A research based on recognition algorithm of characteristics of body surface of infected fish

摘要

The contents of this research is how identify an infected fish through analysis of it's image to get from the monitoring aquaculture process by computer. By studying recognition algorithms about surface images of infected fish, we identifie automatically fish diseases by computer and no need of manual intervention it. The early warning from computer makes it possible to take measures to prevent and save the loss and damages in fishing industry intensive and high-density. In accordance with fish's own body color characteristics of a large contrast, we use Dual-threshold Difference-image Method to extract it's contour, then can make fish's image a structural segmentation according to the proportion of it's head,tail and body, and further extract their color and texture feature vector in order to compare the surface feature vectors with that normal fish have ownned, we can find some infected fishes or no, and even the type of fish diseases.