Font Size: a A A

Research On Technology Of Ground Object Tracking For Imaging Seeker

Posted on:2015-07-21Degree:DoctorType:Dissertation
Country:ChinaCandidate:L R ShenFull Text:PDF
GTID:1108330479479552Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Imaging guidance plays an important role in the field of precision guidance for its advantages, such as strong anti-interference ability, high precision, high combat cost-effectiveness ratio and easy to carry. Based on the imaging seeker, this thesis studies the key technology of the imaging seeker for the ground target tracking, in order to hit the ground still targets(such as airports, docks, islands, bridges, power plants, transportation hub, etc.) and slow moving targets(surface ships, tanks, etc.). The main tasks and innovations of this thesis are as follows:1. A Mean Shift tracking algorithm based on the space-color feature is proposed. In view of the traditional Mean Shift tracking algorithm using the regional color histogram statistics as the features of target tracking, resulting in losing the target position information, this thesis divides the target area and describes the target with the combined features of space-color. It also introduces a commonly used Bhattacharyya—a more robust and more sensitive similar measure, which can directly calculate the mean value of candidate target pixels and the distance of template figure of each pixel. The similar measure considers the differences of position on the basis of color differences of the original one, and deduces the formula of target tracking and proposes an improved Mean Shift algorithm. The results of the study show that the algorithm proposed in this thesis has a better tracking performance when dealing with some rigid body or non-rigid body deformation, the change of target scale and obscured situations.2. A Mean Shift tracking algorithm with Lucas-Kanade algorithm is put forward. The original Mean Shift has two shortcomings, one is that its low target tracking accuracy because its nonparametric estimation, which can not get the target motion parameters. The other is that Mean Shift does not have a template updating strategy, and it may have template drift after long time tracking. This thesis can solve the two problems by introducing the Lucas-Kanande algorithm. Firstly, we can roughly calculate the target position of the current frame using the Mean Shift algorithm. Secondly, we can search the motion parameters between the two target frames through Lucas-Knanade in the neighborhood of this position. The translation parameters are used to determine the target tracking position, and all the kinematic parameters are for the template updating strategy designed in this thesis to calculate whether to replace the template. The results of the study show that the algorithm proposed in this thesis can track targets well and is more accurate to deal with the deformation of the rigid body and non-rigid than the original one.3. A multi-sub template tracking algorithm based on sparse representation is presented. First, this thesis defines the selection principle of the sub template and put forwards a method of selection accordingly. Then it introduces sparse representation to describe the sub template to deal with the shortcoming of histogram’s sensitivity to light, which expands the application of the algorithm. After that, it votes on the target location by constructing a vote table. Whether to update the sub template is by judging the results of sub template and sub target in the rectangle, and it also introduces the training module of the sub template to represent the contribution of different training module. There are two steps of template updating: substitution of training template and updating the proportion. The results of the study show that the multi-sub template tracking algorithm based on the sparse representation is able to deal with many challenges, such as deformation, changes of illumination, occlusions, false target jamming and background interference. It also boasts the high tracking precision and a wide range of application.4. A distribution fields tracking algorithm based on Bayesian mutual information is proposed. It is a challenging problem when there is the position error caused by the target deformation and the features error caused by the changing of light during the tracking process, thus this thesis introduces distribution fields to solve this tracking problem. On the basis of analyzing the distribution fields principle, this thesis proposes a similar measure based on Bayesian framework of mutual information to match the real-time image and template image in different feature layers accordingly. During the match process, only the significant area in different feature layers will be calculated, and it is a method based on local characteristics, thus can better deal with the screening problem. Then the author designs a voting algorithm combined all the matching results in different layers for target tracking, and introduces forgetting factor to update the template. The result of the study shows that the distribution fields based on Bayesian mutual information tracking algorithm in this thesis can deal with some cases of deformation, changing of illumination, partial or full occlusions and complex background interference. It also boasts high tracking precision and a wide range of application.5. Tracking experiments via recording videos by imaging seeker are validated. In order to sufficiently verify the campaign efficiency, simulating the environment of missile to test the performances of algorithms is needed after indoor testing. Two method are introduced in this thesis: one is simulating the missile shaking and investing the influence of outdoor factors such as weather, temperature and so on; the other is closed-loop simulating the tracking status with the missile’s pose change in flight and investing the influence of tracking accuracy to hitting accuracy by combining inertial navigation system(INS) information. Experiments results show that the algorithms proposed in this thesis has good tracking performances, can hit the target with high accuracy. These Experiments establish the base for the next outdoor test and hardware-in-loop(HIL) simulation test on image matching processor, and even for the flight experiment.
Keywords/Search Tags:Imaging Seeker, Object Tracking, Global Feature, Local Feature, Mean Shift, Sparse Representation, Distribution Fields, Template Updating
PDF Full Text Request
Related items