Index

General

  1. Matching threshold
  2. A Matching threshold allows you to manipulate the recognition results. The similarity score is a comparative score which shows how likely the test image contains the object. A high similarity score suggests that test image contains the object. On the other hand, a low similarity score implies that test image displays noise.

    The matching threshold is selected experiment way because objects are very different. Default is 40000 and for concrete case it should be determined. Matching threshold can be in range 0 to 500 000.

  3. Enable mask enhancement
  4. Enables mask enhancement. The given mask can be extended in order to eliminate noise and small halls inside the object.

  5. Recognition speed
  6. Use tracking
  7. This option used during object recognition with test images If the option is marked the tracking of the object in a sequence of test images is enabled. Otherwise, if the option is unmarked the tracking is disabled and object recognition uses the other algorithm for comparison.

    This options should be marked only with a sequence of test images where the neighbouring images differ only slightly. Also, constant light conditions and constant background usually improve tracking results.

  8. Transform type
  9. If auto, selects best possible transform, if particular transform is selected but current condition do not allow to perform it simple transform is selected. Transform complexity: Similarity->Affine->Perspective.

  10. Learning mode

Generalization

  1. Use generalization
  2. Enables generalization.

  3. Auto threshold
  4. The value "0" was used for generalization threshold, i.e. the most suitable threshold will be calculated and used in the method.

  5. Generalization threshold
  6. The generalization threshold. If a recommended value "0" was used, the most suitable threshold will be calculated and used in the method.

    The amount of redundant information and the size of the model after generalization may be controlled through a threshold provided to the method. If this value is "0", the most suitable threshold for this model will be calculated and used in generalization process. If value is not zero, this threshold means maximum similarity score of two images that they both still would be kept in the model. Generalization process compares all the images in learnt model with each other and removes those images which are "too similar" to some others. This "too similar" concept means that their similarity score is above a certain threshold.

Colors

  1. Active shape color
  2. Active shape color and model name, score in recognition mode.

  3. Inactive shape color
  4. Inactive shape color and matched object in recognition mode.