Challa Koundinya, AlHmoud Issa W, Jaiswal Chandra, Gokaraju Balakrishna, Tesireo Raymond
North Carolina A&T State University, 1601 E Market St, Greensboro, North Carolina, 27411, 3363347, USA.
MethodsX. 2025 May 28;14:103386. doi: 10.1016/j.mex.2025.103386. eCollection 2025 Jun.
This study introduces an advanced methodology for optimizing HVAC efficiency through real-time classroom occupancy detection by combining video analysis with gas sensor data to enhance accuracy and reliability. The proposed system integrates video feeds captured by a Logitech C20 webcam with data from an MS1100 gas sensor module, ensuring a dual-modal approach to occupancy detection. A YOLOv4 object detection model, trained on a diverse dataset of over 20,000 labeled human face images, achieves over 98 % accuracy in identifying and counting occupants in real time. OpenCV is employed to facilitate efficient and seamless processing of video streams, enabling the system to deliver real-time results crucial for dynamic HVAC control. The integration of gas sensor data addresses scenarios where environmental factors, such as low light or obstructions, could impair video analysis, thereby improving detection reliability under diverse conditions. The combination of these modalities provides a robust and adaptable framework for occupancy detection, which can be scaled for different building types and configurations. This method demonstrates significant potential in reducing energy consumption and enhancing the sustainability of building management systems by providing precise occupancy data for HVAC optimization. The approach offers a practical and scalable solution for the growing demand for energy-efficient infrastructure in smart buildings. The architecture ensures seamless integration between visual and environmental sensing modalities, enhancing real-time responsiveness and occupancy detection reliability.•Utilizes a YOLOv4 object detection model and MS1100 gas sensor for real-time occupancy detection.•Achieves over 98 % accuracy with a dataset of over 20,000 labeled human face images.•Offers a scalable and efficient solution for energy-efficient HVAC systems.
本研究介绍了一种先进的方法，通过将视频分析与气体传感器数据相结合进行实时教室占用检测，以优化暖通空调（HVAC）效率，从而提高准确性和可靠性。所提出的系统将罗技C20网络摄像头捕获的视频流与MS1100气体传感器模块的数据集成在一起，确保采用双模式方法进行占用检测。一个在超过20000张带标签的人脸图像的多样化数据集上训练的YOLOv4目标检测模型，在实时识别和统计占用者方面的准确率超过98%。采用OpenCV来促进视频流的高效无缝处理，使系统能够提供对动态HVAC控制至关重要的实时结果。气体传感器数据的集成解决了环境因素（如低光照或障碍物）可能影响视频分析的情况，从而提高了在各种条件下的检测可靠性。这些模式的结合为占用检测提供了一个强大且适应性强的框架，可针对不同的建筑类型和配置进行扩展。该方法通过为HVAC优化提供精确的占用数据，在降低能耗和提高建筑管理系统的可持续性方面显示出巨大潜力。该方法为智能建筑中对节能基础设施不断增长的需求提供了一个实用且可扩展的解决方案。该架构确保了视觉和环境传感模式之间的无缝集成，提高了实时响应能力和占用检测可靠性。
•利用YOLOv4目标检测模型和MS1100气体传感器进行实时占用检测。
•在超过20000张带标签的人脸图像数据集上实现了超过98%的准确率。
•为节能HVAC系统提供了一个可扩展且高效的解决方案。
