Practical methods for geometric distortion correction of aerial hyperspectral imagery

To utilize an aerial hyperspectral image acquired from a push broom type image sensor, the image geometric distortion must be first removed. This article reports on two methods for distortion correction. One method was a simplified sensor augmentation approach that used the sensor roll information during image acquisition. The other was an image processing approach that used automatically extracted distortion information. The sensor augmentation approach showed a correction error (RMSE – root mean square error) of 6.5 pixels when using the roll information for distortion correction. It is expected that further integration of an attitude sensor into the imaging system would produce even better results. The image processing approach used natural linear features and edge detection techniques achieving 1.5 pixels RMSE. For practical consideration, this method is especially useful for agricultural applications, as square fields and straight roads are common in rural areas.