Eyeball driving method and electronic device

Abstract

The invention discloses an eyeball driving method and an electronic device, and belongs to the field of image processing. The method comprises the steps of obtaining a plurality of 2D key point coordinates of a human face in a current video frame, and determining a human face region according to the 2D key point coordinates; filtering the human face region through a left and right eye filter to obtain 2D coordinates of eyeball positions of the human face in the current video frame; according to a conversion relationship between the 2D key point coordinates and corresponding 3D key point coordinates, converting the 2D coordinates of the eyeball positions into 3D coordinates of the eyeball positions; and according to the 3D coordinates of the eyeball positions and 3D coordinates of initial positions of eyeballs, driving the eyeballs of a target object. According to the method, the eyeball positions of the target object are driven according to the 3D coordinates of the eyeball positions and the 3D coordinates of the initial positions of the eyeballs, so that the accuracy and stability of eyeball locating are improved; and the method can efficiently run in a mobile terminal, so that the user experience is improved and the demand of a user in controlling the target object through eyeball rotation is met.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an eyeball driving method and electronic equipment. Background technique [0002] With the popularization of face recognition in computer technology, the technology of face recognition by computer has been applied in various fields. In face recognition, the recognition of eye state is very important, and the eyeball can be positioned through eye state recognition. Realize the user's needs for controlling objects in the game through eye recognition. [0003] The human eye positioning methods in the prior art mainly include the following: methods based on geometric features, such as grayscale projection method, Huffman transform, etc., are to first calculate the human eye edge contour with the Canny operator in the eye area of ​​interest, Then use Huffman transform to locate the eyeball; based on statistical methods, such as using neural network or support vector machine, knowledge...

AI Technical Summary

Problems solved by technology

[0004] However, the methods based on geometric features in the prior art are very susceptible to the influence of illumination, and the stability and accuracy are not enough. Although the methods based on statistics have certain adaptability to illumination, they need a large number of samples to train the classifier, making the use of This method has poor real-time performance when applied to mobile terminals. At the same time, the knowledge-based method uses a large amount of prior knowledge for human eye positioning, which is not universal enough.

Images