Sensor Fusion and Convolutional Neural Networks for Indoor Occupancy Prediction Using Multiple Low-Cost...

Area of Science:

• Building Science
• Artificial Intelligence
• Sensor Technology

Background:

• Indoor occupancy prediction is crucial for smart building systems like energy management, security, and health.
• Automated systems considering occupancy can reduce building energy consumption by over 50%.
• High-resolution sensors and cameras pose privacy concerns for occupancy prediction.

Purpose of the Study:

• To propose a novel, privacy-preserving solution for indoor occupancy prediction.
• To utilize multiple low-cost, low-resolution heat sensors for occupancy estimation.
• To evaluate data fusion techniques and Convolutional Neural Network (CNN) performance for this task.

Main Methods:

• Development of two distinct data fusion and processing methods for heat sensor data.
• Implementation of a Convolutional Neural Network (CNN) model for occupancy prediction.
• Experimental assessment of prediction accuracy and the impact of sensor field-of-view overlap.

Main Results:

• The proposed solutions demonstrate high accuracy in indoor occupancy prediction.
• Real-time processing capabilities were achieved with the implemented methods.
• Analysis provided insights into the effect of sensor overlap on prediction performance.

Conclusions:

• The novel approach effectively predicts indoor occupancy using low-cost heat sensors, addressing privacy concerns.
• The fusion methods and CNN model offer a viable, accurate, and efficient solution for smart buildings.
• This technology supports enhanced building management through reliable, real-time occupancy data.