Yong Zhang

Bio: Yong Zhang is an academic researcher from Wuhan University. The author has contributed to research in topics: Orientation (computer vision) & Digital elevation model. The author has an hindex of 5, co-authored 8 publications receiving 64 citations.

Automatic cloud detection for high resolution satellite stereo images and its application in terrain extraction

Abstract: The automatic extraction of terrain from high-resolution satellite optical images is very difficult under cloudy conditions. Therefore, accurate cloud detection is necessary to fully use the cloud-free parts of images for terrain extraction. This paper addresses automated cloud detection by introducing an image matching based method under a stereo vision framework, and the optimization usage of non-cloudy areas in stereo matching and the generation of digital surface models (DSMs). Given that clouds are often separated from the terrain surface, cloudy areas are extracted by integrating dense matching DSM, worldwide digital elevation model (DEM) (i.e., shuttle radar topography mission (SRTM)) and gray information from the images. This process consists of the following steps: an image based DSM is firstly generated through a multiple primitive multi-image matcher. Once it is aligned with the reference DEM based on common features, places with significant height differences between the DSM and the DEM will suggest the potential cloud covers. Detecting cloud at these places in the images then enables precise cloud delineation. In the final step, elevations of the reference DEM within the cloud covers are assigned to the corresponding region of the DSM to generate a cloud-free DEM. The proposed approach is evaluated with the panchromatic images of the Tianhui satellite and has been successfully used in its daily operation. The cloud detection accuracy for images without snow is as high as 95%. Experimental results demonstrate that the proposed method can significantly improve the usage of the cloudy panchromatic satellite images for terrain extraction.

Fully automatic generation of geoinformation products with chinese zy-3 satellite imagery

Abstract: The advantages of continuously and soundly obtaining large multidimensional, multiscale and multitemporal observation datasets from satellite remote sensing make it indispensable in building a national spatial data infrastructure. This paper introduces the ZY-3 satellite developed in China and discusses a fully automatic data-processing system to generate geoinformation products, such as digital elevation models (DEMs) and digital orthophotomaps (DOMs), based on ZY-3 imagery. The key technologies of automatic geoinformation product generation, including strip image-based bundle adjustment together with creating DEMs and DOMs, are illustrated. The accuracies of the georeferencing and automatically generated geoinformation products are also discussed. This automatic dataprocessing system is shown to provide a good foundation for near real-time derivation of such geoinformation products and for the promotion and application of Chinese domestic satellites.

Estimation of Transformation Parameters Between Centre‐Line Vector Road Maps and High Resolution Satellite Images

Abstract: A method for automatically estimating the transformation parameters between road centre-line vector maps and high resolution satellite images is proposed. The advantages of the method are that global image feature extraction is avoided and feature extraction and matching are achieved simultaneously by using the vector data as guidance. The road width, as estimated by the algorithm, together with the road direction are used as constraints to refine the matching results. Arbitrarily chosen road nodes contribute to improving the adjustment. Map-to-image matching has advantages over image-to-image matching and could be a good method for the rapid updating of geographical information system (GIS) data. Resume Une methode est proposee pour estimer automatiquement les parametres de transformation entre des cartes routieres en mode vecteur et des images satellitales a haute resolution. Les avantages de cette methode sont qu'elle n'a pas besoin d'extraction globale d'objets dans les images, et que l'extraction et l'appariement se font simultanement en s'appuyant sur les donnees vectorielles. La largeur des routes, qui est estimee par l'algorithme, ainsi que leur direction, sont utilisees comme contraintes pour affiner les resultats de l'appariement. Des nœuds routiers choisis arbitrairement contribuent a ameliorer le recalage. Le recalage de la carte sur l'image presente des avantages par rapport au recalage entre images, et pourrait etre un moyen approprie pour une mise a jour rapide des donnees dans les systemes d'information geographique. Zusammenfassung Dieser Beitrag stellt eine Methode zur automatischen Bestimmung der Transformationsparameter zwischen Strasenachsen aus vektoriellen Kartendaten und hochauflosenden Satellitenbilddaten vor. Merkmalsextraktion und -zuordnung werden, gesteuert durch die Vektordaten, simultan gelost. Strasenbreite und -richtung werden durch den Algorithmus bestimmt und dienen als Bedingungen, um die Zuordnungsergebnisse zu verbessern. Beliebige Strasenkreuzungen konnen fur die Ausgleichung unterstutzend wirken. Die Zuordnung von Karte zu Bild hat Vorteile gegenuber einer Bild zu Bildzuordnung und konnte somit fur eine zugige Fortfuhrung von GIS Daten vorteilhaft sein. Resumen Se propone un metodo para estimar de forma automatica la transformacion entre la linea central de carretera en mapas vectoriales e imagenes de satelite de alta resolucion. Las ventajas del metodo son que se evita la extraccion global de caracteristicas de la imagen y la extraccion de caracteristicas y su correspondencia se realizan simultaneamente usando el mapa vectorial como guia. El ancho de la via, estimada por el algoritmo, conjuntamente con la direccion de la carretera son usados como restricciones para refinar el resultado de la correspondencia. Los nodos arbitrarios de la carretera contribuyen a mejorar el ajuste. La correspondencia mapa a imagen tiene ventajas frente a la correspondencia imagen a imagen y puede ser un buen metodo para la actualizacion rapida de datos SIG.

Geometrical Consistency Voting Strategy for Outlier Detection in Image Matching

Abstract: False matches in tie-point image matching are common. This paper introduces a straightforward and effective prepossessing method to reject false matches from initial matches. The method is based on the idea of Hough transform using only two geometrical consistency parameters, namely, the scale parameter and the rotation parameter between two images. A weighted voting strategy is employed, and it can further improve the robustness of the algorithm. The method can handle a large rate of outliers and produce more robust matches with low complexity. No assumptions with regard to the relative pose between two images are necessary, and large perspective deformation can be handled as well. Experiments with ground reference data show that the algorithm works effectively even when the ratio of inliers is below 10 percent. In these data, the ratio of inliers can be improved from 5 percent to 40 percent on average.

Registration of CBERS-02B satellite imagery in quick GIS updating

Abstract: Three key techniques in registration of CBERS-02B image are introduced in this paper. Including an improved image matching strategy aided by a voting algorithm, the voting algorithm directly estimates the transformation parameters of image instead of finding the one to one corresponding, making the image matching much more efficient and reliable. Beside that, the method of handling the interior deformation problem caused by the incorrect splicing of three sub CCD arrays is introduced. At last, a virtual combination of satellite image technique is introduced, the adjacent satellite images in the same orbit are virtually combined to form a large image which is taken as the orientation unit, purposed on accelerating the orientation process speed and decreasing the probability of orientation failure. With the techniques mentioned above, the two main problems about CBERS-02B image: the inaccurate ephemeris parameters and the poor internal accuracy caused by the incorrect splicing of three sub CCD arrays are solved successfully, the absolute positioning accuracy reaches 1:10000 mapping accuracy and the internal accuracy reaches sub-pixel after the orientation process. It indicates that extensive usage of Chinese high resolution satellite imagery may become possible and practical.

Deep learning based cloud detection for medium and high resolution remote sensing images of different sensors

Abstract: Cloud detection is an important preprocessing step for the precise application of optical satellite imagery. In this paper, we propose a deep learning based cloud detection method named multi-scale convolutional feature fusion (MSCFF) for remote sensing images of different sensors. In the network architecture of MSCFF, the symmetric encoder-decoder module, which provides both local and global context by densifying feature maps with trainable convolutional filter banks, is utilized to extract multi-scale and high-level spatial features. The feature maps of multiple scales are then up-sampled and concatenated, and a novel multi-scale feature fusion module is designed to fuse the features of different scales for the output. The two output feature maps of the network are cloud and cloud shadow maps, which are in turn fed to binary classifiers outside the model to obtain the final cloud and cloud shadow mask. The MSCFF method was validated on hundreds of globally distributed optical satellite images, with spatial resolutions ranging from 0.5 to 50 m, including Landsat-5/7/8, Gaofen-1/2/4, Sentinel-2, Ziyuan-3, CBERS-04, Huanjing-1, and collected high-resolution images exported from Google Earth. The experimental results show that MSCFF achieves a higher accuracy than the traditional rule-based cloud detection methods and the state-of-the-art deep learning models, especially in bright surface covered areas. The effectiveness of MSCFF means that it has great promise for the practical application of cloud detection for multiple types of medium and high-resolution remote sensing images. Our established global high-resolution cloud detection validation dataset has been made available online ( http://sendimage.whu.edu.cn/en/mscff/ ).

Efficient structure from motion for large-scale UAV images: A review and a comparison of SfM tools

Abstract: Unmanned aerial vehicle (UAV) images have gained extensive attention in varying fields, and the Structure from Motion (SfM) technique has become the gold standard for aerial triangulation of UAV images. With increasing data volume caused by the use of multi-view and high-resolution imaging systems and the enhancement of UAV platform’s endurance, the capability for orientation of large-scale UAV images is becoming a prominent and necessary feature for SfM-based solutions. A classical SfM pipeline consists of three major steps, i.e., (i) feature extraction for an individual image, (ii) feature matching for each image pair, and (iii) parameter solving based on iterative bundle adjustment. Most of the time costs are consumed in the second and third steps. This can be explained from three main aspects. First, for feature matching the large number of images and high overlapping degrees cause high combinational complexity of match pairs. Second, the efficiency of commonly utilized techniques for outlier removal would be seriously degenerated because of high outlier ratios of initial matches. Third, for parameter solving of camera poses and scene structures, the iterative execution of bundle adjustment (BA) leads to high computational costs in the incremental SfM workflow. Thus, this paper gives a systematic survey of the state-of-the-art for match pair selection from both ordered and unordered datasets, for outlier removal of initial matches dominated by outliers, and for efficiency improvement of BA, and conducts an experimental evaluation for six well-known SfM-based software packages on UAV image orientation.

Cloud/shadow detection based on spectral indices for multi/hyperspectral optical remote sensing imagery

Abstract: Cloud and cloud shadow detection is a necessary preprocessing step for optical remote sensing applications because of the huge negative influence cloud and cloud shadow can have on data analysis. However, most of the existing cloud/shadow detection methods are designed based on specific band configurations of certain sensors, and their working mechanisms are relatively complex and computationally complicated, which limits their application. In view of this, in this paper, a unified cloud/shadow detection algorithm based on spectral indices (CSD-SI) is proposed for most of the widely used multi/hyperspectral optical remote sensing sensors with both visible and infrared spectral channels. On the one hand, the cloud index (CI) and cloud shadow index (CSI) are proposed to indicate the potential clouds and cloud shadows based on their physical reflective characteristics. In addition, considering the spatial coexistence of cloud and cloud shadow, a spatial matching strategy is utilized to remove the pseudo cloud shadows. The effectiveness of the proposed CSD-SI algorithm is demonstrated on eight different types of widely used multi/hyperspectral optical sensors, with different spectral and spatial resolution levels. Overall, in the experiments undertaken in this study, CSD-SI achieved a mean overall accuracy of 98.52% for cloud, with a mean producer’s accuracy of 93.13% and a mean user’s accuracy of 98.13%. For cloud shadow, CSD-SI achieved a means producer’s accuracy of 84.33% and a mean user’s accuracy of 89.12%. The experimental results show that the proposed CSD-SI method based on spectral indices can obtain a comparable cloud/shadow detection performance to that of the other state-of-the-art methods.

DABNet: Deformable Contextual and Boundary-Weighted Network for Cloud Detection in Remote Sensing Images

Abstract: In recent years, deep convolutional neural networks (DCNNs) have made significant progress in cloud detection tasks, and the detection accuracy has been greatly improved. However, most existing CNN-based models have high computational complexity, which limits their practical application, especially for spaceborne optical remote sensing. In addition, most of the methods cannot make adaptive adjustments based on the structural information of the clouds, and blurred boundaries often occur in the detection results. In order to address these problems, this article proposes a lightweight network (DABNet) to achieve high-accuracy detection of complex clouds, not only a clearer boundary but also lower false-alarm rate. Specifically, a deformable context feature pyramid module is proposed to improve the adaptive modeling capability of multiscale features. Besides, a boundary-weighted loss function is designed to direct the network to focus on cloud boundary information and optimize the relevant detection results. The proposed method has been validated on two data sets: the public GF-1 WFV benchmark and our self-built GF-2 cloud detection data set with higher spatial resolution. The experimental results exhibit that DABNet achieves state-of-the-art performance while only using 4.12M parameters and 8.29G multiadds.

In-orbit geometric calibration and validation of ZY-3 three-line cameras based on CCD-detector look angles

Abstract: A simple and feasible orientation method for the three-line cameras of the ZiYuan-3 (ZY-3) satellite is introduced. The method calibrates the CCD-detector look angles in the attitude determination reference (ADR) coordinate system rather than modifying the laboratory-calibrated orientation parameters, which are not necessary for the geometric calibration. Additionally, the unknown parameters of the model are virtually uncorrelated. Experimental results show that the direct georeferencing accuracy of ZY-3 nadir images have been improved from the kilometre level to better than 9 m in planimetry, and that of the forward-and-backward stereopair was also raised to better than 18 m in plan and 4 m in height. After a bundle adjustment using four ground control points, a higher accuracy of about 3·4 m in plan and 1·6 m in height was achieved. Resume Cet article presente une methode simple a mettre en oeuvre pour l'orientation des cameras a trois barrettes du satellite ZiYuan-3 (ZY-3). La methode consiste a etalonner les angles de visee des detecteurs CCD dans le repere de determination d'attitude au lieu de modifier les parametres d'orientation determines en laboratoire, qui ne sont pas necessaires pour l’etalonnage geometrique. En outre, les inconnues du modele sont virtuellement decorrelees. Les resultats experimentaux montrent que la precision de georeferencement direct des images ZY-3 prises en visee verticale est amelioree d'un ordre de grandeur kilometrique a mieux que 9 m en planimetrie, et que celle des couples stereoscopiques avant-arriere atteint mieux que 18 m en planimetrie et 4 m en altimetrie. Apres une compensation par faisceaux utilisant quatre points d'appui, la precision est encore amelioree et atteint 3,4 m en planimetrie et 1,6 m en altimetrie. Zusammenfassung Dieser Beitrag stellt eine einfache und praktikable Orientierungsmethode fur die Dreizeilenkameras des ZiYuan-3 (ZY-3) Satelliten vor. Diese Methode kalibriert die Aufnahmerichtungen der CCD-Detektoren in dem Referenzkoordinatensystem (ADR) der Neigungsbestimmung, anstatt die Parameter der Laborkalibrierung zu modifizieren, die fur die geometrische Kalibrierung nicht notwendig sind. Zusatzlich sind die unbekannten Parameter des Models praktisch unkorreliert. Experimentelle Ergebnisse zeigen, dass die Genauigkeit der direkten Georeferenzierung der Nadirbilder des ZY-3 Satelliten sich vom km Niveau auf besser als 9 m in der Lage gesteigert hat, und die Genauigkeit des Stereopaares aus vorwarts und ruckwarts Blick sich auf unter 18 m in der Lage und 4 m in der Hohe verbessert hat. Nach einer Bundelausgleichung mit vier Passpunkten konnte eine noch hohere Genauigkeit von ca. 3·4 m in der Lage und 1·6 m in der Hohe erzielt werden. Resumen Se introduce un metodo simple y asequible de orientacion para el satelite tri-linear ZiYuan-3 (ZY-3). El metodo calibra los angulos de apuntamiento de los detectores CCD en el sistema de referencia de actitud (ADR) en lugar de modificar los parametros de orientacion medidos en laboratorio, los cuales no son necesarios para la calibracion geometrica del satelite. Ademas, los parametros desconocidos del modelo estan practicamente libres de correlacion. Los resultados experimentales demuestran que la precision de la georeferenciacion directa de las imagenes del nadir del ZY-3 han mejorado de nivel kilometrico a niveles mejores que 9 m en planimetria, y que en el caso estereo de los puntos de vista anterior i posterior tambien ha alcanzado niveles mejores que los 18 m en planimetria y 4 m en altimetria. Una vez realizado un ajuste de haces con cuatro puntos de control, se alcanzan precisiones mayores: 3·4 m en planimetria y 1·6 m en altimetria. 摘 要 提出了一种简单 、可行的资源三号卫星三线阵相机几何定位方法。该方法在姿态测量参考坐标系下对CCD探元指向角进行几何检校,而无需重新计算实验室检校获得的定位参数。而且,在几何检校过程中,实验室检校参数并非必需的,检校模型的未知数之间也不存在相关性。实验结果表明,资源三号下视影像直接定位的平面精度由公里级提高到了优于9米,前后视立体像对的平面与高程定位精度也分别提高到了优于18米和4米。利用4个地面控制点进行光束法平差后,可获得的平面精度约3·4米、高程精度约1·6米。