Thet Su Win, Sao Hone Pha
UAV photography system is implementation of efficient data acquisition techniques, high resolution texture, low cost and using in commercial application. In research paper, we got several images from UAV (unmanned aerial vehicle) and produced image mosaic. Firstly, this paper proposes image mosaic method based of SIFT (scale invariance feature transform) feature to detect key points, scaling, rotation and matching between two images in the image mosaic. RANSAC (random sample consensus) method is used to find homography, transformation and adjust colour for rgb or grey scale for grey image with SIFT (scale invariance feature transform) matching location and then produced mosaic image from combination of SIFT (scale invariance feature transform) and RANSAC (random sample consensus). The results from experiment based on four pairs of images with 70% overlapped captures by using the camera of UAV show that our method has many feature points and matching points for image mosaic. In this paper, the combination of SIFT and RANSAC algorithm is used to produce mosaic image and then this image is used in geometric correction. Finally, this paper proposes before and after geometric correction of mosaic image because raw digital images contain geometric distortion and cannot be used directly as a map. In general, there are two approaches for the geometric correction. The parametric approach is model-based while the non-parametric one makes use of ground control points (GCPs). This paper contains the geometric correction of UAV image which is non-parametric approach.In this paper, geometric correction is considered two conditions. These conditions are that distributions of GCPs (Ground Control Points) are considered with distinct location and indistinct location based on geometric metric correction accuracy that is Root Mean Square Error (RMSE).
UAV; SIFT; RANSAC; mosaic; geometric correction; non-parametric; RMSE.