DOT DETECTION ON BRAILLE IMAGE USING CIRCLE HOUGH TRANSFORM WITH SAUVOLA THRESHOLDING

Prihantono, Silvanus (2026) DOT DETECTION ON BRAILLE IMAGE USING CIRCLE HOUGH TRANSFORM WITH SAUVOLA THRESHOLDING. Undergraduate thesis, UPN Veteran Jawa Timur..

	Text (Cover) 20081010223 - Cover.pdf Download (2MB)
	Text (Bab 1) 20081010223 - Bab 1.pdf Download (132kB)
	Text (Bab 2) 20081010223 - Bab 2.pdf Restricted to Repository staff only until 15 September 2028. Download (250kB) \| Request a copy
	Text (Bab 3) 20081010223 - Bab 3.pdf Restricted to Repository staff only until 15 September 2028. Download (293kB) \| Request a copy
	Text (Bab 4) 20081010223 - Bab 4.pdf Restricted to Repository staff only until 15 September 2028. Download (3MB) \| Request a copy
	Text (Bab 5) 20081010223 - Bab 5.pdf Download (109kB)
	Text (Daftar Pustaka) 20081010223 - Daftar Pustaka Bibligraphy.pdf Download (146kB)
	Text (Lampiran) 20081010223 - Lampiran (Turnitin Bebas Plagiasi) .pdf Restricted to Repository staff only until 15 September 2028. Download (88kB) \| Request a copy

Abstract

This research investigates parameter sensitivity of Braille dot detection pipeline combining Sauvola adaptive thresholding and Circle Hough Transform applied to a selected Braille document image dataset. Automated detection of Braille dots presents a unique challenge in document image processing due to low-relief embossed nature of Braille cells and the susceptibility of captured images to local illumination variation and surface texture irregularities making choice of binarization and detection parameters a critical determinant of system performance. Proposed pipeline processes each image through Sauvola adaptive thresholding as a binarization stage, whose output is passed directly to Circle Hough Transform for circular dot detection. To identify optimal configuration for each stage, a joint full Cartesian grid search was conducted over Sauvola parameters window_size and k with all evaluated parameter combinations assessed against ground truth annotations using Precision, Recall, and mean F1-Score across all images in dataset. Results are visualized as two-dimensional heatmaps and individual sensitivity curves for each parameter. Sauvola grid search identified an optimal configuration of window_size=13 and k=0.049, achieving a mean Precision of 0.8047, mean Recall of 0.7809, and mean F1-Score of 0.7925. For the Circle Hough Transform, internal Canny threshold parameter param1 was empirically validated to have no meaningful influence on detection performance when applied to pre-binarized images and was fixed at 1. Accumulator threshold param2 was identified as sole variable of interest, with an optimal value of 3 producing the same peak mean F1-Score of 0.7925. These findings confirm that Sauvola adaptive thresholding combined with Circle Hough Transform constitutes an effective and quantitatively validated approach to Braille dot detection with optimal parameter configurations and their performance boundaries documented through heatmap and sensitivity curve visualizations as a practical reference for datasets sharing similar image characteristics.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Anggraeny, Fetty Tri	NIDN0711028201	fettyanggraeny.if@upnjatim.ac.id
Thesis advisor	Sihananto, Andreas Nugroho	NIDN0012049005	andreas.nugroho.jarkom@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA75 Electronic computers. Computer science

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Silvanus Prihantono

Date Deposited:

15 Sep 2026 07:02

Last Modified:

15 Sep 2026 07:07

URI:

https://repository.upnjatim.ac.id/id/eprint/60223