122 
spatial orientation of its own scanner head. Therefore, 
having measured mutual location and orientation 
parameters of camera and IMU sensor, provided their 
mutual immovability during flight and synchronization of 
camera and locator operation based on absolute GPS 
time, each aerial photo may be provided with quite 
sufficient set of exterior orientation parameters. 
With reference to principal limitations of 
stereotopography method reviewed in the Section 1, 
now these can be analyzed from another point of view, 
that is availability of synchronous laser and digital photo 
data, GPS referenced. 
It is obvious that photo and GPS data themselves is a 
complete set to proceed with classic 
stereophotogrammetric processing. What does 
appearance of DTM formed by laser location method 
principally change in this scheme? 
In any case, such DTM is another independent 
measurement in addition to stereophotogrammetric 
model, and, therefore, ever reliable criterion of 
correctness, even without sophisticated analysis 
methods. Not this only, however. 
Significant technologic success in improving accuracy 
of optical inertial systems operating per formula (IMU)+ 
(differential GPS) allows possible direct geopositioning 
of aerial photos based on direct use of inertial system 
data as exterior orientation parameters (Lithopoulos, 
1999), which in our case with true DTM available, 
makes it possible to completely refuse of ground 
controls points. However, such optimistic conclusion 
looks somewhat too early due to the reasons below. 
3. OFFERED SOLUTION 
The proposed digital technology of automatic 
production of orthophotomaps upon integral processing 
of laser, GPS and photo data is to a big degree an 
alternative to the classic stereophotogrammetry 
method. The following motivation was presented in 
technologic aspect: 
The major problem of technology cycle of classic 
stereotopography method is, a necessity of on-ground 
geodetic support, on the one hand, and, inevitable 
manual labor on stages of frames mutual orientation, 
DTM production, and correct connection of 
orthorectified photos on the other hand. 
Logically the first is a procedure of true DTM separation 
from full laser image of scene. As it was mentioned, 
implementation of such procedure is a sophisticated 
task. Solution is reached upon application of special 
topologic analysis algorithms classifying laser point per 
criteria "belong/not belong" to true ground. Such 
algorithms are built upon two obvious postulates: 
1) True ground point has minimal value of geodetic 
elevation in comparing with ones in its vicinity. 
2) Spectrum of spatial frequencies of true ground 
surface has no high frequencies. 
Specific realization of such algorithms is normally 
carried out by construction of mathematical surface 
limiting the bottom of the entire scope of points of given 
survey scene. Limitations per spectrum for the given 
surface are presented in limitation on values of its first 
and second differential. 
Such approach provides practically satisfactory results, 
however: 
1) Separation procedure is to a big extent based on 
heuristic principles and requires frequent operator's 
participation in selection optimal setup parameters for 
given landscape - size of scan cell, limitation values, 
and other. Moreover, in complex scenes different area 
fragments require essentially different setup that leads 
to necessity to use specialized interactive DTM 
synthesis software. The latest circumstance is clearly in 
a contradiction with imperative idea of full automation of 
technological cycle. 
2) In their mathematical contents such algorithms are 
close to procedures of high-frequency spatial filtration. 
As a result, with no special provisions made, actual 
relief fragments with abrupt elevation changes would be 
deleted from final DTM, or deformed. 
3) At last, all such currently developed algorithms are of 
significant work duration. Computer processing 
connected with true ground separation takes times 
longer than aerial survey data collection. 
Consider a stereopair made by two digital aerial photos 
provided with true spatial coordinates of principal 
points, and initial approximation of optical axis 
orientation angles. Presume availability of accurate data 
on camera exterior orientation parameters and 
photographic distortion. Then with good photo quality 
and presented texture of scene surface, with modern 
digital photogrammetric processing programs (like 
ERDAS Imagine + OrthoBase) automatic mutual 
orientation of frames making a stereopair is possible via 
correlation search of corresponding points. For this 
purpose algorithms are used, working on principle of 
parametric maximization of mutual correlation moment 
of two minor fragments of both images, while the 
parameters are unknown values of scale, mutual 
orientation and photometric brightness of these 
fragments. As it was mentioned procedure of 
stereomodel production for given stereopair is built on 
the same mathematical principles. 
In the perfect case the output shall be a scaled (due to 
availability of GPS coordinates of principal points) 
stereomodel correctly spatially oriented with only free 
angle of turning around survey basis vector. 
Implementing the described approach faces two 
general difficulties: