Structured modeling based face key point positioning method

Abstract

The invention provides a structured modeling based face key point positioning method, which mainly solves a problem that an existing multispectral face recognition method is slow in computing speed. The scheme of the method comprises the steps of 1, collecting a face image and manually marking key points; 2, performing structured modeling on the face by using data acquired by the key points; acquiring a front face image, and segmenting a face region; 4, sequentially setting a binary threshold from 0 to 255, and performing binarization processing on the face region image; 5, solving edge points of the acquired binary images, performing clustering, solving a category center points and acquiring corresponding structured models; and 6, comparing the models with a model acquired in the step 2, and determining key points through solving the minimum difference. The structured modeling based face key point positioning method has the characteristics of high stability and high computing speed, and can be applied to the fields such as identity authentication, security monitoring and intelligent man-machine interaction.

Description

technical field [0001] The invention belongs to the field of biological feature recognition, and in particular relates to a method for locating key points of a human face, which can be used for identity authentication, safety monitoring and intelligent human-computer interaction. Background technique [0002] As an important feature of the human body, the face has played an increasingly important role in the field of biometrics in recent years. Face key point positioning is to further determine the center points of key organs such as eyes and mouth on the basis of the detected face. The accuracy of key point positioning directly affects the accuracy of face recognition results. [0003] At present, the common face key point location methods can be mainly divided into four types: gray-scale projection method, template matching method, active curve method and neural network-based method. in: [0004] The gray projection method is to count the total gray value in the horizont...

AI Technical Summary

Problems solved by technology

[0005] The template matching method uses a regular geometric image as a template to search and distinguish in the face area to locate facial features, such as eyes, mouth, etc. Although this method can accurately extract facial organs, it is very complicated;

[0006] The active curve method is a curve trained with specific parameters, such as a circle, a parabola, etc., to make it approach the shape of the target feature, so that the curve will converge in the target organ area, but its robustness needs to be strengthened;

[0007] The neural network method mainly uses its better self-learning function to enable it to obtain face and facial features. Although this method has achieved good results, it needs further exploration in key point positioning. For example, 2015 Nian Yang Haiyan and others designed and implemented a parallel convolutional neural network in "Research on Face Key Point Location Method Based on Parallel Convolutional Neural Network" ("Computer Application Research", Vol. 32, No. 8, 2517-2519). The face image, the upper half of the face and the lower half of the face are respectively sent to the convolutional network with the same structure for training and learning. By performing local convolution and downsampling on the image, the detailed features near the key points of the face are extracted, and the The positioning results of the three-level parallel network are weighted and synthesized to realize the positioning of key points of the face. However, due to the need for parallel convolution operations, the computational complexity is high, resulting in slow positioning of the key points of the face.

[0008] In summary, none of the existing research methods described above can meet the requirements for precise positioning of key points of the face in most cases.

Images