| The exploitation and utilization of marine resources, particularly oil and gas resources, is a very important part of China's "marine power" strategy. Sediment structure detection based on acoustic method plays an indispensable role in this work. Multibeam and sub-bottom profiler system are capable of detecting large-area seabed sediment rapidly, while the former can detect seabed sediment, the latter can detect seabed and sub-bottom sediment with depth of a few hundred meters. Compared with the traditional drilling and sampling method, acoustic detection method based on multibeam and sub-bottom profiler system not only improves detection efficiency but also significantly reduce operating costs. Nevertheless, it is undeniable that this method still has many unsolved problems. Therefore, after analyzing the existing research status quo of seabed sediment acoustics detection, this paper emphatically conducted the research of key issues of seabed and sub-bottom sediment classification, improved the existing acoustic detection theories and methods, and then realized the accurate detection for seabed and sub-bottom sediment.The main works and contributions of this paper are as follows:(1) The acoustics principles, system composition and operating principles of the multi-beam and sub-bottom profiler system are discussed, and the principle of sediment classification based on multibeam and sub-bottom profiler backscatter intensity are given.(2) Explains the data processing process of multibeam backscatter intensity image,analysis the deficiency of the existing sound ray tracing method and the influence of sea bottom projection point calculation error on backscatter intensity. Then put forward the three dimensional sound ray tracing method with consideration of attitude angle. The tracing method achieved an average tracking precision of 5%o and improved the calculation precision of position and backscatter intensity of beam projection point on seabed.(3) To weaken the influence of multiple wave on sub-bottom profiler's effective primary wave, this paper first analyses the advantages and disadvantages of predictive deconvolution method and feedback loops method for suppressing the multiple wave. Then on the basis, a comprehensive suppression method based on the combination of the above two methods was proposed. The proposed method improved the PSNR (peak signal noise ratio) of the sub-bottom profiler data by 15% and the image fidelity reached to 0.87 after suppression, which significantly increases the quality of the sub-bottom profiler's echo intensity data.(4) To overcome the weakness of single peak and valley method for boundary demarcation, this paper presented a comprehensive boundary demarcation method for sub-bottom sediment combining the curve peak and valley of return loss and sediment quality fector Q. Referenced to drilling data, the comprehensive demarcation method achieved a relative boundarydemarcation accuracy better than 2%, which is obviously superior to the traditional peak and valley method.(5) To weaken the influence of energy attenuation of sound wave propagating in the sub-bottom on echo intensity and classification, with the help of inverse Q filtering sound energy compensation method based on acoustic wave theory, this paper calculated the phase operator and amplitude operator of sound energy compensation, and compensated the sound energy attenuation in the propagation process. The experimental results show that amplitude increased by about 18% on average after acoustic compensation, and the reconstructed echo image reflected sediment boundary more clearly.(6) According to the correlation between sub-bottom sound energy changes and seabed sediment, this paper proposed and defined two parameters which are highly associated with sediment, namely the amount of sound energy attenuation BL*i in sediment layer and the amount of sound energy residual compensation ar* i. This paper also put forward the selection principle for sample characteristics, and obtained significant parameters(BL*i, ?r*I, mean, third moment, moment invariants, etc.)for sub-bottom sediment classification, combined with experiment, the method for width determination and optimal selection of sample were given out. Experimental results show that high quality sample and significant statisticalparameters were used, sediment classification accuracy rate is increased by 6%.(7) Wavelet BP neural network classifier is considered the best seabed sediment classifier after analyzing the pros and cons of support vector machine, BP neural network, and wavelet BP neural network, and comparing the sediment classification accuracy by means of the three methods. To solve the problem of slow convergence, falling into local minimum easily, and bigger error jitter at the steep, this paper improved the wavelet BP neural network by adding the self-adaptive learning rate and momentum factor, and optimizing the initial connection weight and thresholds between layers. The experimental results show that the training convergence speed of the improved wavelet BP neural network is increased by 40%, and the fitting error is controlled within 4%, which means the reliability of the seabed sediment classification is significantly improved.(8) With the help of sample optimation, feature parameters extraction, and the improved wavelet BP neural network training, the classifier achieved the seabed sediment supervised classification based on multi-beam scatter intensity data and the sub-bottom sediment supervised classification based on sub-bottom profiler echo intensity data. Experiments show that in the seabed sediment classification, sample recognition accuracy and the classification accuracy were 94% and 98.8% respectively. For the sub-bottom sediment classification, the sample recognition accuracy was 96.26%, and the internal and external accord accuracy of sediment classification were 97.98% and 84.76% respectively.(9) After analyzing the merits and demerits of the seabed sediment classification based on multi-beam scatter intensity data and the sub-bottom sediment classification based on sub-bottom profiler echo intensity data respectively, the strong complementation between these two classification results is found. Therefore, this paper put forward the sediment structure obtaining method for seabed and sub-bottom by mean of the superposition of these two classification results and the method for three-dimensional sediment structure model,thus obtain three-dimensional sediment structure for the seabed and sub-bottom... |
PDF Full Text Request