Papers
Change detection and motion detection is theoretically possible using image fusion—imagery of a common area from multiple sensor platforms, however perspective differences represent a significant obstacle both to human and machine correlation, particularly in the case of side oblique imagery. Orthorectification provides a common perspective and thus a practical first step for follow-on correlation algorithms.
Today’s modern image processing software that removes pointing angle and platform anomalies through the photogrammetric orthorectification process offers some utility that if migrated to hardware could provide near real-time on platform or on sensor capability. The orthorectification process, however, is so computation intensive and time consuming that real time operation is generally not available.
This paper describes a low-cost means of performing real-time orthorectification, a brief overview of the orthorectification process and how it relates to targeting and location measurement. Also included in the presentation is a dynamic demonstration of two commercial software packages being used to extract geocoordinate information from a high resolution digital image.
Even though film is often thought to be a dinosaur in modern airborne acquisition systems it is still unsurpassed in capability to provide the most resolution and detail to the analyst who is chartered to extract the highest level of intelligence possible. The inability to quickly provide information from acquired film imagery has been one reason stated by field commanders as to their preference for "all digital" camera systems. Digitally scanning the film and adding modern digital processing to scanned images would enhance the "data mining" of archived imagery and could also maintain the exceptional quality of image data from today's film systems. New software developments, if applied, could also shorten the time line between the acquisition and the user.
Airborne sensors provide a valuable source of imagery by virtue of the location of the deployed platform. However, obtaining precise and rapid information from such imagery has been difficult due to the enormous amount of post processing required by existing software and hardware to extract the geo-location for targets of interest. Commercially available software packages running on high-level computer systems generally take several minutes and typically hours of human intervention to provide credible target location.
This paper describes a method of locating every pixel and every target of interest in an image to within a few meters as the images are acquired and some of the additional benefits that would result from such processing on low cost PC computers and in real or near real time.
Targeting from airborne imagery based solely on GPS and IMU data is subject to potentially large errors due to the effects of uneven terrain—particularly with low flying platforms. A proposed system would substantially reduce such errors and provide additional benefits such as target tracking and instant pointing to specified lat-long.