Facial analysis, filtering method, device, embedded device, medium and integrated circuit

Abstract

The invention provides a face analysis and filtering method, device, embedded device, medium and integrated circuit. The face filtering method includes: obtaining a picture to be detected; using a face detector to detect the picture to be detected to obtain a detection result; when the detection result indicates that the picture to be detected is a face picture, using a trained random The fern classifier classifies the picture to be detected; and determines the category of the picture to be detected. The invention can use the random fern classifier to classify the detection results of the face detector, and combine the characteristics of the embedded device to reduce the number of false detections of faces in the face detection results under the premise of ensuring a high recall rate.

Description

technical field [0001] The invention relates to the technical field of face detection, in particular to a face analysis, filtering method, device, embedded device, medium and integrated circuit. Background technique [0002] The face detector is used to identify the location of the face from the signal generated by any type of sensor, generally using a rectangle or ellipse as a marker. In the prior art scheme, mainly include the face detection scheme of following 3 kinds of modes: [0003] (1) A face detection scheme based on traditional algorithms. [0004] Face detection schemes based on traditional algorithms generally use traditional hand-designed features for face description, such as Haar features, Histogram of Oriented Gradient (HOG) features, Local Binary Patterns (LBP) features , GABOR function, pixel difference (Pixel Difference, PD) feature, normalized pixel difference (Normalized Pixel Difference, NPD) feature, etc., and then use Boosting algorithm for classifi...

AI Technical Summary

Problems solved by technology

[0008] Although the above three schemes can perform face detection, they cannot guarantee the accuracy of face detection and reduce the number of false detections under the premise of high recall rate.

[0009] In addition, when the above-mentioned face detection scheme is applied to embedded devices, due to the limited computing power of embedded devices, false detections often occur, and it is even more difficult to achieve high detection accuracy under the condition of high recall rate.

Images