RIASSUNTO
Physical tagging for animals has been a key requirement to discriminate their individuality. However, in small swimming animals like zebrafish, physical tagging is considered a painful, costly and a time-consuming process. In this paper, a computer-vision based method for identifying and extracting of a new feature in zebrafish (Danio rerio) is proposed as an alternative to the physical tagging. The proposed method depends on the extraction of a blue stripes pattern exists around the central left or right side of the fish. The region of interest is initially identified using various image preprocessing steps including segmentation of the fish object, center finding and cropping. The default RGB colour of the identified area is transformed into a lab colour space (also called CIELAB space), which provides an enhanced pixel-level presentation for the colour of the area of interest. Next, a Kmeans clustering with 2 main clusters of colours is used to extract the pattern of blue stripes of the object under study. The obtained results from this pilot study with a relatively small dataset showed that the patterns of blue stripes vary from one object to another. Thus, they are considered as a promising fingerprint for zebrafish.