Analysis of masonry work activity recognition accuracy using a spatiotemporal graph convolutional network across different camera angles

다국어 초록 (Multilingual Abstract)

Human activity recognition (HAR) in construction has gained attention for its potential to improve safety and productivity. While HAR research has shifted toward vision-based approaches, many studies typically use data from a specific angle, limiting understanding of how camera angles affect accuracy. This paper addresses this gap by using AlphaPose and Spatial-Temporal Graph Convolutional Network (ST-GCN) algorithms to analyze the impact of various camera angles on HAR accuracy in masonry work. Data was collected from seven angles (0° to 180°), with the frontal view only used for training. Results showed consistently high recognition accuracy (>80%) for side views, while accuracy decreased as the camera shifted toward rear views, especially from directly behind due to occlusion. By quantifying HAR accuracy across angles, this study provides baseline data for predicting performance from various camera positions, improving camera placement strategies and enhancing monitoring system effectiveness on construction sites. Keywords: Human Activity Recognition (HAR); Camera Angle; AlphaPose; Spatial- Temporal Graph Convolutional Network (ST-GCN)

목차 (Table of Contents)

• Ⅰ. Introduction 1
• 1.1 Introduction 1
• 1.2 Literature review 5
• 1.2.1 Vision-Based Human Activity Recognition with ST-GCN 5
• 1.2.2 Hurdles for Utilizing Computer-Vision in Construction Site 7
• 1.2.3 Needs for Investigating Ambiguity of ST-GCN in Activity Recognition 9
• Ⅱ. Methodology 12
• 2.1 Methodology 12
• 2.2 Experimental design 14
• 2.3 Data Collection and Labelling 18
• 2.4 Modification of AlphaPose and ST-GCN 19
• 2.5 Model Training and Performance Evaluation Methods 26
• Ⅲ. Results 28
• 3.1 Results 28
• 3.2 Average accuracy by angles and classes 28
• 3.3 Accuracy trends by angles and classes 31
• 3.4 Confusion matrix 35
• 3.5 Impact of Training Dataset Angle Diversity on Recognition Accuracy 41
• Ⅳ. Discussion 43
• 4.1 Impact of shooting Angle on Activity Recognition Accuracy 43
• 4.2 Symmetrical Accuracy Patterns and the Influence of Worker's Dominant Hand 44
• 4.3 Limitations, Contribution and Future Application 46
• Ⅴ. Conclusion 50
• Reference 52
• 국문 초록 65