get_training_components — Return the initial or model components in a certain image.
get_training_components( : TrainingComponents : ComponentTrainingID, Components, Image, MarkOrientation : Row, Column, Angle, Score)
get_training_components returns all initial components (if
Components = 'initial_components') or all model
components (if Components = 'model_components')
in TrainingComponents in form of contour regions as well as
in numerical form. Alternatively, by directly passing the index of
an initial component, all found poses of that initial component
(i.e., the poses before solving the ambiguities in
train_model_components) are returned.
The pose of the returned components corresponds to their pose in the
model image (if Image = 'model_image' or
Image = 0) or in a training image (if
Image >= 1). In order to obtain the
components in the pose at which they were found in the ith
training image, Image must be set to i. Furthermore,
when dealing with rotationally symmetric components, one may wish to
mark the current orientation of the found component. This can be
achieved by setting MarkOrientation to 'true'. In
this case, the contour region of each component is complemented by an
arrow at its reference point pointing in the reference
direction. The reference direction of a component is based on the
orientation of the component in the model image and is represented
by an arrow that starts at the reference point and points to the
right in the horizontal direction.
In addition to the contour regions, the pose and the score of all
found components is returned in Row, Column,
Angle, and Score (see
find_shape_model). If Components was set to
'initial_components' or 'model_components', the
tuples TrainingComponents, Row, Column,
Angle, and Score always contain the same number of
elements as initial components or model components contained in
ComponentTrainingID, respectively. If one component was not
found in the image, an empty region is returned in the corresponding
element of TrainingComponents and the elements of the four
output control parameters are set to the value 0. In
contrast, if the index of an initial component is passed in
Components, these tuples contain as many elements as
matches of the corresponding initial component were found in the
image.
The operator get_training_components is particularly useful
in order to visualize the result of the training
ComponentTrainingID, which was performed with
train_model_components. With this, it is possible to
evaluate the suitability of the training images or to inspect the
influence of the parameters of
train_model_components. Sometimes it might be reasonable to
restart the training with train_model_components using
a different set of training images or after adjusting the
parameters.
TrainingComponents (output_object) region(-array) → object
Contour regions of the initial components or of the model components.
ComponentTrainingID (input_control) component_training → (handle)
Handle of the training result.
Components (input_control) string → (string / integer)
Type of returned components or index of an initial component.
Default value: 'model_components'
Suggested values: 'model_components', 'initial_components', 0, 1, 2, 3, 4, 5
Image (input_control) string → (string / integer)
Image for which the components are to be returned.
Default value: 'model_image'
Suggested values: 'model_image', 0, 1, 2, 3, 4, 5, 6, 7, 8
MarkOrientation (input_control) string → (string)
Mark the orientation of the components.
Default value: 'false'
List of values: 'false', 'true'
Row (output_control) point.y(-array) → (real)
Row coordinate of the found instances of all initial components or model components.
Column (output_control) point.x(-array) → (real)
Column coordinate of the found instances of all initial components or model components.
Angle (output_control) angle.rad(-array) → (real)
Rotation angle of the found instances of all components.
Score (output_control) real(-array) → (real)
Score of the found instances of all components.
* Get the model image.
read_image (ModelImage, 'model_image.tif')
* Define the regions for the initial components.
gen_rectangle2 (InitialComponentRegions, 212, 233, 0.62, 167, 29)
gen_rectangle2 (Rectangle2, 298, 363, 1.17, 162, 34)
gen_rectangle2 (Rectangle3, 63, 444, -0.26, 50, 27)
gen_rectangle2 (Rectangle4, 120, 473, 0, 33, 20)
concat_obj (InitialComponentRegions, Rectangle2, InitialComponentRegions)
concat_obj (InitialComponentRegions, Rectangle3, InitialComponentRegions)
concat_obj (InitialComponentRegions, Rectangle4, InitialComponentRegions)
* Get the training images.
gen_empty_obj (TrainingImages)
for i := 1 to 4 by 1
read_image (TrainingImage, 'training_image-'+i+'.tif')
concat_obj (TrainingImages, TrainingImage, TrainingImages)
endfor
* Extract the model components and train the relations.
train_model_components (ModelImage, InitialComponentRegions, \
TrainingImages, ModelComponents, 22, 60, 30, 0.6, \
0, 0, rad(60), 'speed', 'rigidity', 0.2, 0.4, \
ComponentTrainingID)
* Visualize the result of the training.
count_obj (InitialComponentRegions, NumInitComp)
count_obj (TrainingImages, NumTrainings)
for i := 1 to NumTrainings by 1
select_obj (TrainingImages, TrainingImage, i)
for j := 0 to NumInitComp-1 by 1
* Visualize the ambiguous poses of each initial component.
get_training_components (TrainingComponents, ComponentTrainingID, \
j, i, 'false', Row, Column, Angle, Score)
endfor
* Visualize the final poses of the initial components.
get_training_components (TrainingComponents, ComponentTrainingID, \
'initial_components', i, 'false', \
Row, Column, Angle, Score)
* Visualize the final poses of the model components.
get_training_components (TrainingComponents, ComponentTrainingID, \
'model_components', i, 'false', \
Row, Column, Angle, Score)
endfor
If the handle of the training result is valid, the operator
get_training_components returns the value 2 (H_MSG_TRUE). If
necessary an exception is raised.
Matching