| CPC G06Q 50/265 (2013.01) [B64C 39/024 (2013.01); G06V 20/52 (2022.01); G08B 21/02 (2013.01); B64U 2101/30 (2023.01)] | 6 Claims |
|
1. An early warning method for a regional public security management in a smart city based on an Internet of Things (IoT), wherein the IoT includes a user platform, a service platform, a public security management platform, a sensing network platform and object platforms interacted in turn, the early warning method comprising:
obtaining, based on the user platform, a user's inquiry instruction of each region, and sending the inquiry instruction to the public security management platform through the service platform;
in response to the inquiry instruction, obtaining, from at least one monitoring device of at least one target region based on a sensing network sub-platform of the sensing network platform, a monitoring image of the at least one target region by the public security management platform; the at least one monitoring device being configured in the different object platforms, wherein the sensing network platform uses different sensing network sub-platforms to store, process and/or transmit data from different object platforms, and the sensing network sub-platforms correspond to different target regions; the public security management platform uses different management sub-platforms for the data storage, the data processing and/or the data transmission, and performs a data summarizing, the data processing and the data transmission through a general database of the public security management platform;
sending, based on the sensing network sub-platform, the monitoring image of the corresponding target region to the management sub-platform;
processing, based on the management sub-platform, the monitoring image to determine a risk index of the at least one target region, wherein the determining the risk index of the at least one target region based on the monitoring image comprises:
determining a suspicious index of each of one or more persons in the monitoring image, wherein the suspicious index grows with an increase of time of the at least one person staying in the at least one target region, a growth rate of the suspicious index is related to a region type of a region where a suspicious person stays and a track suspicion, and the track suspicion is determined through the following operations:
obtaining an action track of each person in the monitoring image;
extracting a track feature based on the action track, wherein the track feature is expressed by a track diagram, a node of the track diagram-corresponding to each location, an attribute of the node including times when person appears and leaves, an edge of the track diagram being a one-way edge, a direction of the edge indicating that the person goes from one place to another place, and an attribute of the edge including a number of times of the person goes from the one place to the another place; and
by processing the track diagram based on a track suspicion determination model, determining the track suspicion, wherein the track suspicion determination model is a graph neural network model, the track suspicion is output based on a node corresponding to a last place where the person stays;
determining one or more suspicious persons in response to the determination that the suspicious index meets a preset condition; and
determining, based on the one or more suspicious persons, the risk index of the target region, wherein the determining the risk index of the target region based on the one or more suspicious persons comprises:
determining, based on monitoring images of a plurality of adjacent frames, a distance between the one or more suspicious persons;
determining, based on the distance, a suspicious group to which the one or more suspicious persons belong; and
determining, based on the suspicious group, the risk index of the target region, wherein the determining the risk index of the target region based on the suspicious group comprises:
determining, based on a sum of suspicious group indexes of suspicious groups in the target region and an amplification factor, the risk index of the target region, wherein a suspicious group index is a product of a sum of the suspicious index of each suspicious person in the group and the amplification factor;
generating early warning information in response to the risk index of the target region greater than a first threshold, and sending the early warning information to the user platform through a general database of the management platform and the service platform; and
obtaining a management instruction based on the user platform, and controlling, according to the management instruction, an unmanned aerial vehicle (UAV) to go to the target region for monitoring.
|