Heikkinen, Jussi 
  
5 OBJECT RECONSTRUCTION 
What are the advantages of this kind of image block formation? The construction of camera configuration is very well 
controlled and with sensible parameterization the solution of the problem can be stabilized. The result is numerous 
camera poses with known position and orientation in the same co-ordinate system. Now we can have multiple 
observations to one object point as in the traditional camera configuration the point could be seen only on two images. 
Also, with this geometry of the image block we can reach nearly equal positioning accuracy in all directions. 
This kind of imaging configuration is applicable for tasks where the object is surrounding the camera station. Like in 
caves and other inner space measurements. Also measuring tasks where no exterior co-ordinates system is present are 
favourable for the presented method. 
The greàt number of images is partly an advantage and party a disadvantage. The huge amount of images needs a lot of 
processing but on the other hand numerous observations can improve the accuracy of 3D measurements. Because the 
imaging is performed in a controlled way we can take advantage of this (Heikkinen, 1995) (Heikkinen, 1996). The 
knowledge of the turning direction of the camera lets us search the correspondent point on the next image at the defined 
direction on image. Also the fact that subsequent images have only few degrees of difference on orientation angles 
gives us an opportunity to use area based signal matching, which would not normally work with divergent images and 
highly three-dimensional objects. 
The object modelling by using parametric models has been proved to be beneficial when we have a lot of observations 
(Heikkinen, 1994). Parametric models include the geometric properties of the object in a very compact form. That is 
why it is very important that we choose the right parametric model for the reconstruction. With the parametric object 
modelling from images we understand the procedure first introduced by Mikhail and Mulawa as Linear feature based 
photogrammetry (Mulawa, 1989) (Mulawa-Mikhail, 1988). In this procedure the estimation of object parameter is based 
directly on image observations. Traditionally the object points are measured first and after that the object parameters are 
estimated on these resolved 3D points. With linear features based methods the edge measurements are used directly to 
estimate parameter values of the object. 
6 FORESTRY APPLICATION 
Our pilot application area belongs to the field of forestry. In Finland there is a long tradition of forest inventation. From 
the first forest inventation a systematic sampling method has been used. The sampling method called “relascope 
sampling", also known as Bitterlich sampling, has been the main method in field measurements. The forest volume 
estimation method is based on theory of the stochastic processes. Even though the satellite image and laser scanning 
measurements are substituting field measurements nowadays, there is still need and place for them. New airborne 
methods give good estimates of tree heights, but the accuracy of tree stem measurement is still quite poor. 
The observations needed in point sampling on field measurements are the relative distances of tree stems from the 
measuring centre. This means that we have to measure the distance and the diameter of the stem included in the sample. 
Those trees are included in the sample whose viewing angle is larger or equal to the selected angle a as depicted in 
NN 
QS 
Figure 6 Relascope sampling method 
Currently the measurements are done by using measuring tape. Our goal is to apply instead this presented method to 
measure the stem distances and diameters from the plot centre. A parametric cylinder or cone is going to be used for 
estimating tree stem volumes from multiple images. 
  
364 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 
(09 MM 70980 e d