Helsinki University of Technology
Department of Surveying
The reliability of vehicle interpretation is studied. Interpretation has been made from aerial photographs of 11 towns of the southern Finland. The scale of the colour photographs was 1:12500 and they were digitized to the pixel size of 21 Ám. A GIS-software was used to the interpretation of the digital photographs and to the manipulation of the measured data.
Aim of the interpretation was to observe all vehicles in traffic on the photographed areas, recognize types of these vehicles and calculate two quantities: a distance that a vehicle has moved between two consecutive photographs (time lap about 5,8 s) and the speed related to the distance in question. Thus the components of the interpretation's reliability to be examined are completeness of the observations, recognizance of vehicle types and accuracy of the calculated quantities.
The determination of the reliability of the vehicle interpretation is not unambiguous. The numerous sources of errors which weaken the reliability of the interpretation are examined as separate factors and theoretical estimates of their magnitudes are presented. Furthermore, the calculating methods for the estimates of the components of the reliability in question are introduced. Finally some possibilities to still improve the reliability of the interpretation are proposed.
For the performed interpretation the completeness of the vehicle observations was 98 %. Other vehicle types than vans were recognized over 90 % probability. The accuracy of the calculated distances was on average better than 2,9 m and the accuracy of the speeds 1,8 km/h respectively.
Keywords: aerial photograph, geographic information system, traffic, vehicle.