Study On The Method Of Frying The Fry Of Turbot (Paralichthys Olivaceus) Based On Computer Vision

In recent years, turbot aquaculture industry develops rapidly and has become the pillar aquaculture industry in north China, which has brought the huge economic and social value.In the process of turbot breeding, we need to count turbot fry accurately for the scientific control of turbot fry growth, development and sales. So the research of counting method of turbot fry has very important practical significance for turbot aquaculture industry. Traditional fish counting methods mostly are manual counting methods, manual counting methods always are slow, error prone and easy to damage the fry. With the rapid development of computer vision and image processing technology, fry counting method based on computer vision has become an important means of study,because of its high efficiency, high accuracy, simple needed equipment and other advantages, The fish counting method research based on computer vision also has the vital significance to enhance the level of information modernization in the process of aquaculture.In this paper, the turbot fry counting methods are studied by the computer vision and image processing technology. We capture the video of turbot fish fry flowing through the channel by the video capture equipment which is developed by laboratory, and then extract each frame from the video image, we make gray level images processing, filtering processing and binarization processing to the image, extracting the fish object, then we study the single frame images and continuous frame images counting method respectively. for single frame image we use the methods of area, connected domain method and the curve evolution method to count, the level set method is used to model curve evolution, the initial evolution curve is the contour curve of fish image. The experimental results show that the curve evolution method has good counting effect. Continuous frame counting model is from moving target tracking algorithm based on feature, we use the feature of centroid position to track and count fry target, the experimental results show that the counting effect is very good in the absence of adhesion target image; When there is adhesion, counting result is inaccurate, therefore the curve evolution method is adopted to improve the method of counting, the experimental results show that the counting effect is superior to direct counting after the curve evolution.