Dynamic Decision Threshold Fire Warning Based on Neyman-Pearson Criterion

Abstract

In the problem of selecting the fire warning device discrimination threshold, we designed a dynamic judgment threshold fire warning device combining the YOLO human flow statistical monitoring system and the Neyman-Pearson criterion. The traditional fire early warning device is to make a choice between the leakage alarm rate and the false alarm rate, sacrificing the performance of one party in exchange for the better performance of the other party. After the threshold selection, the dynamic change cannot be realized according to the demand situation. However, most fire events are caused by human factors. Therefore, we combined with the YOLO target tracking and monitoring system to dynamically adjust the judgment threshold of the fire early warning device by monitoring the flow of people. In the period of large flow of people, the judgment threshold of fire warning is reduced, and the minimum leakage alarm rate is selected. In the period of small flow of people, the judgment threshold of fire warning device is improved and the minimum false alarm rate is selected. And use the Neyman-Pearson criterion to achieve the ideal lowest leakage alarm rate and false alarm rate under the demand condition, and improve the accuracy of fire prevention. Through this improvement, the problem of too rigid threshold of fire warning device is solved, better adapted to the practical application needs, and provides a more optimized solution for balancing the protection of the safety of people's lives and property and reducing the waste of human and material resources.