Automatic target detection in satellite images is a challenging problem due to the varying size, orientation and background of the target object. The traditionally engineered features such as HOG, Gabor feature and Hough transform do not... more
Automatic target detection in satellite images is a challenging problem due to the varying size, orientation and background of the target object. The traditionally engineered features such as HOG, Gabor feature and Hough transform do not work well for huge data of high resolution. Robust and computationally efficient systems are required which can learn presentations from the massive satellite imagery. In this paper, a target detection system for satellite imagery is proposed which uses EdgeBoxes and Convolutional Neural Network (CNN) for classifying target and non-target objects in a scene. The edge information of targets in satellite imagery contains very prominent and concise attributes. EdgeBoxes uses the edge information to filter the set of target proposals. CNN is a deep learning classifier with a high learning capacity and a capability of automatically learning optimum features from training data. Moreover, CNN is invariant to minor rotations and shifts in the target object. Encouraging experimental results have been obtained on a large dataset which shows the optimum performance and robustness of our system in complex scenes.
Earthquakes are one of the most catastrophic natural phenomena. After an earthquake, earthquake reconnaissance enables effective recovery by collecting data on building damage and other impacts. This paper aims to identify... more
Earthquakes are one of the most catastrophic natural phenomena. After an earthquake, earthquake reconnaissance enables effective recovery by collecting data on building damage and other impacts. This paper aims to identify state-of-the-art data sources for building damage assessment and provide guidance for more efficient data collection. We have reviewed 39 articles that indicate the sources used by different authors to collect data related to damage and post-disaster recovery progress after earthquakes between 2014 and 2021. The current data collection methods have been grouped into seven categories: fieldwork or ground surveys, omnidirectional imagery (OD), terrestrial laser scanning (TLS), remote sensing (RS), crowdsourcing platforms, social media (SM) and closed-circuit television videos (CCTV). The selection of a particular data source or collection technique for earthquake reconnaissance includes different criteria depending on what questions are to be answered by these data. We conclude that modern reconnaissance missions cannot rely on a single data source. Different data sources should complement each other, validate collected data or systematically quantify the damage. The recent increase in the number of crowdsourcing and SM platforms used to source earthquake reconnaissance data demonstrates that this is likely to become an increasingly important data source.
Remote sensing data and image classification algorithms can be very useful in the identification of beach patterns and therefore can be used as inputs in beach classification models. In this work, one aerial photograph, one IKONOS-2 image... more
Remote sensing data and image classification algorithms can be very useful in the identification of beach patterns and therefore can be used as inputs in beach classification models. In this work, one aerial photograph, one IKONOS-2 image and one FORMOSAT-2 image were applied to a part of the northwest coast of Portugal. Several image processing algorithms were employed and compared: pixel-based approach, object-based approach, Principal Components Analysis (PCA), Artificial Neural Network (ANN) and Decision Trees (DT). The ANN and DT algorithms employed conduced to better results than the traditional classification methodologies (pixel-based, object-based and PCA), allowed a more accurate identification of rip currents. Regarding the data used, the high spatial resolution of aerial photograph allows for the better discrimination of different micro patterns. The FORMOSAT-2 image presents a lower spatial resolution, which did not allow for the identification of small microforms. Concluding, the conjugation of better spatial and spectral resolution of IKONOS-2 data and the data mining algorithms seems to be the better approach to accurately identify beach patterns through remotely sensed data.
Earthquakes are one of the most catastrophic natural phenomena. After an earthquake, earthquake reconnaissance enables effective recovery by collecting data on building damage and other impacts. This paper aims to identify... more
Earthquakes are one of the most catastrophic natural phenomena. After an earthquake, earthquake reconnaissance enables effective recovery by collecting data on building damage and other impacts. This paper aims to identify state-of-the-art data sources for building damage assessment and provide guidance for more efficient data collection. We have reviewed 39 articles that indicate the sources used by different authors to collect data related to damage and post-disaster recovery progress after earthquakes between 2014 and 2021. The current data collection methods have been grouped into seven categories: fieldwork or ground surveys, omnidirectional imagery (OD), terrestrial laser scanning (TLS), remote sensing (RS), crowdsourcing platforms, social media (SM) and closed-circuit television videos (CCTV). The selection of a particular data source or collection technique for earthquake reconnaissance includes different criteria depending on what questions are to be answered by these data....
An ongoing challenge in high-resolution satellite image (HRSI) processing has been the establishment of accurate epipolar geometry over the entire HRSI image area, which is a pivotal step of any stereo-image processing, including 3D... more
An ongoing challenge in high-resolution satellite image (HRSI) processing has been the establishment of accurate epipolar geometry over the entire HRSI image area, which is a pivotal step of any stereo-image processing, including 3D topographic mapping. However, the push broom camera which is used by most high-resolution satellites does not produce straight epipolar lines. Furthermore, in contrast to the well -known frame cameras, the epipolar curve pair does not exist for the entire stereo image area. These properties make it difficult to establish the epipolar geometry of a push broom camera, resulting in a limited accuracy of the epipolar image resampling. In this study, a new method of epipolar curve pair determination and epipolar image resampling of space borne push broom imagery, based on the popular RPC (Rational Polynomial Coefficients), is proposed.
