Sistem Deteksi Postur Duduk Ergonomis Berdasarkan Video Menggunakan Openpose Dan Bi-LSTM

Chandra, Rhenaldy (2026) Sistem Deteksi Postur Duduk Ergonomis Berdasarkan Video Menggunakan Openpose Dan Bi-LSTM. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221190-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Gangguan muskuloskeletal (Musculoskeletal Disorders/MSDs) akibat postur duduk yang tidak ergonomis merupakan masalah kesehatan kerja yang signifikan, terutama bagi individu yang menghabiskan waktu lama di depan komputer. Metode penilaian ergonomi konvensional seperti Rapid Upper Limb Assessment (RULA) seringkali dilakukan secara manual atau terbatas pada analisis gambar statis (frame-by-frame), sehingga gagal menangkap dinamika pergerakan tubuh dan durasi pembebanan otot secara akurat. Penelitian ini mengusulkan pengembangan sistem deteksi postur duduk ergonomis otomatis berbasis video yang mengintegrasikan Computer Vision dan Deep Learning. Sistem menggunakan teknologi estimasi pose tubuh OpenPose untuk mengekstraksi 25 titik persendian tubuh (keypoints), yang kemudian ditransformasikan melalui proses joint angle estimation menjadi 14 fitur biomekanik yang sesuai dengan parameter penilaian RULA. Fitur-fitur tersebut terdiri atas sudut persendian, parameter penyesuaian (adjustment) RULA, serta Muscle Use Score yang merepresentasikan karakteristik spasial dan temporal postur kerja berdasarkan deteksi postur statis dan gerakan repetitif. Prediksi tingkat risiko ergonomis dilakukan menggunakan model Bidirectional Long Short-Term Memory (Bi-LSTM) yang mampu mempelajari pola temporal dari urutan gerakan secara dua arah. Pendekatan ini memungkinkan sistem untuk memberikan penilaian risiko RULA (Level 1 - 7) yang lebih stabil dan representatif dibandingkan metode klasifikasi statis, dibuktikan dengan tingkat presisi Mean Absolute Error (MAE) sebesar 0,0433 dan R² 0,9329. Ketangguhan sistem juga tervalidasi pada pengujian dunia nyata dan dataset augmentasi dengan perolehan MAE masing-masing 0,0335 dan 0,0533 dengan R² 0.9539 dan 0,9049. sehingga efektif digunakan sebagai alat bantu pemantauan kesehatan kerja, sehingga dapat digunakan sebagai alat bantu pemantauan kesehatan kerja yang efektif.
=========================================================================================================================================
Musculoskeletal Disorders (MSDs) caused by non-ergonomic sitting postures represent a significant occupational health issue, particularly for individuals who spend prolonged periods working in front of computers. Conventional ergonomic assessment methods, such as the Rapid Upper Limb Assessment (RULA), are generally performed manually or limited to static image (frame-by-frame) analysis, making them incapable of accurately capturing body movement dynamics and the duration of muscle loading. This study proposes the development of an automated video-based ergonomic sitting posture assessment system by integrating Computer Vision and Deep Learning. The system employs the OpenPose pose estimation framework to extract 25 body keypoints, which are subsequently transformed through a joint angle estimation process into 14 biomechanical features corresponding to the RULA assessment parameters. These features consist of joint angle measurements, RULA adjustment parameters, and Muscle Use Score features that represent both the spatial and temporal characteristics of working postures based on static posture detection and repetitive movements. Ergonomic risk prediction is performed using a Bidirectional Long Short-Term Memory (Bi-LSTM) model capable of learning bidirectional temporal patterns from movement sequences. This approach enables the system to provide more stable and representative RULA risk assessments (Levels 1–7) than conventional static assessment methods, achieving a Mean Absolute Error (MAE) of 0.0433 and a coefficient of determination (R²) of 0.9329. The proposed system was further validated using real-world recordings and augmented datasets, achieving MAE values of 0.0335 and 0.0533, with corresponding R² values of 0.9539 and 0.9049, respectively. These results demonstrate that the proposed system possesses good generalization capability and has the potential to serve as an effective tool for automated occupational ergonomic risk assessment.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Bi-LSTM, Ergonomi, Muscle Use Score, OpenPose, RULA, Sistem Deteksi Postur, Bi-LSTM, Ergonomics, Muscle Use Score, OpenPose, Posture Detection System, RULA
Subjects:	Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > T Technology (General) > T57.5 Data Processing T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Rhenaldy Chandra
Date Deposited:	27 Jul 2026 04:05
Last Modified:	27 Jul 2026 04:05
URI:	http://repository.its.ac.id/id/eprint/137836

Actions (login required)

View Item