Multi-class multi-scale multi-target snapshot method and system

Abstract

The invention discloses a multi-class multi-scale multi-target snapshot method and system, and belongs to the technical field of artificial intelligence and computer vision, and the method comprises the following steps: obtaining a panoramic video processing frame from a miscellaneous scene; performing intelligent real-time multi-category multi-scale multi-target detection; carrying out on-line multi-target tracking category by category; carrying out deduplication and preferential selection on the snapshots; and transmitting a snapshot result to a server or a data center. According to the invention, on one front-end camera, real-time detection and analysis can be carried out on videos with multiple categories, multiple scales and multiple targets, and target object snapshot results which are classified, high in image quality and low in repetition and redundancy can be efficiently obtained.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and computer vision, and in particular relates to a multi-category, multi-scale and multi-object capture method and system. Background technique [0002] The traditional way of security monitoring with human naked eyes, or the way of sending the monitoring screen back to the server can no longer "digest" the continuous mass video surveillance data in time, and the demand for intelligent target capture is becoming more and more urgent. People have taken some measures to improve traditional security monitoring technology: [0003] In the patent CN201911235029.7, a target capture method, device and system are disclosed, which adopts the method of combining the bolt camera and the dome camera, and carries out the target feature of the first monitoring screen of the bolt camera and the second monitoring screen of the dome camera. Matching, to capture at least one target to be captured ...

AI Technical Summary

Problems solved by technology

Although the system uses the Tornado framework or the Flask framework to build a data queue for sending video frames from multiple clients to the server, it is still impossible for the server to process huge amounts of data in a timely manner, facing the risk of system crashes in the case of large concurrent processing. Unable to effectively reduce data transmission pressure and storage pressure

[0005] The disadvantage of the existing technology is that many cameras on the market can only capture a single category or even a specific single target. For complex scenes of pedestrians, motor vehicles, and non-motor vehicles, multiple cameras are needed to capture them. Simultaneous deployment and control, in a complex scene where pedestrians, motor vehicles, and non-motor vehicles are mixed, it is impossible to realize real-time detection and analysis of videos with multiple categories, multiple scales, and many targets on one front-end camera. Obtain classified, high-quality image quality, and low-redundancy target object capture results

And, because the accuracy of its detection and tracking is not satisfactory, it is easy to cause more missed captures or repeated redundant captures, and the delay is relatively large

Images