Font Size: a A A

Scheme For Moving Multicolor Object Tracing In Dynamic Scene

Posted on:2009-07-31Degree:MasterType:Thesis
Country:ChinaCandidate:Z LiuFull Text:PDF
GTID:2178360242980212Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
This article proposed an idea that first extracts moving target with abtracting partial background under static scenes, and then adopted of rotary scanning and tracking contour with chain code to track moving targets in the dynamic scene for moving target tracking system. Based on these, designed several important part in a tracking system.Firstly, research background and the status quo are introduced, as well as the basic characteristics of principle of moving target tracking both at home and abroad. Summed up the moving target tracking system, the basic framework, and some of the typical moving target tracking method, discussed the advantages and disadvantages of these algorithms as well as the scope of application.Secondly, on the basis of introducing color space, image morphology and adjacent interframe difference theory, leads to a improved algorithm of simple difference between frames and its advantages, which put forward the algorithm of extracting object in this paper. Then a variety of algorithms were simulated to test the algorithms of this paper has better characteristics than the traditional difference method. The traditional frame difference has simple performance,fast computing speed and better test results, but more sensitive to the dynamic changes of scenes such as illumination or external conditions; The target location of symmetric difference extraction are fairly accurate, but follow-up frames are needed, in this way extracting object in real-time monitoring will have a certain time delay. The method of the local background difference proposed in this paper based on the adjacent frame difference can extract more complete objects, more accurate location, and no time delay. However it can only used when objects left completely from the initial position, therefore it is not appropriate to the objects which have small movement scope.Thirdly, the typical characteristics of moving targets are introduced. Thereout the characteristics of moving targets of this paper are choosed on the basis of the application, mainly containing color features and configuration characteristics. whereafter a algorithm of target tracking based on rotation scanning and contour tracking with chain code is put forward. Research leads from the rotating search to the rotating scan algorithm, and analyzes its advantages and disadvantages, gives concrete realization. Subsequently, the research outline of the chain codes tracking principle, given the chain codes tracking outline concrete steps.Finally, based on the above, given the overall moving target tracking algorithms, and conduct a feasibility analysis and simulation. Subsequently matching and storage of moving target features is studied, to address detection of recurring target. Started with several issues in matching chain codes, by that analyzed several improved chain codes, and then proposed an idea of variable length codes based on the relative chain codes. The traditional method of linear chain codes can not guarantees invariability under rotation and stretching. By the analysis, variable length codes based on the relative chain codes proposed by this paper can adapt to the starting point changes of codes, and can remove tiny noise, reduce match computation,and improve the accuracy of the match.The paper mianly including the following four aspects.(1) According to the characteristics of tracking moving targets designed moving target tracking system model, which proposed that abtracting moving target location and characteristics from the static environment and tracking target in dynamic scene, and set initial state to solve the problem losing target.(2) In order to abtract integrated location and characteristics of moving target, the method of the local background difference is proposed in this paper from analysis frame difference and improved algorithms.(3) In accordance with principles of the camera work and color space compared RGB spatial information is choosed to to process images.Then an idea was made that dividing a target into regions respectively abtracting contour according to the color characteristics. To extract modality characteristics, chain code tracking contour method is choosed. In order to complete the border of the region and the characteristics of color region, The rotational scanning algorithm is put forward from the principle of rotating search algorithm, and concrete realization is given.(4) In order to solve problem of moving target storage and matching, an idea that the variable length chain codes based on the variable length chain codes is proposed, with the thinking of the analysis proved that this method can resolve problems chain complexity of the codes matching affected by the beginning and length of the chain codes, and small noise can be removed.Through a large number of simulation experiments and comparison with other algorithms proved that the algorithm presented in this paper can address a number of issues and have adaptability in a certain degree, can be effective tracking moving targets, and carry out the better characteristics of the moving target storage and match.
Keywords/Search Tags:Moving target, Multicolor, Dynamic scene, Local background difference, Color region, Rotational scanning algorithm, Variable length chain codes
PDF Full Text Request
Related items