Research On Multiresolution Analysis Method For Sonar Image Processing

Sonar imaging is the only effective way to obtain underwater target information from long distances.It is one of the most important tools for exploring underwater unknown worlds.However,its existence is vulnerable to environmental interference,poor image quality and low contrast,which seriously affect the subsequent target recognition It is necessary to preprocess the sonar image before the sonar target is identified to achieve the visual recognition target requirements.Simultaneously,the sonar imaging is different from traditional optical imaging methods.Using traditional image processing methods is not suitable for sonar image processing.So,the study starts from the perspective of human visual perception and information processing,using multi-resolution graphical analysis tools based on wavelet to multi-scale analyze sonar images.Under the deep research of multi-resolution theory,the paper focuses on the foflowing sonar images processing problems.First,Large numbers of sonar image data returned during target detection in broad waters do not always contain targets,so it is necessary to distinguish the target before the mimage is further processed,if merely rely on artificial eyesight to determine the presence of the target of this traditional detection method,Obviously,there are practical problems such as low efficiency,high cost,and long subjective detection cycle.And not all the sonar images returned contain targets in actual underwater detection.If all sonar images are identified,it is bound to cause a large amount of computing resources to be wasted and increase the target false recognition rate.So in this paper proposes texture difference between sonar image with target and background sonar image without a target,Through the idea of "matching" can effectively judge whether there is a goal or not.Secondly,the low-resolution sonar grayscale image is susceptible to noises such as background and environment,resulting in sonar images blurring and unclear target information,which affects the recognition of sonar target.In this paper,a two-dimensional dual-tree high-density wavelet structure is proposed on the basis of one-dirensional dual trees high-density wavelet analysis,and noise reduction is performed on the sonar image in combination with the bivariate shrinkage function.The detail edges of sonar images are preserved as much as possible while reducing noise.Finally,sonar image segmentation is a key research issue in sonar image processing.Its research focuses on the segmentation of target areas and shadow areas from complex underwater reverberation zones without destroying the original image edge information.Most of the target images are debris,and the corrosive effect of the environment further blurs the target edge contours.Therefore,the sonar image segmentation is more complicated than that of the optical image segmentation.In this paper,the multi-resolution geometric analysis tool curvelet transform is used.Considering the spatial correlation relationship between textures,the multi-scale Gaussian Markov level set in the curve let domain is used to segment the sonar image.