Intelligent Baiting Decision System For Captive Bass Based On Multi-source Data Fusion

China is the world’s first aquaculture country,with the increasing demand for aquatic products,aquaculture scale gradually expanded,since the reform and opening up,the aquaculture industry has achieved rapid development,relative to the current scale of domestic aquaculture,as well as foreign fishery aquaculture equipment,China’s fisheries modernization development is relatively backward,mechanization,low degree of intelligence,especially in the baiting,the common domestic The common baiting method in China is manual baiting and mechanical baiting,which is difficult to ensure the uniformity of baiting,and this method is time-consuming and laborious,and it is easy to underfeed or overfeed,which affects the growth and development of fish.With the continuous development of fishery machinery,baiting machines are widely used to gradually replace manual baiting.However,the simple baiting machines generally only meet the function of timing and quantification and cannot increase the feed utilization rate.The study of fish feeding behavior can guide bait feeding,but the fish culture environment is complex and uncontrollable,and different modes of culture and changes in climate can affect fish feeding behavior,so there are still some difficulties in intelligent feeding based on fish feeding behavior.To address the above situation,this paper designs an intelligent baiting decision system to realize intelligent feeding of outdoor captive bass by studying the aquaculture water quality parameters and feeding behavior of surface largemouth bass(Micropterus salmoides).The main research work and conclusions of this paper are as follows.(1)Largemouth bass detection and size and quality estimation.An underwater video acquisition system was set up to collect underwater video of largemouth bass under the "zero discharge" captive breeding mode.For the freshwater fish aquaculture environment with uneven illumination,high noise,blurred and turbid shooting quality,in order to achieve the detection of underwater image bass,the image enhancement algorithm is used to enhance the underwater image using computer vision technology,comparing HE,CLCHE and MSRCR image enhancement algorithms,and comparing the parameters after image enhancement,MSRCR algorithm is selected for image enhancement.The SSD target detection network is used for underwater bass detection,and the network is improved.The improved SSD algorithm is compared with other image detection algorithms,and the accuracy of the improved SSD algorithm is significantly higher than several other models,and the improved model can have better results in the underwater environment.The experimental results show that the method can achieve 92.85% accuracy for the network trained using the captured dataset.Using the calibrated data for estimation of body length,the fish size is predicted according to the current state of research.(2)Construction of intelligent baiting decision system for largemouth bass.The water quality parameters collection system was built to collect the water quality parameters in real time,analyze the factors related to the appetite level of largemouth bass,select the fuzzy controller as the baiting decision controller,analyze the factors in the fuzzy controller,design the structure of the fuzzy controller,determine the water temperature,dissolved oxygen and fish size as the input function of the fuzzy controller,and the baiting rate,baiting frequency and baiting correction coefficient as the output function of the fuzzy controller.The output function of the controller is determined.To establish fuzzy control rules,the three-dimensional coordinates of fuzzy control were established to reflect the relationship between each variable based on the reference literature and the experience of breeders.And establish the theoretical domain and affiliation function of the input and output functions.By combining the growth pattern of largemouth bass,the feeding behavior of the fish was analyzed,and the appetite level of largemouth bass was divided into "strong","medium","weak" and "no" by analyzing the feeding time and feeding behavior of each round."and none.The deep learning model Shuffle Net V2 was selected to classify the appetite classes,and the accuracy of the training model reached 97.20%.The model was used to identify the appetite classes of fish and guide the bait feeding.(3)Design of intelligent baiting decision system.Can baiting decision system design and construction,the baiting machine is improved and its control system is changed to STM32 main controller baiting system,STM32 control system mainly includes core controller,water temperature sensor,dissolved oxygen sensor,Node MCU-8266 module,relay module and AC-DC step-down power supply module,LCD capacitive screen,etc.According to the demand design controller and wireless communication module and water quality sensor between the serial interface circuit,and design each independent module circuit,and at the same time according to the circuit design to complete the corresponding software development,to achieve automatic water quality data collection,transmission function.Graphic interactive interface design,image interactive interface including LCD touch screen graphic interactive interface design and UI graphic interactive interface design,where the LCD touch screen graphic interactive interface includes: water quality parameters display module,quantitative baiting module,intelligent baiting module and button baiting module;UI graphic interactive interface includes: basic information display module,video display module,feeding prediction module,feeding control module,weight prediction module and feeding module.module,weight prediction module and feeding record module.(4)Intelligent baiting decision system test.The intelligent baiting decision system was tested under the "zero-emission" captive breeding mode,and the baiting decision and method were specified and compared with the manual test.The test results show that the error between the feeding amount of intelligent baiting decision system and manual feeding amount is controlled within 5%,which is within the acceptable range.The intelligent baiting decision system established in this paper can better classify the appetite level of largemouth bass,which can reduce the waste of feed and save the cost of breeding,thus replacing the feeding decision of breeding personnel and providing a reference for the intelligent feeding of outdoor intensive breeding.