Tracking Pemain Sepak bola Pada Video Menggunakan Metode Deep Learning dan Multi Object Tracking

Zein, Bilal Khabibullah (2024) Tracking Pemain Sepak bola Pada Video Menggunakan Metode Deep Learning dan Multi Object Tracking. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022221014-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (16MB) | Request a copy

Abstract

Pelacakan adalah prosedur yang ditetapkan yang memerlukan penetapan identifikasi ke objek tertentu dan kemudian secara konsisten mengenali objek tersebut tanpa mengubah identifikasi yang ditetapkan melalui urutan gambar bingkai dan mengasosiasikannya dengan tepat Saat melakukan penelitian tentang pelacakan objek, terutama dalam olahraga di mana objek yang diminati adalah manusia, diperlukan teknologi yang tangguh untuk memfasilitasi proses pelacakan. Ketika pendekatan deteksi objek yang canggih, YOLOV8, digabungkan dengan algoritma DeepSORT, maka diharapkan dapat menghasilkan hasil yang sangat akurat dan tepat dalam pelacakan dan pendeteksian objek. Tantangan dalam pelacakan multi-objek termasuk ketahanan, oklusi, dan pergeseran identitas. Dalam penelitian kami, kami memanfaatkan perpaduan algoritma YOLOV8 dan DeepSORT untuk mencapai solusi pelacakan yang sangat andal dan tepat. Implementasi prediksi gerakan berbasis filter Kalman di DeepSORT memungkinkan pencapaian lintasan yang mulus, sedangkan jaringan saraf dalam YOLOV8 yang digunakan membantu dalam mengenali kemunculan objek di lapangan secara tepat. Hasil percobaan kami menunjukkan pelacakan yang kami dapatkan adalah 38% HOTA, 47% DetA, 31% AssA, 68% DetPre, 35% AssRE, 61% AssPr dan 79% LOcA.
=================================================================================================================================
Tracking is a set procedure that entails assigning an identification to a certain object and subsequently consistently recognizing that object without altering the assigned identification over a sequence of frame images and associating it accordingly.When performing research on object tracking, especially in sports where the object of interest is a human, a resilient technology is necessary to facilitate the tracking process. When the state-of-the-art object detection approach, YOLOV8, is combined with the DeepSORT algorithm, it is anticipated to produce highly accurate and exact outcomes in the tracking and detection of objects. Challenges in multi-object tracking include robustness, oculusion, and identity shifts. In our research, we take advantage of a fusion of YOLOV8 and DeepSORT algorithms to achieve a highly reliable and precise tracking solution. The implementation of the Kalman filter-based motion prediction in DeepSORT allows for the achievement of smooth trajectories, whereas the YOLOV8 deep neural network used assists in precisely recognizing the appearance of objects on the field. The result of our experiment shown the tracking we get is 38% HOTA, 47% DetA, 31% AssA, 68% DetPre, 35% AssRE, 61% AssPr amd 79% LOcA.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	YOLO, MOT, CNN, HOTA, Deep Learning, bola, pemain, wasit, kiper
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > Q Science (General) > Q337.5 Pattern recognition systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Zein Bilal Khabibullah
Date Deposited:	26 Jul 2024 02:02
Last Modified:	26 Jul 2024 02:02
URI:	http://repository.its.ac.id/id/eprint/108991

Actions (login required)

View Item