6.4 Hardware Control 
If the interpreter calls functions for image processing or 
sensor manipulation, these commands are sent to the 
controller module (see Fig. 4) and will be executed in the 
controller processes. During run-time the controller 
module consists of different concurrent processes which 
principally enable sensor control in parallel (time 
sharing). This means that it is possible to position a 
sensor (e.g. rotary table) concurrently to a second sensor 
(e.g. RSC) while other independent actions (e.g. 
database access) can be performed. Significant time 
saving during measurement process can be obtained. 
Each sensor group (cameras, table, lamps) are 
processed in a seperate UNIX process. 
Using standardized UNIX and TCP/IP control 
mechanisms it is possible to install the complete software 
package either on one machine or, if necessary, 
implement the controller module on a second processor 
in order to increase computing power. 
7. APPLICATIONS 
Fig. 6 illustrates the measurement of a circular hole for a 
loud speaker in a dash board which is made of black 
plastic. In this case there are two RSC images without 
rotation of the turntable. In both images, a part of the 
  
      
    
   
    
   
     
   
   
   
   
    
    
    
   
   
    
    
  
   
  
   
  
hole contour is extracted. Since it is an incomplete circle, 
an open contour with user-defined start and end points is 
detected. lllumination is by direct spot lighting. 
Fig. 7 shows an example of cylinder measurement 
applied to the inspection of a cooling tube made of black 
rubber. In this case either diffuse illumination or the 
intergrated light table could be used. Two images with 
different cameras and a rotation of 90° have been taken. 
in each case two edges of a cooling tube part have been 
extracted for determination of the cylinder in 3D space. 
Fig. 8 presents the measurement of two straight lines at 
the edge of a car door made of grey sheet metal. A 
combination of diffuse and direct illumination has been 
used. A part of the greylevel contour of the edge has 
been extracted in each image. The corner point of this 
feature can be calculated as a line intersection in space. 
The line element algorithm is sometimes not able to 
detect non-identical object lines if only two images have 
been taken. In this case a minimum number of three 
images is recommended in order to avoid errors in line 
calculation. 
  
Fig. 6. Measurement of circular loudspeaker hole 
left: Countour extracted from image of top RSC 
right: Contour extracted from image of center RSC