In this paper, we exploit prior information from global positioning systems and inertial measurement units to speed up the process of large scene reconstruction from images acquired by Unmanned Aerial Vehicles. We utilize weak pose information and intrinsic parameter to obtain the projection matrix for each view. As compared to unmanned aerial vehicles' flight altitude, topographic relief can usually be ignored, we assume that the scene is flat and use weak perspective camera to get projective transformations between two views. Furthermore, we propose an overlap criterion and select potentially matching view pairs between projective transformed views. A robust global structure from motion method is used for image based reconstruction. Our real world experiments show that the approach is accurate, scalable and computationally efficient. Moreover, projective transformations between views can also be used to eliminate false matching.
The measurement of cloud motion is very useful in weather forecast and natural disaster management. This paper is focus on accurately estimating cloud motion from a sequences of satellite images. Due to the complexity of cloud motion, which is a non-rigid movement and implying non-linear events, we cannot adopt some simple motion models and need to develop new algorithms. We presented a new method for cloud motion measurement based on image matching. We use the Iterative Multigrid Image Deformation (IMID) technique to measure the cloud movement at sub pixel accuracy, and for the alignment of image sub-regions differing in translation, rotation angle, and uniform scale factor, we change the correlation method from discrete Cartesian cross correlations to the phase correlation based on the Fourier-Mellin Transformation (FMT) which is invariant to translation, rotation and scaling. The phase correlation based on FMT can directly estimate the rotation angle and scale factor between satellite images. For cloud regions with large rotation angle or scale factors, our method can get more accurate motion estimation than traditional correlations by searching the deformation parameters using Cartesian cross correlation. In addition, the iterative multigrid framework aims at improving the precision of motion measurement by refining the size of cloud regions. To validate the performance of our algorithm, we process a cloudy satellite image with known geometric transformation, including translation, rotation and scaling to simulate a sequence of satellite images, and apply our method to measure the velocity fields of clouds. We also apply our algorithm to the sequence of real satellite images. Our results show that IMID technique with FMT can significantly decrease the displacement error compared to traditional correlation methods, especially in regions with large velocity gradients or high rates of rotation.