Research On The Method Of Automatic Feature Extraction Of Typical Target By Side Scan Sonar Based On UUV

Unmanned underwater vehicle(UUV)is a kind of self-energizing,autonomous navigation and autonomous operation underwater robot.Because of its advantages of wide scope of work,good stealthiness,high intelligence and little influence form environment has been rapidly developed and widely used.UUV can perform reconnaissance mission,detection mission and data collection task by carrying underwater television,side scan sonar,ADCP and hydrophone,Because of high resolution and far effective working distance,side scan sonar as one of the main underwater sensor has been widely used in marine surveying and mapping,marine geological survey,marine engineering and searching for underwater target,etc.UUV has become one of the most effective and convenient way to acquire the information of submarine geomorphology and search for underwater target.UUV comparing with surface warship by carrying side scan sonar to perform mission has the advantages that there is little influence from climate and wind wave and sweeping effect is good;it can sweep automatically and reduce the amount of staffs;it can be measurement in secret and can be used for military tasks.But UUV still exist some disadvantages in searching for underwater targets by carrying side scan sonar: After sweeping,UUV needs to rise to the surface,export data manually,process data and judge if there is the target or special object.If exist,it need to submarine once again to shoot underwater.The deficiency is the key point of technical improvement,and when the water depth is deeper,rising to the surface and submergence will consume large amounts of time and energy.In order to improve the ability of autonomous searching target,this paper presents a process and relevant methods of automatic feature extraction of underwater target based on UUV.After reading and studying a lot of literature,the main research contents are as followings:(1)According to the research on the working principle of side scan sonar,the format and information of side scan sonar file,and combining with side scan sonar interpretation,the description of the edgetech(.jsf)file format and relevant literature,extract the effective data including the latitude and longitude information,attitude information,side scan sonar work information and side scan sonar sample data.Moreover transform original data to sound intensity by the conversion model.Transform sound intensity to gray level transformation.(2)Design the imaging process of UUV sweeping,meanwhile this paper analyze traditional method of mean filter,median filter,wiener filter and Gaussian filter to reduce or eliminate underwater noise impact on side scan sonar image.According to the characteristics of the side scan sonar image noise,adopt a new method of side scan sonar image filtering.(3)Study the K-means clustering algorithm and the fuzzy C-means(FCM)clustering algorithm,and build up Markov Random Field(MRF)model.According to the MRF model,use K-means clustering algorithm to initialize Markov Random Field model.In order to satisfy the rapidity,use ICM(iterative condition model)algorithm for side scan sonar image segmentation.(4)After segmentation,extract the feature of underwater aircraft target in side scan sonar image.According to the characteristic of sonar image after segmentation,select the corner feature and line feature as the underwater aircraft characteristics.Through the study of the traditional methods,use a new model-mean method to extract the line feature and combine with corners detector method to complete feature extraction.Making use of the feature points extracted and stored,and combining with the latitude and longitude information and attitude information,through the transformation of the relationship between coordinate system,undertake the study on the process of point calculation,record the latitude and longitude information.