Long-time stable human face detection and tracking method in bank scene and long-time stable human face detection and tracking device in bank scene

Abstract

The invention provides a long-time stable human face detection and tracking method in a bank scene. The method comprises the following steps that a human face detection module is adopted for detecting human faces in a certain frame of image; a moving target detection module is adopted for detecting targets moving in several continuous frames; the human face frame credibility is calculated according on the basis of the results of the human face detection results and the moving target detection results; the human face frame with higher credibility is output to a target tracking module; the target tracking module adopts an area matching method for tracking the human face frame with higher credibility; and the target tracking module adopts the area matching method for tracking the human face frame with higher credibility. The invention also provides a long-time stable human face detection and tracking device in the bank scene. Through the method and the device, the human face in any angle can be detected in a video, the specific human face can be output, the more stable and more continuous human face image output can be realized, the error detection condition of human face detection is effectively avoided, and the condition of detection missing capability of the human face detection is overcome to a certain degree.

Description

technical field [0001] The invention relates to a face detection and tracking method, in particular to a long-term stable face detection and tracking method in a banking scene. Background technique [0002] The usual method of existing face detection is to extract a feature map from a single frame image by means of pattern recognition, and use a pre-detector to perform face detection on the feature map. This method has certain limitations: first, the method of pattern recognition is that a probability event cannot achieve zero false positive rate and 100% detection rate; second, due to the limitations of the detector, it is impossible to detect faces from any angle ; Third, the traditional face detection lacks the use of inter-frame information. When the face has a large angle or the face is blurred, the detector cannot detect the face, so the detection is intermittently unstable. Contents of the invention [0003] Aiming at the problems of missed detection and false de...

AI Technical Summary

Problems solved by technology

This method has certain limitations: first, the method of pattern recognition is that a probability event cannot achieve zero false positive rate and 100% detection rate; second, due to the limitations of the detector, it is impossible to detect faces from any angle ; Third, the traditional face detection lacks the use of inter-frame information. When the face has a large angle or the face is blurred, the detector cannot detect the face, so that the detection is intermittently unstable.

Images