Search

Fire poses a significant risk across industrial and domestic settings, especially to firefighters who must tackle the blaze. Current technology for detection in indoor environments are smoke detectors and flame detectors. However, these detectors have several limitations during the ignition phase of a fire and propagation. These systems cannot detect an exact position of the fire nor how the fire is spreading or its size, all of which is necessary information for fire services when dealing with these incidents. A potential solution is to use artificial intelligence techniques such as computer vision, which has shown the potential to detect and recognise objects and activities in indoor spaces. This study aims to develop a vision-based fire and smoke detection system. A deep learning technique that incorporates convolutional neural networks (CNN) was utilised to develop the real-time detection approach that can potentially provide necessary information for fire services, including identifying the position and size of the fire and how the fire spreads. A transfer learning approach using a pre-trained model was used to train the detector. Based on the detection and recognition tests using indoor fire and smoke videos, results indicated that the fire detection achieved up to 92.37% correct detections while the smoke detection did not perform as well. Hence, further improvement and evaluation of the detection approach will be conducted in future work, focusing on the impact of different parameters such as the detection model, building type, indoor space size and positioning of the detection camera. The present study provides an insight into the capabilities and potential applications of the concept.

Building ventilation accounts for up to 30% of the heat loss in commercial buildings and 25% in industrial buildings. To effectively aid the reduction of energy consumption in the building sector, the development of demand-driven control systems for heating ventilation and air-conditioning (HVAC) is necessary. In countries with temperate climates such as the UK, many buildings depend on natural ventilation strategies such as openable windows, which are useful for reducing overheating prevalence during the summer. The manual opening and adjustment of windows by occupants, particularly during the heating season, can lead to substantial heat loss and consequent energy consumption. This could also result in the unnecessary or over ventilation of the space, or the fresh air is more than what is required to ensure adequate air quality. Furthermore, energy losses build up when windows are left open for extended periods. Hence, it is important to develop control strategies that can detect and recognise the period and amount of window opening in real-time and at the same time adjust the HVAC systems to minimise energy wastage and maintain indoor environment quality and thermal comfort. This paper presents a vision-based deep learning framework for the detection and recognition of manual window operation in buildings. A trained deep learning model is deployed into an artificial intelligence-powered camera. To assess the proposed strategy's capabilities, building energy simulation was used with various operation profiles of the opening of the windows based on various scenarios. Initial experimental tests were conducted within a university lecture room with a south-facing window. Deep learning influenced profile (DLIP) was generated via the framework, which uses real-time window detection and recognition data. The generated DLIP were compared with the actual observations, and the initial detection results showed that the method was capable of identifying windows that were opened and had an average accuracy of 97.29%. The results for the three scenarios showed that the proposed strategy could potentially be used to help adjust the HVAC setpoint or alert the occupants or building managers to prevent unnecessary heating demand. Further developments include enhancing the framework ability to detect multiple window opening types and sizes and the detection accuracy by optimising the model.