Figure 8.

Examples of applications of the morphological filtering function within the tracking process. (Left) Zoom of the acquired color image. The prediction (red dot) and the search area (black box) are highlighted. (Center) Search window after applying the bottom-hat filter. (Right) Binarized search window. (a) The TM fails since the target is partially outside the box, thus causing a reduction in the Normalized Cross-Correlation (NCC) score. (b) The TM fails since the target is partially hindered by a cloud, thus causing again a reduction in the NCC score. (c) The TM fails due to the sudden change in the background (from the sky to the mountain). The smaller extracted region (on the left) is automatically discarded by the algorithm as an outlier due to its limited size.