The Automated Landmark Navigation of the Polar Meteorological Satellite

The problem of automated landmark navigation for the polar meteorological satellite is addressed. Automated landmark navigation can correct the systematic image navigation errors due to orbit, attitude and alignment matrix disturbance. Traditional automated landmark navigation methods need selecting the base image from long time series satellite imagery; otherwise it can result in registration displacements. Previous methods have shown that having daytime and nighttime (including early morning) base images for each season can minimize the registration error. The processing of one year data would thus require a minimum of eight base images for each location, and it limits the method from being widely applied. There are great needs for an accurate, easily implemented navigation system capable of automated landmark navigation when long time series satellite imagery data are absent.First, cloud detection methods insure that contamination from cloudy pixels is minimized. Second, the base image is composed of content features and structural features. The content features are constructed from the energy distribution of the current satellite image's ocean, land, rivers and so on. By defining the landmark feature point, the landmark's structural features are constructed from the global template. The landmark content and structural features are combined together to form the full base image. Third, the maximum cross correlation (MCC) method produces displacement vectors, which are translated into satellite attitude corrections to be added to the orbital image navigation corrections. Each resulting displacement vector has a correlation coefficient, which quantifies how well a pattern is matched. Displacements with correlations lower than the 95% confidence value is the elimination of error matching. The image navigation accuracies are also closely related to the landmark spatial distribution. The image navigation corrections are more accurate when the landmarks are average distributed through the whole satellite image. Finally, FY-1D satellite data are used to assess the performance. The testing results shows that this method is automatic and is successful to rectify the image navigation errors due to attitude disturbance and the rectified errors are within one pixel. This method does not use long time cache files needed by early methods and thus extend the applicability. The proposed method has been applied in Chinese next generation polar meteorological satellite FY-3 and will be developed further in the next generation geostationary meteorological satellite FY-4.