Font Size: a A A

Research On Underwater Image Matching Based On Hausdorff Distance And Genetic Algorithm

Posted on:2017-12-07Degree:MasterType:Thesis
Country:ChinaCandidate:Q W QiuFull Text:PDF
GTID:2348330512965230Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
Because of the growth of population,land resources cannot meet the needs of mankind so that people have turned to the ocean involuntarily,thus the rising underwater image processing technology gets more and more national attention.As a core technology of image processing,image matching has already applied to the technologies such as deep-sea exploration,target identification and location and so on.Since the complex environments in submarine,low visibility,underwater images are severely affected by noise,edge blur and prone to distortion and occlusion,the underwater images are atomized state because of light scattering effect of the water,all of these will bring enormous difficulties to underwater images match,and the original image matching technology is not necessarily suitable for handling underwater images.Therefore,it has important significance and value that finding a method to match underwater image accurately and faster.One of the key factors of image matching is to select the effective method to evaluate the similarity of the image.Since 1991,Hausdorff distance was proposed as the similarity measure,and then HD distance is widely used in image matching research as an evaluation criterion.However,the traditional Hausdorff distance has the sensitivity of noise and shelter and false edge.In this paper,several matching situation of improved HD distance are analyzed in a variety of environments.The improved STMHD matching method is proposed,which can overcome the influence of the difficulties on the matching accuracy.Genetic Algorithm is used to improve the matching speed of the matching search.In research method,firstly,combining with the characteristics of underwater image,using method based on PCNN to filter out noise,to enhance the image with Retinex method;then selecting Canny operator to extract the image edge;choosing genetic algorithm as the search strategy,the fitness function of genetic algorithm constructed by improved STMHD is a basis for selecting optimal transformation of translation,scaling,rotation.And the genetic operator is improved by using methods of adaptive crossover and mutation as well as individual self-learning.In this paper,the underwater image is simulated,and the optimal parameters of the algorithm are fluctuating near the exact value,matching time is greatly shortened compared with other algorithms.Matching the image with noise,the matching accuracy is remain same;when the image shielding ratio is 10%,matching accuracy rate is up to 96% and shielding ratio is within 25%,algorithm is still effective;when the image rotation is within the range of 30 degrees,matching error can be controlled within 15 pixels;to match the image after narrow five times,matching accuracy assurance will be within 92%.The results show that this algorithm can overcome the influence of translation,scale transform,partial occlusions,and noise on the matching,and matching rate is high,real-time is good.
Keywords/Search Tags:underwater image, Image matcing, Hausdorff distance, Genetic Alogorithm, Individual's self-learning
PDF Full Text Request
Related items