Image georegistration methods: A framework for application guidelines

摘要

With the rapid growth of sensor platforms for imagery collection, from micro-unmanned aerial systems (UAS) to smart phones, an ability to geo-register image data is a fundamental need for many downstream applications. Approaches to georegistration for sensor imagery have deep roots in photogrammetry, and more recently with the integration of computer vision techniques. Georegistration solutions are increasingly sought for inexpensive and non-metric quality sensors and/or those that may lack the metadata needed to support rigorous coordinate transfer with error estimation. This indicates a range of solution quality, with situational awareness at one end, and rigorous accuracy at the other. There are a variety of correspondence and transformation models from which to select, with tradeoffs among simplicity, accuracy, and error estimation. The continually expanding vernacular of terms and methods can lead to confusion of application among the broader community of users. A sorting of representative terminology, processes, and techniques, is proposed as a framework. The goal is to motivate discussion for application guidelines.