Statistical Management of Building Fire Equipment Based on Computer Vision and Deep Learning

The operation and maintenance (O&M) of indoor fire-fighting equipment is the key to ensure indoor fire safety. Among the O&M activities, fire-fighting equipment inspection and counting is a periodic task, and currently such task is usually performed manually, which is time-consuming and labor-intensive. In order to realize the intelligent and efficient management of indoor fire-fighting equipment, this study introduces an integrated detection and counting method of indoor fire-fighting equipment based on computer vision and deep learning, and realizes the statistics of indoor fire-fighting equipment combined with on-site video. The method combines multi-target tracking algorithm Deepsort and YOLOv5 target detector. Firstly, the data set of indoor fire-fighting equipment is established. Then, the YOLOv5 algorithm was used to extract the target boundary box and input it into the Deepsort framework. Kalman filter and Hungarian algorithm were used to predict and track the target trajectory. Finally, the tripwire counting method is used to realize the statistics of fire-fighting equipment. The actual test results show that the mean Average Precision (mAP) of the method is 92.4%, and the detection speed can reach 30 fps, which can basically realize the real-time detection of multi-class and multi-target indoor fire-fighting equipment, improving the performance of the O&M of indoor fire-fighting equipment.