Study On Moment Technique Applications In Underwater Acoustic Image Classification And Recognition

Underwater target classification and recognition have very important applications in the fields of military,ocean resources exploitation,maritime search and rescue.It has always been recognized as a difficult problem at home and abroad.In this paper,monitoring mariculture process using acoustic method based on image sonar is proposed to cover the shortage of optical method.The key problem to be solved is exactly underwater target classification and recognition.In order to effectively recognize the two underwater targets of starfish and scallops,this paper focuses on the application of moment technique in underwater acoustic image feature extraction.Three representative methods of moment feature extraction,the Hu moment technique,the moment technique based on Radon transform and the Zernike moment technique,are used to extract the feature of the underwater acoustic image of starfish and scallops.The influence of four factors which are target type,spatial transformation type,moment technique type and calculation data range on the value of the extracted moment feature are analyzed.Shape expression performance,calculation time and programming difficulty of different methods are compared and summarized.Acoustic image classification test was conducted to discuss the recognition results of different technique methods.Before this,a variety of image preprocessing methods were studied.Methods which are suitable for starfish and scallops are selected and used in the preprocessing of image collected by sonar scanning experiment.In the classification and recognition test,this paper compares and analyses the influence of moment technique type,data domain of calculation and quantity of features on the recognition accuracy of starfish and scallops.The results of the research show that Zernike moment technique is superior to other methods in describing the shape of the image,reducing the calculation time and improving the recognition accuracy.Using Zernike moment technique can distinguish acoustic images of starfish and scallops effectively and obtain high recognition accuracy meanwhile.Summarizing the research,we draw a conclusion that owing to the desirable accuracy of underwater acoustic image recognition using moment technique,monitoring mariculture process based on image sonar is feasible from a technical perspective and valuable for further research.