Non-invasive identification of individuals of Atlantic salmon (tent data)

Non-invasive identification of individuals of Atlantic salmon (tent data). 328 fish were sampled. 30 fish were tagged for next data collection. totally 4 data collection were done. The fish were photographed in tent out of water. The fish were photographed on the green background for simple fish localization.

Each image was processed separately. First the green background was detected based on the defined value of the green in HSV color space. The backgound was used as the area of fish appearance. In the secodn step the fish was segmented on the background using the specific color of the fish in HSV color space. The fish was detected as the largest object. Some morphological operations were used to correct fish shape. The object was rotated to the horizontal position. The fish tail was removed by the algorithm because it was detected on some images and not detected on the rest. Therefore the algorithm detects the begining of back tail as the most narrow place at back part of the fish. The lenght of the fish and the height of the fish was used for the localization of the ROI. The ROI is shown on the image bellow. The ROI was selected to cover the fish body with most visible patter.

Uploaded files - original and processd files for all data collections.