Binocular positioning method and binocular positioning apparatus

Abstract

The invention discloses a binocular positioning method and apparatus. The method comprises the steps of carrying out human face detection on a to-be-detected image, obtaining a human face position, and determining a to-be-detected human face bounding box covering the human face position; according to a correspondence relationship between a preset five-point bounding box and the human face bounding box, determining an estimated to-be-detected five-point bounding box corresponding to the to-be-detected human face bounding box, and determining initial positions of five key points of the to-be-detected image in the to-be-detected five-point bounding box accordingly; according to a preset display shape regression model, carrying out correction on the initial positions of the five key points, and obtaining final positions of the five key points; and extracting position coordinates of a left pupil point and a right pupil point from the final positions of the five key points. According to the method and apparatus disclosed by the invention, the eyes can be accurately positioned under the condition of a blur face image or blocked eyes, thereby improving the accuracy and stability of binocular positioning.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a binocular positioning method and a binocular positioning device. Background technique [0002] Binocular positioning is a very important basic technology in the field of digital image processing and pattern recognition. Many face alignment algorithms, facial features positioning algorithms, eye tracking algorithms, etc. all rely on the results of binocular positioning. The existing binocular positioning technology directly locates both eyes, which is prone to missed and false detections when the face image is blurred, the face has an expression or posture, and sunglasses are blocked. [0003] The prior art provides another human eye positioning method based on Haar-like features and an adaptive enhancement (Adaboost) algorithm. This method trains a large number of human eye samples and non-human eye samples based on Haar-like features, uses the adaptive enhancement al...

AI Technical Summary

Problems solved by technology

Compared with the human eye positioning method based on the gray projection method, this method has good detection effect and fast positioning speed, but it still does not solve the distance limitation between the camera and the face. When the distance between the camera and the face is long When , because the detected human eye features are blurred, it is easy to miss detection and false detection.

Images