Identifying license plate character segmentation boundaries using convolutional neural networks | |
| Author | Bazard, Benoit |
| Call Number | AIT Thesis no.CS-17-05 |
| Subject(s) | Image segmentation Neural networks (Computer science) |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Computer Science, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. CS-17-05 |
| Abstract | The aim of the present paper is to design a way to perform character segmentation on license plates using convolutional neural networks without explicitely performing character recognition so as to keep the segmentation and the recognition distincts. This paper details the implementation of the algorithm and an evaluation of its global performance on a dataset of 545 license plates provided by the Vision and Graphics Lab at AlT. The system successively applies two convolutional neural networks to perform vertical segmentation and then horizontal segmentation. The results suggest that convolutional neural networks on their own are not enough to achieve acceptable performance on a dataset of this size. A more sophisticated classifier is necessary to convert the features given by the neural networks into robust character boundaries. The main improvement needed is to eliminate non-character regions identified by the segmenter. |
| Year | 2017 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis : no. CS-17-05 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Dailey, Matthew N.; |
| Examination Committee(s) | Mongkol Ekpanyapong;Vatcharaporn Esichaikul; |
| Scholarship Donor(s) | TELECOM SudParis, France; |
| Degree | Thesis (M. Sc.) - Asian Institute of Technology, 2017 |