Sarifah, Luluk (2020) Kombinasi Metode Filter Korelasi Dan Filter Kalman Untuk Pelacakan Kendaraan Bergerak Dengan Penanganan Oklusi Pada Video Transportasi. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
06111850010012-Master_Thesis.pdf Download (4MB) | Preview |
Abstract
Perkembangan teknologi pengolahan citra digital memiliki banyak manfaat salah satunya dalam bidang pelacakan objek (object tracking), seperti pelacakan kendaraan bergerak pada video transportasi. Pelacakan pada kendaraan bergerak memiliki tantangan, salah satunya ketika terjadi oklusi, yaitu gangguan yang menyebabkan keakuratan dari suatu pelacakan objek berkurang bahkan mengalami kegagalan. Penelitian ini bertujuan untuk menerapkan kombinasi metode Filter Korelasi dan Filter Kalman untuk melakukan pelacakan kendaraan bergerak pada video transportasi. Metode Filter Korelasi digunakan untuk mendeteksi gerakan atau perpindahan dari objek kendaraan, sedangkan metode Filter Kalman untuk mengantisipasi atau
memprediksi gerakan dari objek kendaraan. Salah satu keistimewaan dari metode Filter Korelasi ialah kecepatan dan perhitungannya yang efisien dalam pelacakan visual, tetapi terkadang kurang robust (handal) dalam penanganan oklusi, sehingga sering mengalami perbaikan dari algoritmanya. Sedangkan metode Filter Kalman memiliki keistimewaan berupa hasil prediksi yang presisi. Persentase akurasi diperoleh dengan mendeteksi frame yang terlacak benar dan frame yang terlacak salah dari setiap frame video, baik video tanpa oklusi maupun video dengan oklusi yang bersifat single object dan multiple objects. Dari simulasi yang dilakukan pada keseluruhan video, diperoleh rata-rata persentase akurasi sebesar 89% dengan metode Filter Korelasi dan 94% dengan metode kombinasi pada video tanpa oklusi, sedangkan rata-rata persentase akurasi pada video dengan oklusi diperoleh sebesar 78% dengan metode Filter Korelasi dan 84.3% dengan metode kombinasi. Adapun dalam penanganan oklusi metode kombinasi mampu menangani oklusi sebagian akan tetapi belum sepenuhnya mampu menangani oklusi secara keseluruhan.
==================================================================================================================
The development of digital image processing technology has many benefits, one of which is in the field of object tracking, such as moving vehicle tracking on transportation videos. Tracking on a moving vehicle has challenges, one of which is when occlusion occurs, which is interference that causes the accuracy of an object tracking to be reduced or even fail. This study aims to apply a combination of Correlation Filter and Kalman Filter methods for moving vehicle tracking on transportation videos. Correlation Filter method is used to detect movement or displacement of vehicle object, while Kalman Filter method is to anticipate or predict the movement of vehicle object. One of the features of Correlation Filter method is that its speed and calculation are efficient in visual tracking, but sometimes it is less robust in handling occlusion, so it often experiences improvements from the algorithm. While Kalman Filter method has the privilege of being precise prediction results. Percentage of accuracy is obtained by detecting frames that are tracked correctly and frames that are tracked incorrectly from each video frame, both
videos without occlusion or videos with occlusions that are single object and multiple objects. From the simulations performed on the entire video generated
an average percentage of accuracy 89% with Correlation Filter method and 94% with combination method on videos without occlusion, while the average percentage of accuracy on videos with occlusion 78% with Correlation Filter method and 84.3% with combination method. As for the handling of occlusion, the combination method can handle partial occlusion but is not yet fully able to handle overall occlusion.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Filter Kalman, Filter Korelasi, Oklusi, dan Pelacakan Correlation Filter, Kalman Filter, Oclussion, and Tracking |
| Subjects: | Q Science > QA Mathematics |
| Divisions: | Faculty of Mathematics and Science > Mathematics > 44101-(S2) Master Thesis |
| Depositing User: | LULUK SARIFAH |
| Date Deposited: | 23 Aug 2020 14:08 |
| Last Modified: | 03 Jul 2023 07:49 |
| URI: | http://repository.its.ac.id/id/eprint/79270 |
Actions (login required)
 |
View Item |