EXPERIMENTAL STUDY OF DIRECT GEOREFERENCING SYSTEM (DGS) 
FOR LARGE-SCALE MAPPING 
Kikuo Tachibana , Tadashi Sasagawa and Thoru Yotsumata 
GIS Institute, PASCO Corporation 
1-1-2 Higashiyama Meguro-ku Tokyo JAPAN 
Kikuo_tachibana@pasco.co.jp, Tadashi_sasagawa@pasco.co.jp, Thoru_yotsumata@pasco.co.jp 
KEY WORDS: adjustment, * camera,* * GPS,* * IMU, triangulation 
ABSTRACT: 
Presently for the orthophoto projects and small-scale mapping (e.g., 1:2,500) the aerial triangulation is no longer essential while 
Direct Georeferencing System (DGS) is available as a complementary technique. Though, it is possible to achieve sufficient accuracy 
by doing simultaneous block adjustment with minimum number of the ground control points, in some cases there are insufficient 
inaccuracies of a direct georeferencing in large scale mapping of 1/1000* 4/500. It is very cost effective by reducing the expenses 
for the ground surveying. Further, quality control and accuracy improvement of the photo block are promising with the DGS. In this 
paper, the investigation of the experimental accuracy of simultaneous bundle adjustment for large-scale mapping is reported. Since 
2002, the PASCO Corporation has introduced two sets of APPLANIX POS/AV for its two units of aerial cameras (RC30) and we 
have conducted several accuracy investigations. According to the specification of product, the accuracy levels of direct 
georeferencing system after post-processing are 5 - 30 cm for the position; 0.005 deg for roll and pitch; and 0.008 deg for heading 
angle in case of high-end system. As the Geoid model is also required to transform it from ellipsoidal height to orthometric height 
(above mean sea level), and to always guarantee the accuracy level of large scale mapping, it was considered that the simultaneous 
block adjustment is essential for 1:1000 — 1:500 scale mapping. Since bundle adjustment improves block geometry, it seems that the 
middle class DGS is accurate enough. In this paper we present the results of the accuracy evaluation using APPLANIX POS AV/DG 
310. The number and configuration of ground control points were also considered. 
1. INTRODUCTION 
Both 510 and 310 have same GPS as position sensor. Table 2 
shows the influence of each rotation angles 0.013 deg and 0.035 
deg on the ground if 150mm lens is used. The influence of the 
heading values were located near the fiducial mark. 
Accuracy evaluation of direct georeferencing system has been 
carried out recently utilizing high-end DGS APPLANIX's POS 
AV 510 and mapping project specification including QC 
method has been established. To achieve accuracy level of large 
scale mapping e.g. 1/500, simultaneous block adjustment is most 
effective methodology. Since bundle adjustment improves block Photo scale Photo scale 
geometry, it seems that the middle class of DGS can be accurate 1/4,000 1/8,000 
enough. The main purpose of this practical study is to evaluate *SAGD:6008 9 4 sGD.1200* 9 > 
accuracy level of simultaneous bundle adjustment using middle 
class DGS APPLANIX's POS AV 310. 
  
  
Roll & Pitch (m) 0.14 0:27 
  
Heading (m) 0.37 0.73 
  
  
  
  
Table 1 shows the specifications of POS AV 310 and 510. The 
specifications of the IMUs are different for 310 and 510. In this 
study, we evaluated 310 
  
Table 2. Influence of rotation elements 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
1 For typical mission profiles. 
2 Attitude will drift at this rate up to a maximum error 
defined by absolute accuracy. 
Table 1. POS AV Specifications 
860 
| POS AV/DG 310 | POS AV/DG 510 Actually, other errors, e.g., GPS positioning error, photo 
Absolute accuracy (Post processed) measurement error are propagated. Required map accuracy of 
Position '(m) 0.05° 9.30 0.05° 8.30 1/500 scale map in Japan is regulated at the standard deviation 
Velocity (m/sec) 0.0075 0.005 of 0.25m for both planimetric and height. In this case, exterior 
Roll & Pitch (deg) 0.013 0.005 orientation parameters provided by DGS is sometimes not 
Heading (deg) 0.035 0.008 accurate enough and simultaneous adjustment can be needed to 
Relative eU improve block accuracy. 
E Sem 043 um 2. PRACTICAL STUDY 
0.5 
2.1 Outline of practical study 
In this study, accuracy evaluation was done by comparing the 
coordinates of check points observed in the image models using 
different approaches. Result of 4GCPs block, 1 GCP and also 
In 
nc 
th 
us 
st 
e 
T, 
P! 
SI 
Pi