Research On Sub-Bottom Profile Data Processing And Seabed Sediment Classification

The exploitation and utilization of seabed sediments provide vital significance in many fields,particularly the marine military,marine resource exploration,underwater archaeology,ocean engineering construction,and marine fisheries fields.The classification of seabed sediments has become a research focus in the field of marine geodesy and cartography.Sediment categorization approaches are divided into the traditional method,drilling and sampling,and the modern acoustic remote sensing detection method according to the means of data acquisition.Compared with the traditional drilling and sampling method,acoustic detection method not only improves detection efficiency but also significantly reduces operating costs,which is suitable for large-scale seabed sediment detection,but the accuracy of sediment classification needs to be further improved.This paper focuses on sub-bottom profile detection in modern acoustic remote sensing detection,and mainly concentrates on the processing of sub-bottom profile data and the classification of seafloor sediments.The main innovations and researches are as follows:(1)In order to implement the denoising processing of the sub-bottom profile images with high fidelity,this paper proposes a denoising processing method based on the improved median filtering.The median filtering is improved from the two aspects of noise point judgment and adaptive template size selection.The result shows that the improved sub-bottom profile image not only has good denoising effect,but also retains the key information in the image.(2)In view of the traditional predictive deconvolution algorithm in the sub-bottom profile images has the problem of incomplete multiple wave suppression,in this paper,the adaptive predictive deconvolution algorithm is put forward based on the partition processing that different suppression methods are adopted in different regions.The result shows that the improved predictive deconvolution algorithm has a better suppressing effect and is more suitable for the multiple wave suppression of sub-bottom profile under complex terrain conditions.(3)To overcome the flaws that Biot-Stoll model and attenuation-based model have low classification accuracy for gravel and mud respectively,this paper introduces a sediment classification method combined with these two inversion models.The reflection coefficient R and quality factor Q are extracted from the sub-bottom profile data,and the Biot-Stoll model and attenuation-based model are employed to determine the mean grain size.The obtained two datasets of mean grain size are input into back propagation neural network(BPNN)for training and classifying.The experimental results show that the classification results obtained by the training of BPNN are significantly improved compared with the single model,which demonstrates that these two inversion models have strong complementarity,and the feasibility of using this combined method in the research of seabed classification.(4)To improve the automatic classification accuracy of sub-bottom profile data,this paper adopts the ant colony(ACO)algorithm to optimize the BPNN(ACO-BP).The global optimization characteristics of ant colony algorithm is applied to solve the problem that the neural network is easy to fall into the local optimal solution during the training.Experimental results show that the accuracy of ACO-BP algorithm is 4%higher than that of traditional BPNN,which verifies the correctness and effectiveness of ACO-BP algorithm.