|
|
|
Sensor Integration and Image Georeferencing in Support of Airborne Remote Sensing Applications
Dr. Naser El-Sheimy
Professor and Canada Research Chair
Department of Geomatics Engineering, The University of Calgary
Alberta, Canada
Email: naser@geomatics.ucalgary.ca
Dr. Sameh Nassar
Post-Doctor Fellow, Department of Geomatics Engineering, The University of Calgary,
Alberta, Canada
Email: snassar@ucalgary.ca
Abstract
Airborne remote sensing considerably extends the capabilities of satellite remote sensing in terms of resolution and operational planning. Where satellite remote sensing at its best achieves accuracies in the meter level, airborne remote sensing has the potential of achieving the decimeter level in positional accuracy. In addition, the usefulness of satellite remote sensing is often restricted by the images available for a certain area and the extent of the intervening cloud coverage, such limitations do not exist in airborne remote sensing. It is therefore possible to optimize the required results by adapting the operational conditions to the task at hand. Research in airborne remote sensing and mobile mapping at The University of Calgary (U of C) has, as one of its goals, the development of a precise positioning and attitude determination system that can be used with a variety of airborne sensors and ultimately eliminate the need for ground control. In this paper, accuracy requirements for such a system are discussed, different sensor configurations are described and the results of the U of C prototype development are analyzed. Airborne remote sensing for mobile mapping applications can be subdivided into three major groups: those where precise positioning is the major requirement, as for instance photogrammetric applications; those where both position and attitude are required with high accuracy, as for instance pushbroom imaging applications; and those where accurate real-time position and attitude estimation for real-time applications are also needed, as for instance forest fire fighting. Hence, sensor configurations for different applications will be discussed in the paper and the first results of real-time airborne tests will be briefly reviewed.
1. Introduction
Since the early seventies, collecting remotely sensed data by satellite remote sensing has been widely used. At that time, the satellite remote sensing resolution was in the order of 80 m-100 m (Stadelmann, 1990). However, even a considerable improvement of accuracy occurred afterwards, the current satellite remote sensing best resolution is still at the meter level. Due to the high demand of a better accuracy in many mapping applications, a major trend was directed towards the development of digital airborne remote sensing systems. In addition, satellite images are usually not localized, i.e. images are not available for all specific areas of interest. Of course, this is not the case in airborne missions since they are planned to cover the required local areas. Finally, due to the large distance between the satellites and the Earth’s surface, satellite image visibility is always limited by the cloud coverage and all images need a long processing for image enhancement and quality improvement.
Airborne sensing sensors, which are usually called imaging sensors, can be a frame-based (analog) aerial camera, Charge Coupled Device (CCD) digital camera, thermal camera, Light Detection and Ranging (LIDAR) laser scanner, pushbroom scanner or a Synthetic Aperture Radar (SAR). Despite the imaging sensor used, all airborne imaging measurements are performed in what is called measurement frame, which is defined in general by the body frame of the moving aircraft. On the other hand, all image information needed for mapping is required in what is called a mapping frame. The transformation process from the measurement frame to the mapping frame is called georeferencing (El-Sheimy, 2000). Therefore, image georeferencing is simply the process of determining the Exterior Orientation Parameters (EOPs) of the imaging sensor for each exposure. However, this also requires the determination of the imaging sensor Interior Orientation Parameters (IOPs) through calibration.
For many years, georeferencing in airborne photogrammetric remote sensing applications was performed using image block adjustment as well as sufficient and well-distributed sets of Ground Control Points (GCPs) through the well-known process of Aerial Triangulation (AT). This procedure is known as indirect georeferencing. In this case, to obtain the EOPs and for error propagation control, each block of images must have established GCPs. However, the establishment of the required GCPs constructs the major part of the AT process budget. Moreover, for many remote areas such as deserts, coastlines, forests, steep mountains and snow-covered grounds, the establishment of GCPs is extremely limited and, if performed, adds a large amount of additional costs.
In addition, the image evaluation in the AT is very time consuming and requires highly skilled personnel. This issue does not affect only the budget in commercial applications but also it is an important factor in emergency and rescue applications that rely on airborne remote sensing such as forest fire fighting for example. In theses cases, a very quick response is required and hence, image processing is needed in real-time, which is obviously not available through the traditional AT process (Ip, 2004). Finally, in some airborne mapping applications that involve a single strip mapping such as pipelines, power lines, highways, coastline etc., establishing GCPs will be a major problem and also impractical in principle. In some of the other airborne remote sensing applications, such as pushbroom linear scanners (which have very weak geometry since each scan line has a different set of orientation parameters), the EOPs are required for each scan line. Hence, a block adjustment process in this case will require a very large number of GCPs.
|
|
|