Research Of Control System Of Bait Feeding Device Based On Dynamic Image Acquisition

With the development of economy,the people's demand of aquatic products is increasing.It is an irresistible trend of aquaculture to use reasonable feeding methods and improve the utilization rate of fodder.The traditional ways of feeding have some disadvantages,such as the lack of feedback of the information about fish and the unscientific feeding which only depends on the subjective judgment of human.The control system of bait feeding device based on dynamic image acquisition in this paper can make great use of fodder and improve the breeding benefit.The emphasis of this paper is mainly about image processing technology and embedded operation system.The main content can be summarized as the following aspects:(1)The images of floating baits are obtained by CMOS camera to solve problem that the failure of bait feeding device to get the information about fish.If the water range is too wide,the camera is not able to get the full information and make images clear at the same time.The rotatable installment is designed to realize dynamic image collection.(2)Using image stitching,we can get the complete information about all floating baits.The mosaic image of floating baits is processed by erosion and expansion techniques.The geometrical characteristics of floating bait can be gained to establish a recognition model based on LS-SVM to count the floating baits in images.(3)It is difficult to use traditional control methods to realize the precise control of feeding.So the theory of predictive control is put forward for the characteristics of nonlinear,time-variant,multivariable coupling aquaculture feeding system.The predictive control model is based on LS-SVM,and the GA technique is used for optimization in the proposed controller.(4)The control system of bait feeding device is designed based on the dynamic image collection and the precise control of feeding.The hardware design mainly includes the structure of bait feeding device,the embedded microprocessor unit control circuit and etc.The software design mainly includes establishing the platform based on Linux system,processing images and sending information.The experiments of recognizing the floating baits in the images and casting the baits are carried to analysis of system's performance.The results of experiments indicate that the control system of bait feeding device based on dynamic image acquisition can monitor a wide range of waters,and the counting error of floating baits is under 5%,which shows that the control system can improve the feeding efficiency and enhance the investment benefit.