Research on Special Vehicle Detection and Passenger Elevator Docking Behavior Recognition in Intelligent Monitoring

Abstract:
Aiming at the problem of time node entry in the airport collaborative decision-making (airport collaborative decision-making, A-CDM) system, a method for temporal Action localization based on the results of target detection is proposed. First, the YOLOv3 target detection algorithm was selected. Since the mainstream dataset does not include airport special ground work vehicles, the data set is recreated and migration learning is performed to detect and identify airport ground targets. The bounding box and category obtained by the target detection can further obtain the on-site target object Speed and pixel coordinates, through mobile learning, near detection and recognition of airport flying targets. Features such as the interaction relationship and speed of events occurring at each moment can make judgments on airport behavior and realize real-time monitoring of the real-time screen to ensure its working process. Now we have realized the temporal action localization of the passenger elevator car docking behavior.