Sistem Pelacakan dan Pengenalan Pelat Nomor Kendaraan Berbasis Video Menggunakan Hybrid-CNN Mean Shift

Rahmat, Basuki (2018) Sistem Pelacakan dan Pengenalan Pelat Nomor Kendaraan Berbasis Video Menggunakan Hybrid-CNN Mean Shift. Doctoral thesis, Institut Teknologi Sepuluh Nopember Surabaya.

[thumbnail of 07111160010002-Dissertation.pdf]
Preview
Text
07111160010002-Dissertation.pdf - Accepted Version

Download (5MB) | Preview

Abstract

Penelitian ini mengusulkan metode baru dalam sistem pelacakan dan pengenalan pelat nomor kendaraan berbasis video secara cerdas. Tiga bagian besar yang diteliti, yaitu: bagian ekstraksi pelat nomor kendaraan, bagian pelacakan pelat nomor kendaraan sepanjang frame video, dan bagian ekstraksi karakter, yang di dalamnya terdapat unit segmentasi karakter dan pengenalan karakter pelat nomor kendaraan.

Pada bagian ekstraksi pelat nomor kendaraan digunakan metode Smearing Algorithm. Pada bagian pelacakan pelat nomor kendaraan digunakan metode Basuki-I Ketut-Mauridhi (BIM) Mean Shift (BIM-Mean Shift) dengan Switching Kernel dan Adaptive Fuzzy Gaussian Kernel. Pada bagian ekstraksi karakter, yang terdiri dari unit segmentasi karakter pelat nomor kendaraan digunakan metode Basuki-I Ketut-Mauridhi (BIM) Hybrid Cellular Neural Network (BIM-HCNN), dan unit pengenalan karakter pelat nomor kendaraan digunakan metode Extreme Learning Machine (ELM).

Dari hasil serangkaian percobaan simulasi menunjukkan hasil-hasil sebagai berikut. Smearing Algorithm dapat digunakan sebagai salah satu metode ekstraksi pelat nomor kendaraan. Pelacakan pelat nomor kendaraan dengan menggunakan metode BIM-Mean Shift dengan Switching Kernel dan Adaptive Fuzzy Gaussian Kernel menghasilkan kinerja yang lebih baik daripada Mean Shift standar. Ekstraksi karakter pelat nomor kendaraan, menggunakan metode BIM-HCNN dengan Fuzzy Adaptif dan Neuro-Fuzzy memberikan hasil segmentasi yang lebih unggul dibandingkan CNN standar. Serta pengenalan Karakter Pelat Nomor Kendaraan dengan ELM dari hasil penelitian ini mencapai kinerja yang sama baik dengan metode template matching. ========== This research proposes a new method for
video-based vehicle license plate tracking and recognition
system intelligently. Three major sections were researched : the vehicle license plate extraction, the vehicle license plate tracking along the video frame, and the vehicle license plate character extraction, in which there are the license plate character segmentation unit and the license plate character recognition unit. In the vehicle
license plate extraction section is by using Smearing
Algorithm method. In the vehicle license plate tracking
section, using Basuki-I Ketut-Mauridhi (BIM) Mean Shift (BIM-Mean Shift)method with Switching Kernel and Adaptive Fuzzy Gaussian Kernel. In the character extraction section, which consists of the vehicle
license plate character segmentation unit, using Basuki-
I Ketut-Mauridhi (BIM) Hybrid Cellular Neural Network (BIM-HCNN) method, and the vehicle license plate character recognition unit, using Extreme Learning Machine (ELM)method. From the results of a series of simulation experiments show the following results. Smearing Algorithm can be used as a method of the vehicle license plates extraction. The vehicle license plates tracking using the BIM-Mean Shift methods with Switching Kernel and Adaptive Fuzzy Gaussian Kernel result in better performance than the standard Mean Shift. The vehicle license plate characters extraction, using the BIM-HCNN method with Adaptive Fuzzy and Neuro-Fuzzy gives superior segmentation results compared to the standard CNN. As well as the vehicle
license plates character recognition using ELM from the results of this research achieved the same performance with the template matching method.

Item Type: Thesis (Doctoral)
Additional Information: RDE 629.836 Rah s
Uncontrolled Keywords: Smearing Algorithm; BIM-Mean Shift; CNN; BIM-HCNN; ELM
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions: Faculty of Electrical Technology > Electrical Engineering > 20001-(S3) PhD Thesis
Depositing User: Basuki Rahmat
Date Deposited: 10 Apr 2018 05:13
Last Modified: 25 Sep 2020 02:12
URI: http://repository.its.ac.id/id/eprint/50724

Actions (login required)

View Item View Item