How does detection work?
Based on user provided parameters (rod length, thickness) macro generates a series of template images:
For each angle orientation, macro calculates normalized cross-correlation image between template and a frame of original movie. Here is an example movie frame:
and example template orientation:
It will produce following cross-correlation map, enhancing positions where template can be found:
In the next step macro looks for the pixel with maximum cross-correlation value. It should be above the threshold specified by user (x times SD of cross-correlation image intensity). Once it found it, it removes line shaped area around it (marking rod detection) and looks for the next maximum:
In addition, to prevent multiple detections, macro adds line image with unique ID to the "detection map" image:
This image is checked on every new detection. If current pixel is occupied by previous detection, based on the maximum correlation value one of the detection is rejected and the other one is kept.
Choosing "Show rejected detections in overlay?" option shows these rejected detections in red and blue (while "true" detections are in green).
