Rectification of aerial images using piecewise linear transformation

8th International Symposium of the Digital Earth

地球科学;计算机科学

Liew, L.H.^1 ; Lee, B.Y.^1 ; Wang, Y.C.^2 ; Cheah, W.S.^2

Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA Sarawak, Samarahan Campus, Malaysia^1

Faculty of Computer Science and Information Technology, Universiti Malaysia Sarawak, Malaysia^2

关键词: Corresponding control points;    Geometric transformations;    Global transformation;    Image rectification;    Nonparametric approaches;    Piecewise-linear transformation;    Remotely sensed images;    Root mean square errors;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/18/1/012009/pdf
DOI  :  10.1088/1755-1315/18/1/012009

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Aerial images are widely used in various activities by providing visual records. This type of remotely sensed image is helpful in generating digital maps, managing ecology, monitoring crop growth and region surveying. Such images could provide insight into areas of interest that have lower altitude, particularly in regions where optical satellite imaging is prevented due to cloudiness. Aerial images captured using a non-metric cameras contain real details of the images as well as unexpected distortions. Distortions would affect the actual length, direction and shape of objects in the images. There are many sources that could cause distortions such as lens, earth curvature, topographic relief and the attitude of the aircraft that is used to carry the camera. These distortions occur differently, collectively and irregularly in the entire image. Image rectification is an essential image pre-processing step to eliminate or at least reduce the effect of distortions. In this paper, a non-parametric approach with piecewise linear transformation is investigated in rectifying distorted aerial images. The non-parametric approach requires a set of corresponding control points obtained from a reference image and a distorted image. The corresponding control points are then applied with piecewise linear transformation as geometric transformation. Piecewise linear transformation divides the image into regions by triangulation. Different linear transformations are employed separately to triangular regions instead of using a single transformation as the rectification model for the entire image. The result of rectification is evaluated using total root mean square error (RMSE). Experiments show that piecewise linear transformation could assist in improving the limitation of using global transformation to rectify images.