|
| |
|
| |
|
|
Robot Localization by 3D map
Mateják, Marek ; Winkler, Zbyněk (advisor) ; Obdržálek, David (referee)
This thesis deals about possibilities of autonomous mobile robot localization by camera in the known environment. In detail the model is analyzed of scene representation by perspective projection including evaluation of the location according to paired points with the scene. This is done especially for the model of the robot with the fixed camera location that moves on the fiat floor. Localization is analyzed from two points of view. From the global view, where is the goal to evaluate position without previous knowledge of the position. And from the local view, where is the location of the robot tracked during its movement in the scene. In the first case is estimation of the position based mainly on the color separation of segments. The case of undistinguishable reference points is analyzed, where the position is estimated according to their field location. On the other hand, when the previous position is known, the two methods of reference points tracking are mentioned. First method is ICP algorithm, which makes the registration of points. The second is the Mean Shift algorithm for tracking image seg1nents. In the end is mentioned the application of Monte Carlo filtration, which assures robustness of segment tracking. Powered by TCPDF (www.tcpdf.org)
|
|
|
Single-Image License Plate Recognition
Vala, Tomáš ; Winkler, Zbyněk (referee) ; Štanclová, Jana (advisor)
The diploma thesis deals with analysis and design of an automatic system for a single-image license plate (LP) recognition. Input data came from a system for car speed measurement on the roads in the emirate of Abu Dhabi (the United Arab Emirates). A goal is to recognize a numeric identi cation code on a LP and a type of the LP. In details there are discussed individual parts of the system for LP recognition. A particular stress is put on segmentation and grouping of individual characters to be able to utilize specific information on the LP from Abu Dhabi during the character recognition. The diploma thesis compares several methods for character recognition - a minimum distance classi er, feed-forward neural networks and hidden Markov models. The methods (and some modifications of them) are tested on real data and compared according to the percentage of correctly recognized digits (at fi rst the digits were manually classi ffied). Wrongly recognized LPs are undesirable. Instead in some cases it is better to mark the LP as "unrecognized". That's why there were introduced sophisticated tests for recognized digits reliability veri cation. A summary of our own results is included.
|
|
|
Semiadaptive 3D modeling
Kubík, Pavel ; Pelikán, Josef (referee) ; Winkler, Zbyněk (advisor)
In first part of this paper, methods for reducing complexity of building 3D maps are introduced. These methods use predictions about scene properties or special hardware devices for reducing complexity such as a laser range finder, an omnidirection camera or both. Furthermore, an algorithm for creating a reconstructed scene using a camera and robot odometry is presented. The method presupposes moving on a straight floor and an indoor environment. A user can interact with the algorithm and help with creating model or with repairing parts of the model, if they were not satisfactorily created by the automatic algorithm. An interaction also allows to apply the method in various environments, which weren't initially planned. So, the adaptation to the new environment may be done by a user. The output model is exported to the common format for virtual reality VRML. This allows to view the model in e.g. a web browser. The presented method was implemented with a user interface to enable an interaction with the algorithm. The method was tested on data from the real world and the effectiveness of each user's interferences was evaluated.
|
guest ::
