Quantitative evaluation of positional accuracy of images

After recording in a remote sensing image, it is important to evaluate, or validate, the positional quality of the recorded image for different sensors from different satellite platforms. The evaluation can be qualitative, in a visual or quantitative comparison, using error metrics, from matrix data, recorded images or vector reference data. The quantitative evaluation is important because it is independent of the user and numerically informs the discrepancies, in relation to a reference considered to be true.

Various error metrics, absolute, mean, quadratic mean, etc., can be reported considering local and global information. The validation can be done from references by point samples or procedures of new criteria for calculating accuracy based on vector features or tracks.

Thus, punctual data and tracks, with high precision and collected in field works, are compared with corresponding information digitized on registered remote sensing images. In this comparison, positional accuracy calculations are performed with vector and matrix approaches between the data collected and the digitized information.

These calculations are reported in local and global error / uncertainty reports that can help a user evaluate the quality of an image recorded within pre-set accuracy parameters.


 

Operations in the SPRING registration window

 

Sobre Registro