The main objective of this research topic is to develop a robust computer vision system that could detect and track a large number of blob-like structures in a video stream for a longer period of time.
Several tasks connected with this objective are:
- segmentation of the image to determine regions for blobs
- separating correct blobs from accidental noise
- connecting blobs with existing detections from the past frames, registering new detections and removing ones that are not present any more
- real-time performance of the entire system
Sample video of the input for the algorithm: