Font Size: a A A

Research On The Target Tracking Algorithm Based On Joint Transform Correlator

Posted on:2011-08-04Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y WangFull Text:PDF
GTID:1118360305990352Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
Joint Transform Correlator (JTC) is now widely used in the field of target recognition due to its high speed, highly parallelizable and programmable characteristics.However, there are two types of weakness in classical JTC.One drawback of classical JTC is that the joint transform power spectrum (JTPS) has a spatially varying average which leads to a large zero order term. The desired crosscorrelation peaks are almost overshadowed. Another drawback of classical JTC is that the size and relative position of target image and reference image of input plane is limited by the wide zero diffraction peak. New types of JTC can be designed by the performance improvement method of weakening or eliminating the zero order term and enhancing the crosscorrelation peak. In these years, improvement for the performance of JTC mainly aims at preprocessing of input plane image and nonlinear processing for JTPS.Swarm intelligence method was inspired by swarm behaviors of some animals. It was designed by simulating these behaviors. Digital image can be taken as a region where many agents live. Then the pixels in the image come into being the resources which the agents depend on. The agents'behaviors can be described by several rules. The pixels that can represent the feature of an image are named"valid resources". The moving, self-reproducing and evolution of the agents are prescribed by the predefined rules. The inquilinous agents in the image represent the feature of the image after all the agents finish searching for the valid resources. In order to improve the performance of JTC the Swarm Intelligence based JTC is proposed in this paper on the basis that deep analysis is performed on the state-of-the-art JTC.The experimental results show that the performance of the proposed Swarm Intelligence preprocessing based JTC is better than those of the JTC based on other preprocessing methods.Relation between the position of the crosscorrelation peak in the output plane and the relative position between the target object and the reference object in the input plane of JTC is deeply analyzed. Then the theory of translation-invariant for JTC is proposed. The peak value is analyzed quantificationally. On the basis of that, a tracking algorithm based on JTC is proposed. Precise target template is unnecessary for the proposed method. The layered matching like classical correlation algorithm is unnecessary in the tracking process either. Joint correlation transform can be performed by taking the former frame as the reference image and taking the later frame as the target image. Then the relative shift between the target in the later frame and the reference object in the former frame is obtained. Tracking is performed by compensating the shift. Because of taking the whole characteristic into account the algorithm is unsensitive to the detail variation. It can track the target well though the target is revolved or distorted. Frequently changing of the template is not necessary for the algorithm and the inherited error of template changing is reduced compared with the classical tracking algorithm. Then the template drift possibility is reduced.
Keywords/Search Tags:Joint Transform Correlator, Swarm Intelligence, image preprocessing, target tracking, joint transform correlation tracking
PDF Full Text Request
Related items