OCess 
d Intelligence 
51S, pages 1- 
lage segmen- 
ate and map 
RS- Congress, 
emantic mo- 
erial images. 
ditors, Auto- 
m Aerial and 
Basel. 
cialist system 
or, 11th Eu- 
ce, ECAI 94, 
of evidence. 
of aerial im- 
and Remote 
CONTROLLING AERIAL PHOTOGRAPHY FROM TOPGRAPHIC MAPPING USING A MULTI-SENSOR 
BUNDLE ADJUSTMENT 
Michael Reading, James Carswell, Phil Kern, Intergraph Corporation, USA 
Commission III, Working Group 1 
KEY WORDS: Photogrammetry, Triangulation, SPOT, Aerial, Mapping, Accuracy 
ABSTRACT: 
Establishing ground control points for a photogrammetric topographic mapping project represents a significant cost, 
especially in remote or inaccessible areas. The goal then is to find methods to reduce this cost by minimizing the amount 
of ground control needed to achieve the desired level of accuracy. We evaluated an approach utilizing a simultaneous 
bundle adjustment which oriented a block of aerial frame photographs to a geometrically corrected SPOT stereo model. 
This approach was found to satisfy U.S. National Map Accuracy requirements for horizontal and vertical positioning with 
only 5 ground control points when compared to a conventional aerial triangulation of the photographic block. But these 
results are contingent on a number of factors, including the distribution of the control throughout the SPOT stereo model, 
the distribution of the tie points between the SPOT stereo model and the aerial photographs, the convergence angle 
between the SPOT images, and the amount of overlap between the aerial photographs and the SPOT images. 
1. INTRODUCTION 
À significant cost in any photogrammetric topographic 
mapping project is establishing the surveyed field control 
to scale and level the stereo models into the ground 
coordinate system. This can account for as much as 50% 
of the entire project's cost. The actual cost of establishing 
the control depends on the remoteness, harshness, and 
accessibility of the area being mapped. In order to 
properly orient a model, there needs to be at least two 
horizontal and four or more vertical control points in the 
model. Thus an economical approach to any project is to 
examine ways to reduce the number of control points 
needed to achieve the desired level of accuracy. The 
conventional approach to reducing this cost is to establish 
a sparse network of surveyed ground control and then 
densify the control through an aerial triangulation of the 
photographic block. In a typical project, this requires a 
minimum of eight horizontal and four vertical control 
points placed around the perimeter of the block, with 
additional horizontal control placed every 5 models and 
vertical control placed every 3 to 4 models. This network 
of control can be reduced further by incorporating Global 
Positioning System (GPS) observations into the 
adjustment to provide accurate estimates for the exposure 
stations' positions and orientation within the ground 
coordinate system. 
This paper explores the feasibility of minimizing the 
ground control by orienting the block of aerial 
675 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
photography to geometrically corrected SPOT imagery in 
a simultaneous multi-sensor bundle adjustment. In this 
adjustment different orbital models are used to describe 
the position and orientation of the imagery within the 
object space. The images are organized in the bundle by 
sensor type and orbital event. For frame photography 
each orbital event is the moment of exposure, and the 
orbital model is the collinearity equations. SPOT imagery 
is collected using a linear array of detectors which scan 
across the terrain in the direction of flight. In a SPOT 
panchromatic image there are 6000 scan lines, and each 
scan line's position and orientation differs based on the 
oribatal path of the platform. The orbital model for the 
SPOT imagery accounts for the time dependent nature of 
the sensor through a modified version of the collinearity 
equations. Each continuous path of SPOT images is 
treated as a unique orbital event with its own set of twelve 
parameters which describe the sensor position, velocity, 
attitude, and attitude bias at an initial epoch. By solving 
for these parameters, and knowing the time mark for each 
scan line, the state vector for each scan line in the SPOT 
image can be determined. 
Besides demonstrating that this approach can achieve the 
compilation accuracy necessary to satisfy United States 
National Map Accuracy Standards, we also intended to 
examine the various factors which affect the accuracy of 
the results, such as control point accuracy, number of 
control points, distribution of control points, the 
convergence angle between the SPOT images, and the