Application Research Of Computer Vision Technology In The Swimming Crab Cultivation

Swimming crab is one of the important species for aquaculture in China. Swimming crab beds, in theponds, are a traditional form of farming.Its survival rate is so low. With the development of China’sindustrialization and urbanization, aquaculture area has been shrinking year by year, forcing the swimmingcrab culture model must transform from extensive to intensive.However, that change will increase the workload for the farmers and informationization is necessary.In recent years, computer vision technology has been widely used in aquaculture, such as aquatic productclassification, quality estimation and welfare monitoring. It has become one of the essential technologies indeveloping aquaculture. The research about applying computer vision technology in the swimming crabcultivation can realize industrial upgrading. That is of great significance for improving productionefficiency and quality.This paper tried to analyze the requirement for swimming crab farming work and the current status onapplying computer vision technology in other biological aquaculture. The following researches werecarried out.1. To apply computer vision techniques in the swimming crab farming work, this paper developed aeffective method for detection and tracking of swimming crab in the aquaculture environment. In term ofdetection, Statistical pattern recognition method was used in this paper. It created a classifier by learning avariety of swimming crab images offline, which could detect the swimming crab successfully in thecomplex farming environment. In the side of target tracking, this paper regarded tracking as a binaryclassification problem.This method could trace the swimming crab stably, by building a classifier withcompressed sensing feature and using the simple Bayesian classifier online learning strategy. Comparedwith the previous studies, the detection and tracking algorithm developed by this paper need not additionalequipment to change the background, which was more economical and practical. Meanwhile, the approachwas applied to other aquatic organisms for detecting and tracking.2. For the biomass estimation, the study measuring the quality of the swimming crab based oncomputer vision was carried out. The digital image was sent to a computer for image correction, imagesegmentation and gets the area of swimming crab. Data fitting between area and weight was using the leastsquares method. The best fitting way was quadratic polynomial, which’s average relative error was6.40%.That proved this method can meet the requirements of swimming crab’s weight estimation.3. For the welfare monitoring, the experiment was tried out, which measured the swimming crabactivity under different salinity. The result showed that the activity of the swimming crab under salt stresswas different with normal salinity, indicating that activity can be used as an indicator of the health ofswimming crab.