervious surfaces has attracted growing attention. However, the vast majority of existing studies have selected pixel-based and object-based methods for impervious surface mapping, with few adopting sub-pixel analysis of high spatial resolution imagery. This research makes use of a vegetation-bright impervious-dark impervious linear spectral mixture model to characterize urban and suburban surface components. A WorldView-3 image acquired on May 9th, 2015 is analyzed for its potential in automated unmixing of meaningful surface materials for two urban subsets and one suburban subset in Toronto, ON, Canada. Given the wide distribution of shadows in urban areas, the linear spectral unmixing is implemented in non-shadowed and shadowed areas separately for the two urban subsets. The results indicate that the accuracy of impervious surface mapping in suburban areas reaches up to 86.99%, much higher than the accuracies in urban areas (80.03% and 79.67%). Despite its merits in mapping accuracy and automation, the application of our proposed vegetation-bright impervious-dark impervious model to map impervious surfaces is limited due to the absence of soil component. To further extend the operational transferability of our proposed method, especially for the areas where plenty of bare soils exist during urbanization or reclamation, it is still of great necessity to mask out bare soils by automated classification prior to the implementation of linear spectral unmixing.     

Predicting forest carbon stocks from high resolution satellite data in dry forests of Zimbabwe: exploring the effect of the red-edge band in forest carbon stocks estimation

In this study, we tested whether the inclusion of the red-edge band as a covariate to vegetation indices improves the predictive accuracy in forest carbon estimation and mapping in savanna dry forests of Zimbabwe. Initially, we tested whether and to what extent vegetation indices (simple ratio SR, soil-adjusted vegetation index and normalized difference vegetation index) derived from high spatial resolution satellite imagery (WorldView-2) predict forest carbon stocks. Next, we tested whether inclusion of reflectance in the red-edge band as a covariate to vegetation indices improve the model's accuracy in forest carbon prediction. We used simple regression analysis to determine the nature and the strength of the relationship between forest carbon stocks and remotely sensed vegetation indices. We then used multiple regression analysis to determine whether integrating vegetation indices and reflection in the red-edge band improve forest carbon prediction. Next, we mapped the spatial variation in forest carbon stocks using the best regression model relating forest carbon stocks to remotely sensed vegetation indices and reflection in the red-edge band. Our results showed that vegetation indices alone as an explanatory variable significantly (p < 0.05) predicted forest carbon stocks with R² ranging between 45 and 63% and RMSE ranging from 10.3 to 12.9%. However, when the reflectance in the red-edge band was included in the regression models the explained variance increased to between 68 and 70% with the RMSE ranging between 9.56 and 10.1%. A combination of SR and reflectance in the red edge produced the best predictor of forest carbon stocks. We concluded that integrating vegetation indices and reflectance in the red-edge band derived from high spatial resolution can be successfully used to estimate forest carbon in dry forests with minimal error.