Study On The Extraction Of Offshore Aquaculture Area Based On Multi-feature Coupled With Object Classification

Offshore fishing facilities is the main form of offshore aquaculture in China,and also an important disaster-bearing body for marine disasters.Therefore,it is important to carry out research on area extraction of aquaculture areas based on remote sensing technology to quickly grasp the structure of aquaculture patterns in coastal areas,scientifically plan and manage coastal aquaculture areas,and avoid coastal natural disasters.However,due to the wide distribution of farming areas,it is difficult to carry out a large-scale field survey.Therefore,the use of satellite remote sensing images to extract farming facilities has become an important means to obtain information on the characteristics of farming areas in the fields of offshore fisheries and ecological environment at present.Sansha Bay is a semi-enclosed harbor in the northeastern part of Fujian Province,and there are large areas of nets and rafts in the nearshore of the bay,with rich marine resources,and the aquaculture industry in the bay is the main economic pillar of the local industry.The aquaculture industry in the bay is the main economic pillar of the local economy.How to quickly obtain information on the distribution of aquaculture areas in the bay has become a pressing problem for the local government to solve.By carrying out research based on remote sensing technology,it can help local governments to quickly grasp the distribution location of aquaculture areas for rational planning and management of coastal aquaculture facilities,solve the problem of difficulty in obtaining data for fishery underwriting and claims system,and improve the efficiency of risk assessment and management of fishery insurance.The article combines the multi-feature technique and Rule-based Object-oriented Image Classification(RBOIC)technique,and proposes a model for aquaculture area information extraction in complex environments.-The paper proposes a multi-feature and Rule-based Object-oriented Image Classification(MROIC)method for aquaculture area information extraction in complex environments.The experiment firstly takes small areas of water as the study area,uses the domestic high-resolution high-fraction-2(GF-2)satellite remote sensing image as the data source,and adopts the object-oriented classification method based on the multi-feature technique and rule set to extract the feature data of net box and raft aquaculture area in this area,and also verifies the extraction accuracy of the model.The basic idea of the method model is: firstly,the feature function is determined according to the wave spectrum characteristics and spatial characteristics of the aquaculture area,then the feature index bands are constructed by using the feature function,and then the constructed feature index bands are filtered by using the baroclinic distance,and the filtered bands constitute the band feature set.Finally,the method of principal component analysis is used to reduce the dimensionality of the feature set,and the reduced dimensional bands are combined with MROIC,so as to realize the accurate extraction of feature information of the breeding area.The results show that the overall extraction accuracy of the method in this paper reaches 90.4%,and the Kappa coefficient reaches 0.80,which can effectively extract the net box and raft farming areas in Sansha Bay Bay.In order to test the extraction effect of large area farming facilities,Sentinel-2remote sensing image was selected as the extraction area for the whole sea area in Sansha Bay,and a total of 3820.50 ha of raft farming area and 1083.69 ha of cage farming area were extracted.The accuracy verification experiment selected three densely farmed areas in the bay located in the southwest of Xiapu County as the extraction accuracy verification sea area,and extracted the area of net box and raft farming area in the sea area.The experimental results show that the average overall extraction accuracy of the three areas is92.78% and the average Kappa coefficient is 0.84,which is similar to the previous experimental results,indicating that MROIC can extract the aquaculture areas in Sentinel-2images with high accuracy.In summary,the MROIC model constructed by coupling multi-feature technique and object-oriented classification technique in this study is able to extract feature information of offshore aquaculture area using GF-2 and Sentinel-2 remote sensing images as data sources.The model can also ensure a good extraction effect in the complex environmental situation of Sansha Bay.Therefore,the technique in this paper can provide to the local government with the characteristic data of the coastal farming area for monitoring and management of the offshore and beach farming area.At the same time,the method can provide data sources for the underwriting system of the local fishery insurance industry to help them improve and refine the existing management system and technical means to meet the market demand and the development trend of the insurance business.