Blink detection method, device and equipment and computer readable storage medium

Abstract

The invention discloses a blink detection method comprising the steps that an eye opening detector and an eye closing detector are trained; a first face area in the first video frame is acquired, andthe first positions of the eyes are located by using the eye opening detector in the first face area; if locating is successful, the first positions are saved; a second face area in the second video frame is acquired, and the second positions of the eyes are located by using the eye opening detector in the second face area; if locating is successful, the video frames are switched in turn to perform locating until locating fails, and the second positions acquired in the last time of successful locating act as the first positions; the eye closing positions of the eyes are located by using the eye closing detector at the first positions; and if locating is successful, successful blinking is outputted. According to the method, the problems of error detection caused by shielding can be effectively avoided and the acquired classifier model is small so that the detection efficiency can be further enhanced. The invention also discloses a blink detection device and equipment and a computer readable storage medium which have the beneficial effects.

Description

technical field [0001] The present application relates to the field of living body recognition, in particular to a blink detection method, and also to a blink detection device, equipment and computer-readable storage medium. Background technique [0002] With the rapid improvement of living body recognition technology, its application fields are becoming more and more extensive. Under the upsurge of smart city development, living body recognition has become a favorable weapon to promote the construction of smart cities. Blink detection is a common recognition method in the field of live recognition. Blink detection in existing living body recognition usually uses a classifier or feature point positioning method. [0003] The feature point positioning method is to locate the position of the eyes through the face alignment algorithm, and calculate the closing degree of the upper and lower eyelids of the eyes according to the positioning points to determine the state of the e...

AI Technical Summary

Problems solved by technology

However, since the extraction of Haar features is generally accelerated through integral graphs, but the calculation of integral graphs cannot be accelerated through SIMD instructions, it may have a certain impact on front-end devices with low performance, resulting in a slow rate, and the Haar features required The number of weak classifiers is relatively large, so the size of the entire model is also large

Images