Image Registration Research Based On The Gradient Threshold Segmentation And Hybrid Evolutionary Algorithm

With the development of ship automation,the requirements of cabin safety and intelligent are higher and higher,so the installation of multi camera has been gradually become the standard configuration in the cabin.After image information is collected and fused,characteristics of the image information will be showed more clearly,it is useful for more accurate artificial processing or computer automatic processing.To gethigh quality image fusion,we must implement the accurate image registration,which is the base of image fusion processing.The presence of a number of external factors(such as camera maintenance is not timely,inside the cabin air pollution,etc.)caused the image contains many kinds of noises,such as Gaussian noise,multiplicative noise,poison noise and impulse noise,etc.,and containing noise image registration is still the main difficulty.This paper is aimed at this kind of engine room monitoring image registration problem with complex noises.The main research contents are as follows:1)The definition of image registration is explained with mathematical model,and then the main frame of the image registration is introduced,including space transform,image interpolation,similarity measure and optimization algorithm,and the main modules is stated for mathematics.2)Analyzes the role of entropy and mutual information in image registration,and according to the relationship of mutual information and f information,using I_??information as image registration measures.The measure relative to the mutual information measure can produce fewer local minima point.Through the experiment to get the best value range of the?,and determine the multi-stage registration,?value is at each stage.This measure is helpful to realize high efficiency,preciseness and good robustness of image registration.3)To solve the problem of containing complex noise of image registration,based on gradient threshold segmentation method was proposed to enhance the image of the contour information,at the same time can improve the space position information of pixels of the gray level information.The experimental results show that this method greatly reduces the noise influence on image registration information,so as to improve the precision of the registration.4)To solve the problems multiple local extreme value point of the objective function,an improved QPSO algorithm is proposed.A kind of?(t)adaptive control strategy for compression expansion factor is designed,thus effectively avoid the premature convergence particles,the global optimization ability is strengthened.The experimental results show that the improved QPSO algorithm achieved the purpose of improve the ability of global optimization.5)Implements the multi-stage multiresolution image registration method.combined the improved QPSO calculate methods and NM simplex method,in the first stage,improved QPSO calculate methods is used,in the second stage of the fine registration process,the NM simplex method is used,so as to achieve the purpose of improve the success rate of registration.The experimental results show that the larger the initial simplex setting ratio,the more advantageous to improve registration.