Research On Key Techniques For Swimbladder-bearing Fish Acoustic Monitoring

With the grand blueprint of national ocean power and blue granary strategy put forward,people’s demand for high-quality fish protein is increasing day by day.How to rationally utilize and manage fishery resources becomes the production demand of fish farmers and fishermen.On our country’s more than 18,000 kilometers of coastline,there are thousands of marine pastures and breeding areas all over the country,plus our country’s rich rivers and lakes,how to efficiently use the fishery acoustic data measured by fish detection sonar and obtain reliable Fishery resource information,assessment and monitoring of fishery resources are the hotspots and difficulties that people pay close attention to.According to the development status of fish detection technology at home and abroad and the mainstream configuration of fish detection equipment,this research focuses on its development trend,focusing on the classic KRM model method,and carries out the scattering model modeling and experiments of individual fish and fish schools.The high-resolution imaging principle and experiment of Bayesian compressed sensing beamforming technology,the track tracking of single fish and the realization of abnormal behavior classification algorithm have been carried out in the main aspects of research work.Through theoretical derivation,computer simulation,experiment and data processing and other research methods,the main research contents of the thesis are as follows:Firstly,the acoustic scattering model of fish with swim bladder and fish school is studied.The target strength of fish is the key parameter to transform the integral value of fish school echo into fish statistics,and the model estimation method is to combine the target strength of fish with fish body and the shape of the swim bladder is correlated and can provide a valid estimate of fish populations.In order to solve the problem of large amount of calculation and low calculation efficiency of the traditional Kirchhoff ray model,this paper proposes a modified KRM model,which eliminates the interference by non-coherently adding the scattering lengths of the fish body and swim bladder term effectively reduces the amount of calculation.The experiment proves that the target strength result of the swimbladder fish body obtained by this model is more consistent with the measured results,and it is more suitable for TS estimation of wild fish schools.This method has higher stability and effectively solves the shortcomings of the KRM model.Secondly,the high-resolution imaging algorithm of fish detection sonar is studied.Aiming at the underwater high-resolution detection requirements of fish detection sonar,based on Bayesian compressed sensing,a multiple sparse representation Bayesian model is proposed to realize the sparse representation of underwater complex targets.The Gaussian prior model of multiple sparse representations is given through more rigorous derivation,and multiple sparse transformations that are independent of each other are introduced simultaneously by analytical methods to preserve the conjugate matching relationship of multi-level Bayesian,which can be solved efficiently by methods such as maximum expectation.Theoretically,the signal reconstruction performance of the algorithm is quantitatively verified by computer simulation,and then the experiment is carried out in the laboratory pool,and the data is obtained in the sea test to verify the feasibility of the algorithm.Finally,the trajectory tracking of fish based on Gaussian particle filter and the abnormal behavior classification algorithm of fish based on SVDD are studied.Facing the complex underwater environment,the tracking complexity of fish targets is high,and the computational complexity is large.This chapter proposes a fish tracking technology based on Gaussian Particle Filter(GPF).Through the tracking of a single fish,extract its movement characteristics,and use this as a basis to judge the abnormal behavior of fish,and then study the classification of abnormal behavior of fish,and propose a classification method for abnormal behavior of fish based on SVDD,and based on the combination of the training set and the test set,the validity of the classification results is verified.Experimental verification shows that the fish abnormal behavior classification algorithm based on SVDD has a higher classification accuracy than the traditional SVM algorithm,which verifies the effectiveness of the method.The above explains the echo counting principle of fish detection sonar from the acoustic scattering model of swimbladder fish,improves the imaging algorithm of fish detection sonar from multiple sparse representation of Bayesian compressive sensing beamforming,and from the trajectory tracking of fish based on Gaussian particle filter And the abnormal behavior classification of fish based on SVDD improves the application function of fish detection sonar,and provides certain methods and technical support for the assessment and management of fish with swim bladders by using fish detection sonar in three aspects,and verifies it through experiments effectiveness and feasibility.It proves that the new method proposed in this paper can be extended to modern marine ranch construction and fish farming management.