The Research Of Algorithms For Moving Target Recognition And Tracking In Image Terminal Guidance Of Missile

With the development of information technologies, the complicated and variable war environment requests precision-guided weapon higher performance. The precision-guided technique is the core of precision-guided weapon, which is needed to detect, recognize and track targets precisely. In this paper the methods of recognize and tracking of moving objects in image terminal guidance of missile are researched based on the advanced theory results in the field of present terminal guidance of missile.Firstly, in respect of target feature recognition, research the diverse features of the target image in movement scene on pixels level and image level respectively, such as color histogram feature, NMI (Normalized Moment Inertia) and edge histogram feature etc. Base on the principle of color histogram, introduce a color self-correlation histogram method which considering both of the statistic and spatial distribution information in order to resolve the misrecognition problem caused by the similar color distribution between target and background.Secondly, in respect of target tracking, comparatively study the existing popular target tracking algorithms and mainly focus on the linear kalman filtering algorithm and nonlinear particle filtering algorithm, test and verify the performance of algorithms through contrastive simulation experiments; For the sequential visible images and infrared images, realize the whole recognition and tracking process by combining the target feature recognition and filtering tracking methods and analyze the experimental results of tracking error; Through the implementation on real scene sequential images of recognition and tracking algorithms, analyze the effects on tracking performance of particle number, resample technology and noise of equation of motion etc., improve the accuracy of tracking algorithm. Finally, through sequential images simulation experiments and principle analysis, propose a improved adaptive particle filter algorithm by tuning the number of particles, resample interval and the noise variance of equation of motion timely, guarantee the accuracy of tracking process and simultaneously reduce the computational complexity, improve the real-time property and robustness of tracking algorithm.