The Matching Assistant of HDevelop is a front-end to HALCON's powerful
matching methods:
- shape-based matching,
- correlation-based matching,
- descriptor-based matching, and
- deformable matching.
For information on these matching methods and how to
choose a suitable method for your application, please refer to the
Solution Guide II-B: Matching.
Using the Matching Assistant you can
- Configure and
test the matching process with
a few mouse clicks and
- Interactively optimize the parameters for your application regarding the
speed and recognition rate.
All you need is a single
model image and a set of
test images.
The Matching Assistant guides and assists you setting up and optimizing your
matching application.
Its individual elements are explained in the part
匹配助手参考.
In this manual, the following special terms are used:
- Matching
- Matching is the process of locating an object described by a
model in an image.
The main results of the matching process are the position and orientation of
the found object instance and its matching score.
- Alignment
- This method can be applied to
transform the position of the matched object corresponding to the
reference image.
Alignment is useful if the following image processing step is not
invariant against rotation or translation, like OCR or the variation model.
Note that by alignment the matched object is only rotated and
translated. More information about alignment using shape-based matching can be
found in the
Solution Guide II-B: Matching.
To remove perspective or lens distortions, for example, if the camera observes the
scene under an inclined angle, you must
rectify
the image first.
- Model
- The model is an internal representation of the object containing only
the information characterizing the object.
This model is created by the assistant from an example image of the object, the
model image,
which is provided by you.
The model is used when searching for the object in the
test images.
You can also provide the Matching Assistant with a
model, see section “The Menu File”.
- Model Image
- This is the image containing your example of the object to be searched
for. This image should be a characteristic image of the object.
The object should appear in its default position and orientation and not be
occluded.
- Reference Image
- If a reference image is selected, the position of the match in this
image is used as reference. This is necessary to perform
alignment.
If no reference image is chosen, the
model image
will be used as basis for
alignment.
- Model Region of Interest (ROI)
- This is the region in the model image which contains the object to
be trained. You can select this region via the menu item
ROI.
- Rectification
- This method can be applied to the search image to transform it such that
the found model and the model in the
reference image
appear as similar as possible.
Rectfying an image is useful for all further processes that rely on
fixed ROIs, like measuring and OCR.
- Test Image
- You can test the performance of the matching process by providing
test images.
These images should be representative images from your matching
application. The object should appear in all allowed variations
of its position, orientation, occlusion, and illumination.