Peningkatan Sistem Navigasi Kursi Roda Berbasis Computer Vision Dan Informasi Sensor Bagi Penyandang Gangguan Penglihatan

Noor, Raiihan Aria Muhamad (2025) Peningkatan Sistem Navigasi Kursi Roda Berbasis Computer Vision Dan Informasi Sensor Bagi Penyandang Gangguan Penglihatan. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023211036-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Mobilitas mandiri bagi penyandang disabilitas netra masih menghadapi banyak tantangan, terutama dalam mengenali rute dan menyadari keberadaan orang di sekitarnya saat berada di ruang publik. Penelitian ini bertujuan untuk mengembangkan sistem pendeteksi paving taktil dan deteksi orang secara otomatis berbasis model YOLOv5 yang diterapkan pada kursi roda pintar. Sistem ini menggunakan dua kamera: kamera bawah untuk mendeteksi empat pola paving taktil (lurus, belok, pertigaan, dan perempatan), serta kamera belakang untuk mendeteksi keberadaan orang di sisi kiri atau kanan kursi roda. Pemrosesan dilakukan pada perangkat Jetson Xavier dengan pendekatan multiprocessing untuk menjalankan dua model deteksi secara paralel di core CPU yang berbeda. Deteksi orang dioptimalkan melalui akselerasi GPU CUDA, dan sistem menggunakan Queue dari library multiprocessing Python sebagai perantara komunikasi antar proses. Output sistem dikirim ke mikrokontroler Arduino yang mengaktifkan motor getar sebagai umpan balik kepada pengguna: angka 1–4 untuk klasifikasi paving taktil, dan 5–6 untuk posisi orang. Sistem berjalan secara real-time dengan kecepatan rata-rata 10,24 FPS pada deteksi paving taktil dan 7,56 FPS pada deteksi orang. Rata-rata latensi pengiriman data citra ke sistem algoritma tercatat sebesar 0,032 detik untuk paving taktil dan 0,030 detik untuk deteksi orang. Selain itu, evaluasi menggunakan metode NASA-TLX menunjukkan nilai beban kerja total rata-rata 51,9, yang tergolong sedang dan masih dapat diterima. Hasil ini menunjukkan bahwa sistem memberikan bantuan yang signifikan dalam meningkatkan kesadaran pengguna terhadap kondisi sekitar tanpa membebani secara kognitif. Dengan integrasi antara visi komputer, pemrosesan paralel, dan sistem umpan balik, sistem ini mampu meningkatkan keamanan dan kenyamanan penyandang tunanetra dalam menggunakan kursi roda pintar secara mandiri, serta memberikan kontribusi nyata dalam pengembangan teknologi bantu yang adaptif.
=======================================================================================================================================
Independent mobility for individuals with visual impairments remains a significant challenge, particularly in recognizing navigation routes and being aware of people in their surroundings in public spaces. This study aims to develop an automatic tactile paving and human detection system based on the YOLOv5 model, implemented on a smart wheelchair. The system utilizes two cameras: a bottom facing camera to detect four tactile paving patterns (straight, turn, T-junction, and crossroad), and a rear-facing camera to detect the presence of people on the left or right side of the wheelchair. Processing is performed on a Jetson Xavier device using a multiprocessing approach to run both detection models in parallel on separate CPU cores. Human detection is optimized through GPU acceleration with CUDA, and the system uses the Python multiprocessing library’s Queue as an inter process communication bridge. The system output is sent to an Arduino microcontroller, which activates a vibration motor as user feedback: numbers 1–4 indicate the tactile paving classification, while 5–6 indicate the position of a detected person. The system runs in real-time with an average speed of 10,24 FPS for tactile paving detection and 7,56 FPS for human detection. The average image data transfer latency to the algorithm system was recorded at 0.032 seconds for tactile paving and 0.030 seconds for human detection. Furthermore, user evaluation using the NASA-TLX method yielded an average workload score of 52.9, categorized as moderate and acceptable. These results demonstrate that the system provides significant assistance in enhancing users' environmental awareness without causing cognitive overload. Through the integration of computer vision, parallel processing, and feedback systems, this smart wheelchair system improves the safety and comfort of users with visual impairments and offers a meaningful contribution to the development of adaptive assistive technologies

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kursi roda pintar, YOLOv5, paving taktil, multiprocessing, Jetson Xavier Smart wheelchair, YOLOv5, tactile paving, multiprocessing, Jetson Xavier
Subjects:	T Technology > T Technology (General) > T58.62 Decision support systems T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7878 Electronic instruments
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Raihan Aria Muhamad Noor
Date Deposited:	04 Aug 2025 01:50
Last Modified:	04 Aug 2025 01:50
URI:	http://repository.its.ac.id/id/eprint/126405

Actions (login required)

View Item