Diandra, Wildan Shafa (2023) Reidentifikasi Kendaraan Menggunakan Convolutional Neural Network (CNN) Pada Citra Unmanned Aerial Vehicle (UAV). Other thesis, Institut Teknologi Sepuluh Nopember.
Text
07211940000049-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (18MB) | Request a copy |
Abstract
Beberapa tahun terakhir jumlah kendaraan bermotor di Indonesia terus meningkat. Sejumlah besar CCTV digunakan di sudut-sudut jalan maupun persimpangan untuk membantu proses pemantauan lalu lintas. Sistem pemantauan tersebut sebagian besar berbasis video real-time dan masih membutuhkan manusia untuk mengawasinya sehingga masih kurang efektif. Selain itu, penggunaan kamera pengintai juga kurang efektif apabila digunakan untuk mengawasi banyak lokasi. Satu buah kamera pengintai hanya bisa digunakan untuk mengawasi 1 titik lokasi. Saat ini penelitian mengenai sistem reidentifikasi kendaraan berbasis CNN mulai banyak dikembangkan. Sistem reidentifikasi kendaraan dapat membantu proses pemantauan lalu lintas khususnya dalam memantau, melacak, maupun mencari kendaraan. Pada saat yang sama, pemanfaatan Unmanned Aerial Vehicle (UAV) mulai banyak digunakan di berbagai sektor kehidupan, salah satunya di bidang visi komputer. Pada penelitian ini, penulis ingin membuat model CNN yang dapat melakukan reidentifikasi kendaraan. Dataset yang akan digunakan pada penelitian ini diambil menggunakan UAV. Penulis berharap dengan adanya penelitian ini, penelitian tentang reidentifikasi kendaraan yang memanfaatkan UAV akan lebih banyak dipelajari oleh para peneliti. Penulis juga berharap sistem reidentifikasi kendaraan dapat diimplementasikan pada sistem pemantauan lalu lintas sehingga memudahkan staf yang bertugas. Pada penelitian ini, digunakan tiga arsitektur model CNN yang berbeda, yaitu ResNet-50, DenseNet-121, dan HRNet-18. Berdasarkan hasil penelitian, didapatkan bahwa arsitektur DenseNet-121 dengan batch size 16 menghasilkan performa terbaik dengan nilai mAP sebesar 57.74%, Rank@1 sebesar 59.87%, Rank@5 sebesar 71.77%, dan Rank@10 sebesar 77.68%.
==================================================================================================================================
In recent years, the number of motor vehicles in Indonesia has been continuously increasing. A large number of CCTV cameras are used at various street corners and intersections to assist in traffic monitoring. The monitoring system is mainly based on real-time video and still requires human supervision, making it less effective. Furthermore, the use of surveillance cameras is also less effective when used to monitor multiple locations since one camera can only cover one location point. Currently, research on CNN-based vehicle reidentification systems has been gaining traction. These systems can aid in traffic monitoring, particularly in vehicle surveillance, tracking, and identification. Concurrently, the utilization of Unmanned Aerial Vehicles (UAVs) has been on the rise in various sectors, including computer vision. In this study, the author aims to create a CNN model capable of vehicle reidentification. The dataset used for this research is obtained using UAVs. The hope is that this research will encourage more researchers to explore vehicle reidentification utilizing UAVs. Furthermore, the author aspires to implement the vehicle reidentification system into traffic monitoring, thus facilitating the work of the staff involved. Three different CNN model architectures are utilized in this study: ResNet-50, DenseNet-121, and HRNet-18. Based on the research results, DenseNet-121 with a batch size of 16 achieved the best performance with an mAP (mean Average Precision) of 57.74%, Rank@1 (top-1 accuracy) of 59.87%, Rank@5 (top-5 accuracy) of 71.77%, and Rank@10 (top-10 accuracy) of 77.68%.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | convolutional neural network, vehicle re-identification, unmanned aerial vehicle, convolutional neural network, reidentifikasi kendaraan, unmanned aerial vehicle |
Subjects: | Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis |
Depositing User: | Wildan Shafa Diandra |
Date Deposited: | 02 Aug 2023 06:46 |
Last Modified: | 02 Aug 2023 06:46 |
URI: | http://repository.its.ac.id/id/eprint/101059 |
Actions (login required)
View Item |