Next: Correspondence Analysis
Up: Line Detection
Previous: Synthetic data
Real data
Different kinds of real objects can be used to evaluate the quality of an algorithm. But it must be proved, that the object used can be correctly and totally described. In the actual case the objects are also limited to those that can be detected thus produce straight lines when mapped onto the image plane. A ball for example will not be detected by the line detector. The objects used can be categorized into
- Cube
- A metallic cube with dimensions (70x100x60,h-w-d,mm).
- Cone
- A wooden cone with dimensions (79×76.5,r-h,mm)
- Plane
- A plastic CD-case with dimensions (124×142,w-h,mm)
The reference values are manually extracted, the worst case is a displacement of +/- 1px. The values of the line detection result from four different images of the same objects. All object lines are investigated. The result is shown in Table 4.3.
|
It can be seen, that the differences between the real and the computed start and end points have remarkable increased in comparison to the line detection using synthetic data. Also the differences between the start and end points of the same object between different images varies with an approximate maximum of about 10% of the line length. Thus the points should not be taken as an important attitude when it comes to the correspondence analysis. In comparison, the difference of the attitude and the attitude itself varies only about 0.1. As a result, the attitude can be seen as a strong property, as well as the gradient direction of the line.
Next: Correspondence Analysis
Up: Line Detection