Analysis of Acoustic Depth Sounder Signals with Artificial Neural Networks (FinalReport)

摘要

Research was conducted on 3 problems involving the analysis of acoustic depthsounder returns that concern hydrographic surveyors, using artificial neural networks. The first problem is the detection of a suspended layer of material called fluff that lies above the top layer of the bottom and poses no obstruction to navigation, but appears to the conventional depth sounder as the hard bottom. The second problem involves classifying the top layer of material as to hard silty sand, hard silty clay, or soft clay. The third problem involves classifying the density of the top layer of material of the bottom. Neural network models based on the back propagation learning method are designed and tested. Hardware is proposed for the implementation of these models. The accuracy of the models developed is compared with a conventional mathematical method.