Research On Classification Method Of Submarine Substrate Type Based On Sonar Data

The scope of human activities is expanding with the development of modern science and technology,as far as the moon and even Mars,there are traces of human technology.However,people's ability to explore the ocean close at hand may not be as good as the exploration of the moon,the characteristics of seabed environment limit the application of many technologies.But the Abundant resources in the ocean is the most important and the most available treasure-house of human civilization under the current development of science and technology.The seabed sediment information is an important reference for understanding the ocean,and the analysis of seabed sediment information is an important research direction in marine research.The traditional contact-type method for seabed sediment detection is expensive and inefficient.The research of oceanographers and hydroacoustic scholars has proved that the use of acoustic remote sensing technology can invert the seabed sediment information by analysing reflected sound waves.This article is supported by the project,this paper mainly carried out the research work on four seabed sediment inversion related contents,including the principle of classification of seabed sediment types,sonar data processing and imaging,backscatter intensity feature extraction,and sediment classification algorithm model.Besides,the research revolved with theoretical analysis,simulation experiments,and implementation of sediment map inversion project with measured data in the offshore pilot zone.The seabed bottom quality map inversion process used in this study provides theoretical basis and technical support for our country's future seabed bottom quality exploration work.The main research contents are as follows:Firstly,this paper investigated the research status seabed inversion technology and the development of quality and seabed classification software.Besides,this paper explained the basic principle of the Multi-beam sonar and side-scan sonar system,and studied the detection method of seabed sediment,the particle size division method of sediment type,the division rule of mixed sediment and the symbol system of sediment type.Secondly,the common formats of sonar data are introduced and parse sonar data obtained from actual experiment,In research,data accuracy is a main influencing factors in the realization of research content.Therefore,in order to improve the accuracy of the data,the factors that introduce errors in the sonar data during the measurement process are analyzed to realize the intensity reduction of the sonar data and the true value calibration of the backscatter intensity,thereby improving the authenticity of the sonar data and laying a good foundation for the inversion of the bottom quality basis.Thirdly,this paper studied the probability distribution characteristics,texture characteristics and Pace characterization based on the characteristics of multi-beam backscatter intensity maps,and verified the characterization of the extracted features on the underlying information contained in the sonar data with the combination of experimental data.In addition to this,this part also proposed a probability ratio peak feature extraction method based on the multi-beam angle response curve.At the same time,the drawn three-dimensional probability ratio map can directly magnify the bottom quality characteristics contained in the angle response curve,and qualitatively characterize the bottom quality distribution in the lateral and longitudinal directions of the multi-beam.Using the method of probability ratio curve and sliding window to analyze and calculate the vertical bottom quality boundary of the multi-beam survey line,which can effectively improve the accuracy of the bottom quality sampling data and correct the error points in the bottom quality sampling data,and reduce the classification accuracy drop caused by the error of the bottom quality label.Moreover,this method can be used as an auxiliary means to reduce the degree of human experience intervention in the processing of sediment sampling data and increase the processing speed.Finally,this paper studied the classification of sonar data with citing CNN and SVM two classifications of sonar data classification algorithm research respectively,at the same time,the data is processed to match the algorithm the characteristics of each algorithm according to the characteristics of each algorithm.At last,the model has achieved excellent fitting effect,and the inverse sediment distribution map is highly matched with the sampling data,so it has practical value through the effective correction of the original sonar data and the bottom sampling data.