How to detect occluded faces

Abstract

The invention discloses a method for detecting a blocked human face, comprising the following steps: constructing an image pyramid of each frame of pictures; Discrimination of occlusion; training with several face and non-face pictures; dividing each weak classifier according to the local position of the face used by each weak classifier when extracting features; based on the weak classifier obtained after classifying classifier collection {h i (j)}, to perform face detection and abnormal face discrimination; merge the face detection sets, and obtain the detection results of human faces or abnormal faces after the merger is completed. The present invention has the characteristics of correctly judging whether the human face is blocked when wearing masks of different colors, eye masks of different colors, dark sunglasses and the like.

Description

technical field [0001] The invention relates to the technical field of ATM machine safety protection equipment, in particular to a detection method that can quickly detect whether a human face is blocked, and a blocked human face that is blocked by the upper half of the face or the lower half of the face. Background technique [0002] All ATMs are equipped with pinhole cameras for taking pictures of people who withdraw cash, to monitor the behavior of those who operate the ATM, and to record the facial images of the people who withdraw money. In order to prevent the camera from capturing the image of his face and identify himself, criminals will deliberately cover their faces with sunglasses, masks, headscarves, etc. when operating ATMs. Therefore, covering the face to operate the ATM machine has always been the focus of bank supervision. [0003] At present, the number of ATM machines in China has reached more than 600,000, and it is still growing. ATM machines work 24 / 7....

AI Technical Summary

Problems solved by technology

[0007] The purpose of the invention of the present invention is to overcome the deficiency that the monitoring method in the prior art cannot accurately detect the situation of blocking a human face to operate an ATM machine, and to provide a method that can quickly detect whether a human face is blocked, and whether the upper half of the face is blocked or not. Detection method of occluded face with lower face occlusion

Images