Classification Of Seabed Sediment And Terrain Complexity Based On Multibeam Data

With the development of acoustic detection technology,the multi-beam system is widely used in seabed topography surveying and mapPing exploration of the seabed sediment,ocean engineering,and other marine exploration aspects with its advantage of full coverage,high precision and high resolution.The multi-beam system mainly records three important information: water depth data,sound intensity data and water column data.Water depth data can be used to describe the submarine topography and geomorphology.Sound intensity data can be used to study the classification and distribution of seafloor sediments.As an important geological interface,the seabed has important geological significance,and its shallow bottom sediments has great significance to explain the evolution of the climate and the deposition process,to detect the submarine minerals and the gas hydrate.The water column data covers all the reflections and scattering of acoustic signal in the process of water body.It can be used to study the suspended matter in the whole water body.In this paper,we discuss the application of the two kinds of data in the classification of sediment and seabed terrain complexity.The main work and contributions are described as follows:(1)This paper introduces the significance of multi-beam data to the classification of seabed and the topographic complexity,and introduces the research status of seabed and terrain complexity classification based on multi-beam at home and abroad.The problems and deficiencies in the data processing,the establishment of classification index and the construction of the classification library are summarized.(2)The principles of underwater acoustics,system composition and classification and working principle of multibeam system are introduced.The principle of seabed sediment classification based on multibeam backscatter data is presented.(3)The process of multi-beam echo intensity data processing and analysis is given from the decoding of multi-beam data.The water depth data is fitted to realize the full extraction of sound intensity data,so the correction model of sound area is improved.Different filtering methods are compared.The Wiener filtering is carried out to filter the sound intensity data and the water depth data,and the bilateral filtering method is selected for sonar image filtering.The sound energy compensation model is constructed.The advantages and disadvantages of different methods are compared,such as inverse distance weighted interpolation,spline interpolation and kriging method.The feature extraction of sonar image is summarized.(4)Using gray value and 6 texture features of sonar image extracted from gray level co-occurrence matrix to study on the classification of sediments by using ISODATA,SVM and BP neural network.(5)The CNN is applied to the classification of CNN classification.Based on the detailed principle of CNN classification,16 image texture features and acoustic intensity gray information are extracted as feature information for the classification of the material.The classification results of ISODATA,SVM,BP neural network and CNN four methods are compared.The classification results demonstrate the superiority of the CNN algorithm and the effectiveness of the classification results.(6)The angular response curves of six types of sediments,such as muddy sand,gravel,bedrock,fine sand,sandy mud and coarse sand,were extracted by using the angle response analysis model.Seven parameters for each category such as mean,mode,range and standard deviation and so on and four model parameter characteristicsand of Hellequin L to sediment classification by using the minimum distance are extracted.For each category,100 Ping data were selected on the establishment of the angle response analysis model to test the effectiveness of the test.(7)Based on the traditional single classification index,the slope and the fluctuation degree are introduced as the new classification index and quantified into the submarine terrain complexity classification.The two-dimensional model complexity method are introduced in the submarine terrain complexity classification.The spatial resolution is unified to 100 m.Using BP neural network to establish the 18 terrain feature database to realize the automatic division of the complexity of the submarine terrain,the classification accuracy is high,not only can the qualitative classification of the experimental area can also be quantitatively analyzed,The classification of advantages.