Efficient individual identification of zebrafish using Hue/Saturation/Value color model

Automated fish species recognition is widely investigated in research but it is not explored for the individuals with the same fish species. A new classifying method for zebrafish individuals that is based on statistical texture and Hue/Saturation/Value (HSV) color features are presented in this paper. Post image acquisition, pre-processing stages and features of sub-images are extracted, using statistical texture and HSV color space domain, and grouped into HSV and statistical sets of features. An artificial neural network (ANN) and K-Nearest Neighbors (KNN) are then used to identify the subjects under test. The impact of using statistical and HSV features on the prediction accuracy and average processing time is then assessed experimentally. An improved performance for the HSV over the statistical model is clearly demonstrated. The combination of HSV model and KNN classifier has also demonstrated a superior performance over the combination of HSV and ANN classifier in terms of the accuracy (KNN = 99.0%; ANN = 97.8%) and average processing time (KNN = 4.1 ms; ANN = 24.2 ms). Such promising findings encourage further testing of the HSV model towards developing a highly-efficient and fully-automated identification system for small species individual like zebrafish. (C) 2018 National Institute of Oceanography and Fisheries.