Previously used solutions - HelmchenLabSoftware/contrast-marker-video-tracker GitHub Wiki

Previously, manual identification of the markings for every time frame were necessary. Software like OCIA and ClickJoint allow user to manually select the markings frame by frame, and save the resulting coordinates to a file. Attempts to automatize the procedure were based on frame-to-frame correlation: the user would specify the location and size of the markings in the initial frame, and the algorithm would try to locate each of the markings in the next frame by shifting the contour of the original markings around within a small distance and correlating the insides with the original. The algorithms break down for the following reasons

Not fault tolerant: If the region is not visible or poorly visible in a given frame, the algorithm will fail to detect it in all the following frames as well. Also, the marking could have high correlation with other markings, spots on the fur or any of the connectors on the wheel. Again, if the algorithm detects it wrong once, it can not continue.