Multi-modal scene perception and data collection method based on smart city

The invention relates to the technical field of multi-modal scene perception, in particular to a multi-modal scene perception and data collection method based on a smart city. The method comprises the following steps: S1, data collection: collecting data of different modes of each key area in a monitored city in real time, and transmitting the collected data to a cloud platform through an Internet of Things technology; s2, data fusion: integrating the collected multi-modal data through a data fusion algorithm; key information is extracted from fused data through computer vision, sound and machine learning technologies and is intelligently analyzed and perceived, so that the reliability and accuracy of the data are improved, and the perceived data are shared and opened to urban managers, research institutions and citizens through data anonymization, encrypted transmission and access control technologies. Therefore, real-time city information and service are provided, and decision making, traffic management and