The Establishment And Application Of Fish Intelligent Transportation Decision-making System

There are many kinds of aquatic products in China.Fish is considered as the best source of animal protein in the 21st century because of its rich nutrition and low fat content.However,fish are often under intense stress during transportation.After long-distance transportation,deterioration of water quality and death of live fish(loss rate of long-distance transportation>10%)often occur,which become a major problem in the live fish transportation industry.Because the transport process of live fish is a high-dimensional and complex system,it will be subjected to multiple stresses such as deterioration of water quality,bacterial infection,anoxia,noise,vibration and so on.Therefore,the selection of water source before transportation,the accurate evaluation of water quality,the monitoring of environmental indicators during transportation and the early warning of death after transportation are all very important for the survival of fish before and after transportation.In order to solve the above problems,this paper intends to creatively construct a new generation of fish intelligent logistics decision-making system by integrating biotechnology and artificial intelligence algorithm.The construction of intelligent decision-making system can not only effectively solve the problem of long-distance transportation of live fish,but also effectively overcome the shortcomings of traditional transportation mode such as high pollution,low efficiency and poor safety,so as to ensure that local specialty fish in China are no longer restricted by regional circulation,which has important practical guiding significance for improving the survival rate of live fish transportation.The main results are shown as follows:1.The effects of different transportation time(0,2,4,6 and 8 hours)on the blood physiological and biochemical indexes of yellow catfish and hybrid yellow catfish"Huangyou-1" during simulated transportation were studied.The lactic acid content,lysozyme content and total superoxide dismutase(SOD)activity of two kinds of fish were compared and analyzed at different transport times.The results showed that the SOD activity of yellow catfish and hybrid yellow catfish "Huangyou-1" increased gradually with the increase of transportation time,and the SOD value of yellow catfish reached the highest value 8 hours after transportation.The overall trend of SOD value of hybrid yellow catfish"Huangyou-1" was consistent with that of yellow catfish,but its activity was lower than that of yellow catfish.During transportation,the lactic acid level of yellow catfish and hybrid yellow catfish "Huangyou-1" increased first and then decreased with the transportation time,and reached the peak level in 6 hours.Compared with common yellow catfish,the lactic acid content of hybrid yellow catfish "Huangyou-1" was lower overall.After transportation,the lysozyme activity of yellow catfish and hybrid yellow catfish "Huangyou-1" changed significantly,showing a downward trend as a whole.The lysozyme activity of two species of fish fluctuated greatly during 2-6 hours of transportation,and showed a lower level after 8 hours.2.BP(Back propagation)and RBF(Radial Basis Function)artificial neural network assessment methods are applied to the precise regulation of water quality in the process of live fish transportation,and a comprehensive comparison is made.The results showed that when the stocking density was 6.26 g/L and the temperature was 20℃,the water quality of Channel Snakehead could not meet the requirement of fishery water when it was transported naturally(without MS-222)for more than 2 hours.When the stocking density was 28.08 g/L and the temperature was 20℃,the water quality of anaesthetic transportation(adding anaesthetic MS-222)of hybrid yellow catfish "Huangyou-1" was better than that of natural transportation(without anaesthetic MS-222)(the water quality grade of C1-C5 group was lower than that of CO group in all time periods).It can be seen that using BP or RBF neural network to evaluate the water quality of live fish transportation can break through the limitations and simplicity of traditional methods of water quality evaluation.3.Combining with the Internet of things technology,a set of BP neural network algorithm based on optimization is designed for non-destructive testing of fish quality.Specifically,passive RFID is used to collect fish traits,filter method is used to select features,and three evolutionary algorithms(GA,PSO and MAA)are used to optimize BP network to predict fish quality.The morphological data and three UCI data of three kinds of fish(river sand pond snapper,yellow catfish and wattle catfish)were used to test the results.After optimization by evolutionary algorithm,compared with the original BP model,the output error of the optimized BP neural network is smaller and the convergence speed is faster.The results show that the proposed method can predict biomass quickly and accurately according to the morphological characteristics of fish.The prediction error(RMSE)is 1.76 and the determination coefficient is 0.974.4.Aiming at the problem that it is difficult to detect endangered fish in real time and with high accuracy,an early warning system for endangered fish based on Transfer Learning and Convolution Neural Network is proposed.The initial weight of convolution neural network is optimized by transfer learning method,so that the model has certain generalization ability at the beginning of training.The total error of network before optimization is 1.75,and after optimization is 0.75.To further improve the performance of the model,two groups of convolutional neural networks(VGG-net and ZFnet)are compared as feature extractors.The results show that ZFnet network(AP=0.902)is slightly better than VGnet network(0.898),and both of them are significantly better than the common HOG+SVM algorithm(AP=0.259).