Sistem Kendali Kursi Roda Berbasis Computer Vision Menggunakan Head Pose Dengan Adaptasi Occlusion Face Pada Jetson Xavier NX

Adawiyah, Robi'ah (2026) Sistem Kendali Kursi Roda Berbasis Computer Vision Menggunakan Head Pose Dengan Adaptasi Occlusion Face Pada Jetson Xavier NX. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5023221038-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Spinal Cord Injury (SCI), khususnya quadriplegia, merupakan kondisi yang menyebabkan kelumpuhan pada tubuh bagian atas dan bawah sehingga menyebabkan penderita mengalami keterbatasan dalam melakukan mobilitas secara mandiri. Sementara itu, penggunaan kursi roda dengan kontrol joystick tidak dapat digunakan. Beberapa metode kendali alternatif telah dikembangkan menggunakan sinyal biologis tubuh, tetapi umumnya memerlukan sensor yang dipasang pada pengguna sehingga dapat mengurangi kenyamanan selama penggunaan. Oleh karena itu, penelitian ini mengembangkan sistem kendali kursi roda berbasis computer vision yang memanfaatkan gerakan kepala sebagai masukan navigasi karena bersifat non-kontak dan hands-free. Sistem menggunakan kamera untuk menangkap citra wajah secara real-time. Deteksi wajah dilakukan menggunakan YOLOv8n-face untuk memperoleh area wajah, kemudian hasil deteksi diproses melalui tahap image enhancement berupa Contrast-Limited Adaptive Histogram Equalization (CLAHE) dan gamma correction sebelum diestimasi menggunakan 6DRepNet untuk memperoleh sudut Euler (yaw, pitch, dan roll). Sudut tersebut selanjutnya diterjemahkan menjadi delapan perintah navigasi melalui mekanisme one-tap control. Sistem diimplementasikan pada Jetson Xavier NX dan dilengkapi mikrokontroler STM32F103C8T6 sebagai pengendali motor serta lima sensor ultrasonik HC-SR04 sebagai sistem keselamatan untuk mendeteksi rintangan dan menghentikan kursi roda secara otomatis. Hasil pengujian menunjukkan bahwa sistem mampu mendeteksi wajah pada berbagai kondisi pencahayaan, baik di dalam maupun di luar ruangan, serta pada kondisi wajah tanpa oklusi, menggunakan kacamata, dan menggunakan masker. Pengujian keseluruhan sistem menghasilkan tingkat keberhasilan rata-rata sebesar 86.2% pada rute tanpa rintangan dan 86.9% pada rute dengan rintangan, dengan tingkat keberhasilan 89.0% pada oklusi kacamata dan 84.1% pada oklusi masker. Sistem pendeteksian rintangan juga berhasil mencegah tabrakan pada seluruh percobaan dengan tingkat keberhasilan 95%. Selain itu, hasil evaluasi NASA-TLX menunjukkan skor rata-rata sebesar 37.7 yang termasuk dalam kategori beban kerja ringan. Hasil tersebut menunjukkan bahwa sistem yang dikembangkan memiliki potensi sebagai solusi kendali kursi roda hands-free yang aman, nyaman, dan sesuai untuk quadriplegia.
=====================================================================================================================================
Spinal Cord Injury (SCI), particularly quadriplegia, is a condition that causes paralysis of both the upper and lower limbs, resulting in limited independent mobility. Conventional wheelchairs controlled by a joystick are often unsuitable for individuals with this condition. Several alternative control methods based on biological signals have been developed; however, most require sensors to be attached to the user's body, which may reduce comfort during operation. Therefore, this study develops a computer vision-based wheelchair control system that utilizes head movements as a non-contact and hands-free navigation input. The system uses a camera to capture facial images in real time. Face detection is performed using YOLOv8n-face to obtain the facial region, after which the detected face undergoes an image enhancement stage using Contrast-Limited Adaptive Histogram Equalization (CLAHE) and gamma correction before being processed by 6DRepNet to estimate the euler angles (yaw, pitch, roll). These angles are then translated into eight navigation commands through a one-tap control mechanism. The system is implemented on the Jetson Xavier NX platform and integrates an STM32F103C8T6 microcontroller for motor control and five HC-SR04 ultrasonic sensors as a safety mechanism to detect obstacles and automatically stop the wheelchair. Experimental results demonstrate that the system is capable of detecting faces under various lighting conditions, both indoors and outdoors, as well as with no occlusion, eyeglasses, and face masks.The overall system achieved average success rates of 86.2% on obstacle-free routes and 86.9% on routes with obstacles. Under facial occlusion conditions, the success rates reached 89.0% for eyeglasses and 84.1% for face masks. The obstacle detection system also successfully prevented collisions in all experimental trials, achieving a 95% success rate. Furthermore, NASA-TLX evaluation yielded an average score of 37.7, indicating a low mental workload. These results suggest that the proposed system has the potential to serve as a safe, comfortable, and hands-free wheelchair control solution for individuals with quadriplegia.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Head Pose Estimation, Jetson Xavier NX, Kontrol Kursi Roda, Occluded Face, Quadriplegia Head Pose Estimation, Jetson Xavier NX, Wheelchair Control, Occluded Face, Quadriplegia.
Subjects:	R Medicine > RM Therapeutics. Pharmacology > RM950 Rehabilitation technology.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Robi'ah Adawiyah
Date Deposited:	01 Aug 2026 02:10
Last Modified:	01 Aug 2026 02:10
URI:	http://repository.its.ac.id/id/eprint/139038

Actions (login required)

View Item