OBJECT DETECTION AND HUMAN SPERMATOZOA MOVEMENT TRACKING IN VIDEO USING YOLOV5 AND STRONGSORT ALGORITHM

Halim, Christina (2025) OBJECT DETECTION AND HUMAN SPERMATOZOA MOVEMENT TRACKING IN VIDEO USING YOLOV5 AND STRONGSORT ALGORITHM. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 21083010124_Cover.pdf Download (1MB)
	Text (Bab 1) 21083010124_Bab 1.pdf Download (439kB)
	Text (Bab 2) 21083010124_Bab 2.pdf Restricted to Repository staff only until 5 December 2027. Download (1MB)
	Text (Bab 3) 21083010124_Bab 3.pdf Restricted to Repository staff only until 5 December 2027. Download (790kB)
	Text (Bab 4) 21083010124_Bab 4.pdf Restricted to Repository staff only until 5 December 2027. Download (1MB)
	Text (Bab 5) 21083010124_Bab 5.pdf Download (422kB)
	Text (Daftar Pustaka) 21083010124_Daftar Pustaka.pdf Download (345kB)
	Text (Lampiran) 21083010124_Daftar Lampiran.pdf Restricted to Repository staff only until 5 December 2027. Download (455kB)

Abstract

Infertility is a global health issue, with approximately 50% of cases attributed to male factors. Sperm quality analysis, particularly the assessment of motility parameters, is essential for evaluating male infertility. However, manual analysis is time-consuming, prone to human error, and often inconsistent. Automated systems such as CASA can assist, but they are costly and still limited in multi-sperm tracking capabilities. This study proposes a deep learning–based track-by detection approach using YOLOv5 as the detector for identifying spermatozoa and StrongSORT as the tracker for monitoring sperm movement. The tracking performance was evaluated using two integration methods: YOLOv5 + Kalman Filter and YOLOv5 + StrongSORT. Experimental results show that the YOLOv5 + Kalman Filter method yields better performance, achieving a MOTA of 56.7%, an IDF1 of 45.4%, a MOTP of 0.891, and a processing speed of 5 FPS. In comparison, YOLOv5 + StrongSORT achieved a MOTA of 50.9%, an IDF1 of 43.1%, a MOTP of 0.887, and a processing speed of 1 FPS. These findings indicate that the YOLOv5-based track-by-detection approach integrated with the Kalman Filter provides more stable, faster, and more accurate detection and tracking results than StrongSORT. This is because the Kalman Filter involves lighter computation, does not require re-identification feature extraction, and can efficiently update object position estimates at every frame.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Saputra, Wahyu Syaifullah Jauharis	NIDN0725088601	wahyu.s.j.saputra.if@upnjatim.ac.id
Thesis advisor	Hindrayani, Kartika Maulida	NIDN0009099205	kartika.maulida.ds@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA76.87 Neural computers
Q Science > QH Natural history > QH301 Biology

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Christina Halim

Date Deposited:

05 Dec 2025 08:27

Last Modified:

05 Dec 2025 08:51

URI:

https://repository.upnjatim.ac.id/id/eprint/47772