Study On Target Orientation And Recognition Of Underwater Security-Defense System

The marine oil-gas fields are specially located and highly concentrated,often being destroyed by neighboring hostile forces.Underwater security-defense system is based on the actual environment of the South China Sea Liuhua oilfield,through a variety of sensing devices to achieve the comprehensive protection of the oil platform and underwater production facilities.In this paper,an active detection scheme with multi-beam sonar is proposed for the underwater vehicle and frogman with small volume and low noise level.Aiming at the problem of orientation and recognition of underwater targets is of strategic significance for engineering personnel to grasp the current situation,early detection of the signs of accidents and take further preventive measures.In order to cope with the complicated marine international environment,the underwater security system,which is composed of the underwater wet end,the water surface equipment and the intelligent monitoring center is constructed,and the data communication between subsystems is realized.Based on the properties of underwater targets and the principle of multi-beam sonar detection,an active detection scheme is proposed,which integrates the target signal acquisition,processing and display functions.By analyzing the advantages and disadvantages of the commonly used underwater acoustic signal processing methods,the EMD decomposition algorithm of local wave theory is used to extract the target feature information.For the orientation problem of underwater targets,a mathematic model of cylindrical matrix is established.Based on the product theorem in beamforming theory,the spatial directivity functions of cylindrical arrays are deduced from the depth direction and horizontal direction respectively.The effect of array element number and layer number on the directivity is studied,by the simulation results and the actual size of the array,the optimal array structure is designed and the main beam width and pitch rotation range are verified on the basis of this formation.Finally,the two-dimensional MUSIC algorithm is used to estimate the azimuth of the multi-target.Aiming at the problem of spatial azimuth ambiguity caused by the close distance between targets,by reducing the search dimension of the MUSIC algorithm,more accurate target azimuth information is obtained.For the recognition problem of underwater targets,the validity of the EMD algorithm is verified by the feature extraction results of the simulation signals.Through EMD decomposition of the target echo signal collected by lake test,several IMF components and margin are obtained.Since the IMF components of different orders contain different echo components,the first eight orders IMFs with larger order are selected as the feature extraction objects according to the decomposition distortion degree,and then the energy features,energy entropy features and singular value characteristics are extracted,so as to realize the inter-class recognition of two kinds of targets: frogman and underwater vehicle.Aiming at the problem of type identification of underwater vehicle,a BP neural network classifier was designed and trained for the eight-dimensional feature samples extracted from the three scales.Based on the test results,the dimensions of the feature vectors in the original samples are compressed,and the classification result with higher recognition rate is obtained.