DEVELOPMENT AND TESTING OF AN INTELLIGENT MONITORING SYSTEM FOR CONSTRUCTION PROCESSES BASED ON COMPUTER VISION AND MACHINE LEARNING

Abstract

This paper presents the design, development, and testing of an intelligent construction monitoring system that leverages computer vision and machine learning techniques to enhance the efficiency and safety of construction processes. The primary focus is on the automated detection of personal protective equipment (PPE) and other safety-related objects at construction sites. To analyze video streams, neural network models YOLOv11 and ResNet-50 were employed, enabling efficient object detection and classification of potential safety violations. The Lightning AI platform was utilized for automated data processing and cloud computing. Testing the system on real-world data demonstrated a detection accuracy of 70–75% (mAP@0.5) for workers and PPE, while the classification accuracy for safety violations reached 92%. Under stricter evaluation criteria (mAP@0.5:0.95), accuracy decreased to 35%, indicating the need for model improvements in complex scenes and cases with partially occluded objects. Error analysis revealed that the most common issues included incorrect recognition of helmets and safety glasses, particularly in scenarios involving object occlusion and low image quality. The proposed intelligent monitoring system enhances the timely detection of safety violations and reduces human error, thereby contributing to lower accident rates and improved construction process management in real-world operational conditions.