Automatic registration of SAR and SPOT imagery based on multiple feature extraction and matching

• Published in: IGARSS 2000. IEEE 2000 International Geoscience and Remote Sensing Symposium. Taking the Pulse of the Planet: The Role of Remote Sensing in Managing the Environment. Proceedings (Cat. No.00CH37120) Vol. 7; pp. 2896 - 2898 vol.7
• Main Authors: Dare, P.M., Dowman, I.J.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 2000
• Subjects: Backscatter; Distortion measurement; Educational institutions; Electromagnetic modeling; Electromagnetic spectrum; Feature extraction; Image registration; Master-slave; Microwave radiometry; Polynomials
• ISBN: 0780363590; 9780780363595
• DOI: 10.1109/IGARSS.2000.860282

Many different models have been developed in the past to automatically register SAR and optical images. The vast majority of these models rely on feature based matching, due to the very different backscattering properties of the terrain in the optical and microwave regions of the electromagnetic spectrum. Even so, the difficulties associated with extracting similar features from radiometrically very different images have always hindered this approach to automatic registration. The model proposed in this paper uses feature based matching, but rather than relying on just one method of feature extraction, many different feature extraction algorithms are employed. This methodology ensures there is a large set of features extracted from each image to be matched. Consequently the chances of locating pairs of correctly matched points, which can be used in either a polynomial or a photogrammetric rectification model, are greatly increased. Application of the proposed algorithm to pairs of both small and large images showed that a substantial number of tie points could be accurately located in each pair of images. More importantly, the approach to feature based registration using multiple feature extraction techniques clearly improved the quality and quantity of the tie points compared to traditional feature based registration techniques which rely on only one feature extraction algorithm.