mI Ue EIRE ES ET TE, 
    
  
  
  
14. 
3.1 Edge Location and Pointing 
The task of pointing to edges of objects on photography is an 
important operation of any photogrammetric organization. Tasks such 
as the detailed plotting of linear features, and the tracing of edges 
of stockpiles of materials which must be located accurately for 
volume computations are just two examples of edge measurements made 
by operations in photogrammetry. Indeed, any task requiring the 
delineation of the borders of objects involves the task of edge 
pointing. However, pointing to edges of targets may even be under- 
taken by operators if asymmetric identifiable ground points are used 
for ground control points, in the absence of pre-marked points. Such 
asymmetric points may be chosen as the changes in density patterns é 0$ 
occurring at corners of fences, due to differences in cultivation or 
grazing on each side of the fence or a particular point on the edge 
of a road. The magnitude of systematic errors resulting from edge 
pointing observations will be discussed in the following sections. 
3.2 Research Data Available on Edge Location 
Research data quoted in this section are those presented by 
Thompson (1972), Martucci (1972) and Welch and Halliday (1973). 
Thompson performed pointing observations on extended asymmetric 
edges in order to derive a measure of image quality which would be 
directly related to the reliability of the location of the subjective 
edge. In doing so Thompson revealed a number of factors which are 
relevant to this paper:- 
(i) there were large systematic errors in the subjective location 
of the edges; 
(ii) the magnitude of these errors varied from one observer to v a 
another, and even between different sets of observations by the one 
observer; 
(iii) The parameter used by Thompson for ranking image quality 
was the standard deviation of the means derived by all observers. He 
deduced that the best measure of image quality was related to the : 
degree of difficulty of arriving at a subjective location of the edge. 
This method of ranking the image quality correlated well with a 
property of the edge referred to by Thompson as "Decutance". 
To amplify on the above points, Thompson found that an average 
value of the systematic errors of all 4 observers was 30 secs of arc 
(150urad or 4um at 10X magnification). The systematic errors of 
different sets of observations by the same observer varied 10 to 20 
secs of arc (50 to 100urad or 1.5 to 3um at 10X magnification). 
Normally the observer pointed to the lighter on less dense side of the 
edge; also the errors increased with contrast of the edges. Even 
though the errors varied each observer had his own personal bias, 
since he tended to point to approximately the same region of the edge 
on all occasions. It is significant that the overall spread or dis-