3. WuCAPS^ DESCRIPTION 
According to the principle mentioned above, the authors 
developed a software package suitable for the combined 
adjustment of photogrammetric and non- 
photogrammetric observations—WuCAPS G/ >s. It starts 
from collinearity equations, combines the stochastic 
model based on statistic theory with strict mathematical 
model of combined bundle adjustment, and then a series 
of special algorithms has been developed in WuCAPS GPS . 
The strictness of theory and the flexibility of practical use 
have indicated sufficiently that WuCAPS G/ > 5 will be 
widely used both in scientific research and in productive 
project. At present, WuCAPS Gra is using as means of the 
control point determination for aerial photogrammetry in 
China. WuCAPS G /> 5 consists of a group of separate 
program modules related to each other, which are written 
in FORTAN. It nearly contains all kinds of basic 
computation tasks on analytical aerotriangulation, and 
each task can be executed independently with result 
output, which is the prerequisite for WuCAPS G/ > 5 to be 
used in teaching and training. WuCAPS Gra is divided into 
two parts, i.e. a main system of combined bundle block 
adjustment for productive purpose and a subsystem of 
quality analysis to serve mainly as the purpose of some 
special scientific research. Figure 2 pictures the overview 
of the design for combined bundle block adjustment 
procedure in WuCAPS G /> 5 . 
WuCAPSgps includes the following basic functions: 
□ Block adjustment by strips 
This is a block adjustment program with a 
nonlinear correcting model based on quadratic 
conformal polynomial expression. It can meet 
the demands for topographic mapping at small 
or medium scale and severe the purpose of low 
precision point determination. 
□ Bundle block adjustment with additional 
parameters 
This is a conventional bundle adjustment 
program for pure photogrammetric observations. 
Three sets of self-calibration models, such as 
Bauer model with 3 parameters, Ebner model 
with 12 parameters and Brown model with 18 
parameters [Wang, 1990], are alternative in the 
different photogrammetric projects. 
□ Combined bundle block adjustment with 
navigation data and/or geodetic observations 
This is the core function of WuCAPS G/ > 5 
software package. If the original geodetic 
observations and/or the other auxiliary data as 
well as some relative constrains in object space 
are introduced into the mathematical rriodels of 
the conventional photogrammetric adjustment, 
the integral adjustment method is called 
combined adjustment in photogrammetry, which 
improves the location accuracy of 
photogrammetric points or reduces the number 
of the essential ground control points. In 
WuCAPS Gra the usable additional measurements 
include 3D GPS-determined positions of camera 
centers, INS-determined attitudes of camera, and 
the geodetic observations such as leveling, 
spatial distances, horizontal angles and vertical 
angles, and so on. 
□ Compensation of systematic errors 
It is well known that the accuracy of the 
conventional bundle adjustment has closed 
connection with the magnitude of the systematic 
errors in image observations. Therefore, during 
the combined bundle adjustment procedure, 
WuCAPS G />s can correct the systematic errors of 
photo observations by affine transformation (six 
parameters) using four fiducial marks, correcting 
formula of radial distortion according to the 
calibration report and self-calibration techniques, 
and compensate that of 3D GPS-determined 
coordinates of camera perspective centers in a 
linear correcting model. In addition, according 
to the coordinate residuals of the ground control 
points between from photogrammetric 
adjustment and from field survey, the adjusted 
coordinates of photogrammetric points can 
register with the coordinate system defined by 
the ground control points by least squares 
collocation after the conventional bundle 
adjustment. 
□ Automatic detection and rejection of the 
blunders 
Gross errors in observations are not avoidable in 
photogrammetry. An advanced adjustment 
program should automatically be able to detect 
and eliminate the blunders to obtain reliable 
computing results. For this purpose, several 
methods of blunder detection are used in 
WuCAPS g/ »5. The blunders in the fiducial mark 
observations are detected by data snooping 
approach [Forstner, 1985] during the inner 
orientation. The blunders in image observations 
are eliminated efficiently by varying weight 
iteration method [Wang, 1990] during the 
procedure of forming free strips and by Danish 
method [Kubik, 1982] in the combined bundle 
block adjustment, respectively. The gross errors 
including the ground control points and GPS- 
determined positions of the cameras are 
automatically detected by varying weight 
iteration method during the procedure of data 
preparation, too. 
□ Calculation for accuracy matrix of unknowns 
and reliability matrix of observations 
Accuracy and reliability are two very important 
factors to evaluate the quality of the adjusted 
results for photogrammetric block adjustment. 
Only results with high-accuracy and high- 
reliability are acceptable. Therefore, two 
3-5-5