Research On Fishing Ground Forecast Of Neon Flying Squid Based On Surface Temperature Feature Extraction

The neon flying squid(Ommastrephes bartrami)is an economically important cephalopod with abundant resources and short life cycle,and it is also an important fishing object in my country’s pelagic squid fishery.Because the biological characteristics of squid are highly susceptible to changes in sea surface temperature,too high or too low temperature will be detrimental to its growth.Researchers usually use the average value of sea surface temperature(SST)to study its effect on the growth of squid,but the monthly average value of SST cannot reflect the dynamic characteristics of SST diurnal variation well.With the continuous advancement of remote sensing technology,especially the application of hyperspectral remote sensing and high spatial resolution remote sensing technology,the data accuracy of ocean remote sensing retrieval has been significantly improved.The application of this technology enables us to obtain more detailed and rich marine environmental information,and further provides a wide range of application space for the research of marine fisheries and the analysis of fish conditions.The features of marine remote sensing data are extracted by deep learning method,which can be extracted layer by layer from continuous convolutional layers,making the data gradually become more abstract and rich.In this paper,the Pacific squid was taken as the research object,and the sea surface temperature three-level inversion product image was used for feature extraction to explore the relationship between the sea surface temperature and the squid,and combined with the chlorophyll concentration(Chl a)to build a GAM model for prediction.The research results are as follows:(1)According to the biological measurement data of Northwest Pacific squid from July to October 2016,the carcass length,body weight and gonad maturity of squid were analyzed,and the power index equation was used to fit the female and male squids monthly The relationship between carcass length and body mass;the Deep Convolutional Embedded Clustering(DCEC)model was used to extract the distribution pattern of sea surface temperature based on the three-level inversion image data of MODIS-Aqua and MODIS-Terra sea surface temperature,which is used to reflect the characteristics and laws of the temporal and spatial distribution of SST corresponding to the survey sites in each month;on this basis,the influence of the temporal and spatial distribution of SST in different months on the growth and development of female and male squid is analyzed.The results show that: during July-October 2016,the growth and development of female and male squid in the Northwest Pacific Ocean were not completely synchronized,the growth rate of females was faster,and the development rate of male gonads was faster;four types of SST image features were extracted based on the DCEC model,which can better reflect the temporal and spatial distribution patterns of SST corresponding to the survey sites in each month,and reflect the influence of El Ni(?)o,Kuroshio and Oyashio on sea surface temperature;from July to August,the average body weight,carcass length and gonad maturity of squid The temperature is low,the difference between male and female individuals is not significant,and the relationship with the spatial and temporal distribution pattern of SST is not significant;from September to October,when the SST changes more uniformly but the overall temperature is lower,the average body weight,carcass length and gonad maturity of squid When the SST changes uniformly but the overall temperature rises,the average body weight,carcass length and the number of sexually mature individuals of squid increase significantly,and the growth rate of female fish is faster than that of male fish,but the maturity rate of male fish is faster than that of female fish.Fast;when the cold and warm water met and formed a relatively obvious boundary,the carcass length and body weight of the female and male cuttlefish reached the maximum,but the gonad development was delayed.Based on the deep learning DCEC model,this study explored the influence of SST spatio-temporal distribution pattern on the growth and development of female and male squid in different months,providing a reference for further understanding of the biological characteristics of squid in the Northwest Pacific.(2)According to the fishery data and environmental data of Pacific squid from January to December 2015,a fishery forecast model for squid was established based on surface temperature feature extraction.First,use MATLAB to crop the SST three-level inversion product image according to the survey station of the fishing area.The cropped image is normalized,fused,screened,repaired,etc.,and finally the SST sub-image is obtained.Then the SST subgraph is input into the model,and after three layers of convolution,it becomes a one-dimensional vector,and features are extracted through the fully connected layer,and the extracted results are received by the decoder and the clustering layer.The decoder is used to calculate the image reconstruction error and output the reconstructed image,and the clustering layer uses the output of the encoder for clustering.Determine the optimal number of clusters by the DBI value,and label each SST submap.The clustering feature values of each image were obtained,and the clustering feature values were grouped by month,and the mode of the feature values of each group of atlases was extracted for each survey point,so as to obtain the SST feature values of each survey station in each month.The SST monthly characteristic value and the monthly average chlorophyll concentration are used to match the catch per unit of effort(Catch per unit of effort,CPUE)in time and space respectively,and a GAM model based on surface temperature feature extraction is constructed,that is,an improved GAM;using SST Based on the monthly mean value and the monthly mean value of chlorophyll concentration and CPUE data,a modular GAM model based on the monthly mean value of SST,that is,the basic GAM,is constructed.It can be seen from the results that the comprehensive factor explanation rate of the GAM model constructed based on the surface temperature feature extraction is higher than that of the GAM model constructed based on the SST monthly average,and the impact on the CPUE of the Pacific squid is more significant;Compared with the model based on SST monthly average,the AIC value of the GAM model constructed by feature extraction was reduced by 9.72%,and the MSE value was reduced by 24.12%,which combined with the chlorophyll concentration got a better fitting result,but the sea surface temperature was still The main reason affecting the growth of softfish proves the effectiveness of extracting sea surface temperature characteristic values based on the DCEC model,and provides a reference for the follow-up research on longline fishing forecast of Pacific softtail.