Target object identification method, system and device, and medium

Abstract

The invention provides a target object recognition method, system and device and a medium. The method comprises the steps of acquiring a single-frame or multi-frame image containing one or more facesor human bodies; inputting an image containing the one or more human faces or human bodies into a hierarchical vectorization model to obtain human face or human body feature vectors of the image; andaccording to the human face or human body feature vector of the image, identifying whether the image contains the human face or human body of one or more target objects. According to the invention, whether a single-frame or multi-frame image contains a human face or a human body of one or more target objects can be identified; and a judgment is made that which image acquisition device the single-frame or multi-frame image comes from. Motion information of one or more target objects is generated according to the geographic positions corresponding to the image acquisition devices, so that cross-border head tracking can be carried out on the one or more target objects.

Description

technical field [0001] The present invention relates to image recognition technology, in particular to a target object recognition method, system, equipment and medium. Background technique [0002] In recent years, face recognition technology has also been widely used in the construction of "smart cities" and "safe cities". However, among the built cameras, more than 80% of the cameras cannot capture clear faces under any circumstances. Coupled with the improvement of the anti-reconnaissance ability of the perpetrators, they will deliberately avoid the cameras, capture face information in time, and It is difficult to deal with the alarm; not only that, in the actual scene, one camera often cannot cover all areas, and there is generally no overlap between multiple cameras. Therefore, this application proposes a target object recognition method, system, device and medium, which can construct a complete pedestrian trajectory by recognizing faces in a video, and realize cross-...

AI Technical Summary

Problems solved by technology

However, among the built cameras, more than 80% of the cameras cannot capture clear faces under any circumstances. Coupled with the improvement of the anti-reconnaissance ability of the perpetrators, they will deliberately avoid the cameras, capture face information in time, and It is difficult to deal with the alarm; not only that, in the actual scene, one camera often cannot cover all areas, and there is generally no overlap between multiple cameras

Images