GEOSCAN Search Results: Fastlink


TitleAutomated Control Point Acquisition in Radar-Optical Image Registration
AuthorGuindon, B
SourceCanadian Journal of Remote Sensing 11, 1, 1985 p. 103-112,
Alt SeriesEarth Sciences Sector, Contribution Series 20040516
Alt SeriesRESORS 1049481
PublisherInforma UK Limited
Mediapaper; on-line; digital
File formatpdf
Subjectsgeophysics; remote sensing
Released2014 08 01
AbstractA study has been carried out to assess the usefulness of automated image matching algorithms in the registration of radar-optical image pairs. Registration experiments were conducted with an airborne image data-set which included an optical reference scene (simulated TM bands 1, 2, 3 and 4) and 3 multitemporal CVV SAR scenes acquired over an agricultural area in western Canada.

The matching procedure investigated has included those algorithms which have proven useful in the past for the registration of multitemporal LANDSAT imager, namely the extraction and matching of gradient image segments employing normalized correlation as a measure of goodness of match. Adaptive filtering has been applied to the SAR data prior to edge extraction in order to reduce the effects of speckle noise.

Automated control point acquisition was found to provide registration accuracies which approached those achievable by manual means (i.e. average errors of 2.8 and 2.3 pixels were observed for automated and manual acquisition, respectively). The registration noise introduced by the matching process was estimated to be 1.7 pixels.

The accuracy of a control point acquired by automated means was found to be related to its observed peak correlation. Application of a correlation threshold of 0.65 resulted in the elimination of mismatches while still providing acceptable success rates (63 to 79%).


Date modified: