Rational Polynomial Coefficient-based Generic Sensor Model
Technology description
Geometric sensor models are used in image processing to model the relationship between image space and object space and to transform image coordinates to ground coordinates in a map projection. Rational Polynomial Coefficient (RPC) is a sensor model commonly used by the remote sensing industry to determine the ground coordinates of pixels in high resolution satellite imagery.
The present invention - RPC-based Generic Sensor Model - is a further extension of the RPC sensor model. It extends the application scope of the RPC model from cameras with narrow field of view (such as high resolution satellite cameras) to cameras or sensors with any field of view (such as optical satellite sensor/camera, airborne cameras, and radar sensors) through a RPC refinement. However, the existing industry standard method for RPC refinement, i.e. the Bias Compensation method, is only feasible to high resolution satellite cameras with a narrow field of view.
This invention relates to a computer implemented method of adjusting original coefficients of a RPC model to obtain a new set of RPC coefficients by using one or more ground control point(s) (GCPs). The new set of RPC coefficients can better fit the geometry of the sensor, improving its geometric accuracy.
The basic idea of the algorithm is: (1) restore the light ray for each GCP based on the RPC coefficients provided by an image provider to obtain the camera's exterior orientation parameters (EOPs); (2) adjust these EOPs using the GCPs by least squares solution; and (3) generate a new set of RPC coefficients based on the adjusted camera EOPs. In other words, this algorithm is a direct method to refine the RPC coefficients, because it directly modifies the coefficients to better fit the sensor.
Advantages
The novel features of the RPC-based Generic Sensor Model are:
- Extended application scope and stable results, i.e. accurate ground coordinates can be obtained regardless of the magnitude of the field of view and the angular and position errors of the sensor.
- Geometric accuracy reaching to sub-pixel level under all conditions.
- Feasible to a broad variety of remote sensing sensors including airborne and satellite sensors.
Stage of development
A prototype is developed and beta tests have been successful in the laboratory.
Intellectual property
- US Patent Pending
- Canadian Patent Pending
This technology is available for licensing.
Contact information
Office of Research Services, Industry-Government Services
University of New Brunswick
Phone: (506) 453-4674
partner@unb.ca
