Training model generation method and human face detection method and device

Abstract

An embodiment of the invention discloses a training model generation method comprising the following steps: a plurality of human face positive samples and human face negative samples are obtained, characteristics of pixel difference between each pixel in each sample and other pixels of the sample are calculated, one object pixel difference characteristic is chosen from all the calculated pixel difference characteristics, point coordinates and pixel value difference corresponding to the object pixel difference characteristic are set as decision nodes for a decision making tree and are used for determining and distinguishing the positive samples and the negative samples, the decision nodes of the decision making tree are used as weak classifiers that are then cascaded, steps from a step of choosing the object pixel difference characteristic from all the calculated pixel difference characteristics to a step of using the decision nodes of the decision making tree as the weak classifiers that are then cascaded are subjected to iterative execution operation, and finally a strong classifier is generated. Point coordinates and pixel value difference corresponding to each decision node of the decision tree in the strong classifier are stored, and therefore a training model can be generated. The strong classifier is formed via the decision tree, a human face can be subjected to secondary classifying operation, and time needed for human face detection can be reduced when the same precision remains unchanged.

Description

technical field [0001] The invention relates to the field of human face detection, in particular to a training model generation method, a human face detection method, a training model generation device and a human face detection device. Background technique [0002] Face detection technology is a research hotspot in the field of pattern recognition and artificial intelligence, and is widely used in security systems, medicine, file management, video image processing, online payment, human-computer interaction, etc. With the development and application of smart phones, people gradually have a strong demand for face-related interaction methods, so high-precision and real-time face detection technology has become crucial. [0003] Traditional fast face detection devices based on HAAR+Adaboost have been widely used in PCs and digital cameras, but they are all combined with tracking to perform real-time face area positioning. [0004] The current way of extracting features is usu...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the embodiments of the present invention is to provide a training model generation method, a face detection method, a training model generation device, and a face detection device to solve the problem of embedding problems based on mobile phones in order to ensure accuracy and real-time performance in the prior art. The time-consuming problem of face detection of traditional mobile communication equipment is too long, and it is difficult to apply to the technical problems in actual scenarios

Images