Research on grid based fire warning algorithm with YOLOv5s for palace buildings

Abstract

In response to the early warning requirements of fire security technology in the Imperial Palace & large Ming and Qing ancient architectural complexes in China, a grid based fire warning algorithm is proposed by combining neural network YOLOv5s smoke detection technology. In this algorithm, the inverse proportional gridding algorithm based on building density is used to optimize the grid of buildings, and compared with the results of the equidistant grid algorithm, the risk distribution division is more detailed and reasonable. The smoke detection part uses YOLOv5s based smoke detection technology to detect the distribution of fire smoke in various areas, and the positioning of this area in the overall grid realized by the remote transmission module. With detection experiments on relevant datasets, the results show that its accuracy and mAP both reach 0.99. By utilizing the collaborative effect of inverse proportional gridding algorithm and smoke detection technology, a grid based visualization of smoke warning is achieved.

Keywords

• fire warning
• grid based algorithm
• smoke detection
• yolov5s