A Fast Locating Method of Facial Feature Points

Abstract

The invention discloses a method for quickly locating facial feature points, comprising the following steps: 1) using Haar-Like features as the basis, using a trained cascade classifier to sequentially detect the left eye and the right eye; 2) using the left and right human eyes to detect Based on the results, obtain the precise position of the inner corners of the left and right eyes; 3) Based on the inner corners and their distances, use the geometric features of the three courts and five eyes of the human eye to quickly determine the key feature points on the outer contours of the mouth and eyes, and save them These points are used as a reference feature table, as a reference for feature point search in the next frame; 4) take the left eye coordinates of the previous frame saved in the reference feature table as the initial position, and the inner corner distance as the benchmark, according to the geometric structure of the face, Set the human eye search range of the current frame in proportion, repeat steps 1) to 3), when the left eye cannot be detected within the set search range, start the full image search again, and repeat steps 1) to 3). The present invention can quickly and accurately locate facial feature points without special devices.

Description

technical field [0001] The invention belongs to the technical field of image processing and pattern recognition, and in particular relates to a method for rapidly locating facial feature points according to real-time positioning of specified feature points in facial video. Background technique [0002] At present, there are existing methods of judging the state of the eyes based on facial images, measuring the direction of the person's line of sight, and estimating the degree of arousal of the person. In order to judge the state of the eyes, it is necessary to correctly detect the eye feature points of the face image. In addition, when judging the awakening state of the vehicle driver, it is necessary to detect facial feature points in real time, including eye feature points and mouth feature points. [0003] For example, in Patent Document 1 (CN102254151A, a driver fatigue detection method based on facial video analysis), technologies such as facial feature point location ...

AI Technical Summary

Problems solved by technology

Although this method can accurately locate the feature points in most cases, the iterative matching process increases sharply with the increase of the number of feature points, and the matching time is too long, so it is not suitable for practical use.

[0006] Patent Document 2 uses a specific shape constraint on the detected eye edge image to detect the edge of the eye. There are two problems in practical application. One is that the fixed shape constraint is difficult to satisfy human eyes of different sizes and shapes at the same time, and the other is that The edge shape of the same human eye is greatly affected by different face poses. This method requires a special device proposed in Document 2 to ensure the reliability of the algorithm.

Images