CN104794275A - Face and hair style matching model for mobile terminal - Google Patents

Abstract

The invention discloses a face and hair style matching model for a mobile terminal. The face and hair style matching model based on the mobile terminal comprises two parts including the mobile phone mobile terminal and a server, wherein the mobile phone mobile terminal collects face outline feature vectors and sends the face outline feature vectors to the server, and the server feeds back to the mobile phone mobile terminal. The mobile phone mobile terminal collects a face image, the image is obtained by selecting existing photographs or photographing, a face outline is calculated automatically, and the face outline feature vectors are extracted; the feature vectors are submitted to the server, and a facial form category returned by the server is received; a hair style is recommended according to the facial form category. The server receives the face outline feature vectors sent by a mobile phone, the facial form category is judged based on the K nearest neighbor algorithm, and the facial form category result is sent to a mobile phone client side. Facial forms are classified in virtue of face recognition and the K nearest neighbor algorithm, and the calculation model is concise and efficient. The hair style is recommended according to the facial form, in addition, the face and hair style matching model for the mobile terminal is applied to a mobile phone platform, the operation is simple, and the applicability is strong.

Description

A kind of face of mobile terminal and hair style Matching Model

Technical field

The invention belongs to Mobile solution exploitation and field of image recognition, relate to a kind of face shape of face identification based on mobile terminal and hair style Matching Model.

Background technology

Traditional face recognition technology is mainly based on the recognition of face of visible images, and this is also the recognition method that people are familiar with, the development history of existing more than 30 year.But this mode has the defect being difficult to overcome, especially when ambient lighting changes, recognition effect can sharply decline, and cannot meet the needs of real system.The scheme solving lighting issues has 3-D view recognition of face, and thermal imaging recognition of face.But these two kinds of technology are also far immature, and recognition effect is unsatisfactory.

The a solution developed rapidly is the multiple light courcess face recognition technology based on active near-infrared image.It can overcome the impact of light change, and achieved remarkable recognition performance, the overall system performance in precision, stability and speed exceedes 3-D view recognition of face.This technology development in nearly 1 year rapidly, makes face recognition technology move towards practical gradually.

Face is equally inherent with other biological characteristic (fingerprint, iris etc.) of human body, its uniqueness and the superperformance be not easily replicated are that identity verify provides necessary prerequisite, compare recognition of face with the bio-identification of other type and there is following features: be non-imposed: user does not need to coordinate face collecting device specially, almost under unconscious state, just can obtain facial image, such sampling mode does not have " mandatory "; Untouchable: user does not need directly to contact with equipment just can obtain facial image; Concurrency: the sorting of multiple face, judgement and identification can be carried out under practical application scene; In addition, also meet visual characteristic: the characteristic of " knowing people with looks ", and the feature such as simple to operate, visual result, good concealment.

Current, there are some hair style design softwares in Mobile solution field, such as based on the hairdresser's of Android platform, smug hair style with send type etc. elsewhere based on ios platform, these are applied or recommend instantly popular hair style to user, and user maybe can be allowed to pass through hair style on probation of taking pictures, and function also comprises substantially chooses photo (by shooting or local acquisition), select hair style, place hair style, pull hair style, adjustment hair style etc.

Mobile terminal is also in conceptual phase by the hair style exemplary application of face recognition technology, and matured product is less.Therefore, in good time suitable proposition and improve existing face identification method, itself and mobile terminal are matched has very important Research Significance simultaneously.

Summary of the invention

Be add hair style on the photo of mobile phone based on the face of mobile terminal and hair style Matching Model, automatically identify personage's shape of face in photo, recommend hair style according to shape of face.Through size and the position thereof of suitably adjustment hair style material (or photo), rotate its angle, truly ideally experience hair style wearing effect.Model by means of face recognition technology, has the Functional Design of hommization, and operation is succinct, is the new opplication of field of mobile terminals.The technical matters solved mainly contains:

(1) face shape of face identification;

(2) hair style of different shape of face is recommended;

(3) friendly mobile terminal operation: for adding users is experienced, devise flexile gesture operation, facilitates user to choose photo and adjustment hair style.

For achieving the above object, the technical solution used in the present invention is a kind of face and hair style Matching Model of mobile terminal, mobile terminal of mobile telephone and server two parts are comprised based on the face of mobile terminal and hair style Matching Model, wherein, mobile terminal of mobile telephone gathers facial contour feature vector and sends to server, and server feeds back to mobile terminal of mobile telephone to mobile terminal of mobile telephone.

Mobile terminal of mobile telephone gathers facial image, and image chooses existing photo or take pictures obtain and automatically calculate facial contour, extracts facial contour feature vector; Proper vector is submitted to server, the shape of face classification that reception server returns; Hair style is recommended according to shape of face classification.

Server receives the facial contour feature vector that mobile phone is sent, and judges shape of face classification, shape of face category result is sent to cell-phone customer terminal based on k nearest neighbor algorithm.

Comprise the following steps based on the face of mobile terminal and hair style Matching Model:

Step one, mobile phone mobile client obtains facial image, and image sources has been taken in existing photo or startup camera function.

Step 2, mobile phone mobile client obtains facial contour by ASM (active shape model) algorithm, extracts facial contour feature vector.Owing to having bang, the top of face can be blocked, and therefore facial contour feature vector adopts profile below face, is made up of 20 pixels, [x1, y1, x2, y2, x3, y3 ... x20, y20].

Step 3, facial contour feature vector is sent to server by mobile phone mobile client.

Step 4, server receives face Outline Feature Vector, analyzes shape of face classification.Analytical approach is as follows:

(1) proper vector normalization; Calculate the center (x of proper vector _c, y _c)

$x_c=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_i$

$y_c=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i$

In formula, x _crepresent the transverse center point of face feature vector; y _crepresent longitudinal center's point of face feature vector; I represents each location point of face feature vector,

Each point does translation transformation, calculating reposition (x ' _i, y ' _i)

x′ _i＝x _i-x _c

y′ _i＝y _i-y _c

X ' _irepresent the transverse center point after face feature vector translation transformation; Y ' _irepresent the longitudinal center's point after face feature vector translation transformation;

The scope that vector is normalized to [0-1], method for normalizing is as follows:

$x_i^{\′}=\frac{x_i-x_{\min }}{x_{\max }-x_{\min }}$

$y_i^{\′}=\frac{y_i-y_{\min }}{y_{\max }-y_{\min }}$

In formula, x _minrepresent the normalized horizontal smallest point of face feature vector, x _maxrepresent the normalized horizontal maximum point of face feature vector; y _minrepresent the normalized longitudinal smallest point of face feature vector, y _maxrepresent the normalized longitudinal maximum point of face feature vector;

(2) adopt k nearest neighbor Algorithm Analysis facial contour generic, method is as follows:

Server end store marked classification circle, ellipse, square and triangle face proper vector Sample Storehouse.All proper vectors in sample to be sorted and Sample Storehouse calculate Euclidean distance, choose front K nearest sample.In K sample, by the shape of face classifications being designated unknown sample maximum for classification.

Step 5, shape of face analysis result is sent to cell-phone customer terminal by server.

Step 6, mobile client receives shape of face category result, recommends hair style according to shape of face.

Step 7, hair style is shown, supports gesture operation, and comprise hair style and recommend or freely choose hair style, picture amplifies, reduce, rotate, translation.

The invention has the beneficial effects as follows: the face shape of face identification based on mobile terminal that the present invention proposes and hair style Matching Model, consider face shape of face feature, classify to shape of face by recognition of face and k nearest neighbor algorithm, computation model is succinctly efficient.Recommend hair style according to shape of face, and be applied to cell phone platform, simple to operate, applicability is strong.

Accompanying drawing explanation

Fig. 1 is system of the present invention composition structural drawing;

Fig. 2 is facial contour feature schematic diagram of the present invention;

Fig. 3 is effect figure of the invention process.

Embodiment

Below in conjunction with drawings and Examples, the present invention is described in further detail.

A kind of face of mobile terminal and hair style Matching Model, mobile terminal of mobile telephone and server two parts are comprised based on the face of mobile terminal and hair style Matching Model, wherein, mobile terminal of mobile telephone gathers facial contour feature vector and sends to server, server feeds back to mobile terminal of mobile telephone to mobile terminal of mobile telephone, and the relationship of the two is shown in Fig. 1.

Mobile terminal of mobile telephone gathers facial image, and image chooses existing photo or take pictures obtain and automatically calculate facial contour, extracts facial contour feature vector; Proper vector is submitted to server, the shape of face classification that reception server returns; Hair style is recommended according to shape of face classification.

Server receives the facial contour feature vector that mobile phone is sent, and judges shape of face classification, shape of face category result is sent to cell-phone customer terminal based on k nearest neighbor algorithm.

Comprise the following steps based on the face of mobile terminal and hair style Matching Model:

Step one, mobile phone mobile client obtains facial image, and image sources has been taken in existing photo or startup camera function.

Step 2, mobile phone mobile client obtains facial contour by ASM (active shape model) algorithm, extracts facial contour feature vector.Owing to having bang, the top of face can be blocked, and therefore facial contour feature vector adopts profile below face, is made up of 20 pixels, [x1, y1, x2, y2, x3, y3 ... x20, y20], effect is shown in Fig. 2.

Step 3, facial contour feature vector is sent to server by mobile phone mobile client.

Step 4, server receives face Outline Feature Vector, analyzes shape of face classification.Analytical approach is as follows:

(1) proper vector normalization; Calculate the center (x of proper vector _c, y _c)

$x_c=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_i$

$y_c=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i$

In formula, x _crepresent the transverse center point of face feature vector; y _crepresent longitudinal center's point of face feature vector; I represents each location point of face feature vector,

Each point does translation transformation, calculating reposition (x ' _i, y ' _i)

x′ _i＝x _i-x _c

y′ _i＝y _i-y _c

X ' _irepresent the transverse center point after face feature vector translation transformation; Y ' _irepresent the longitudinal center's point after face feature vector translation transformation;

The scope that vector is normalized to [0-1], method for normalizing is as follows:

$x_i^{\′}=\frac{x_i-x_{\min }}{x_{\max }-x_{\min }}$

$y_i^{\′}=\frac{y_i-y_{\min }}{y_{\max }-y_{\min }}$

In formula, x _minrepresent the normalized horizontal smallest point of face feature vector, x _maxrepresent the normalized horizontal maximum point of face feature vector; y _minrepresent the normalized longitudinal smallest point of face feature vector, y _maxrepresent the normalized longitudinal maximum point of face feature vector;

(2) adopt k nearest neighbor Algorithm Analysis facial contour generic, method is as follows:

Server end store marked classification circle, ellipse, square and triangle face proper vector Sample Storehouse.All proper vectors in sample to be sorted and Sample Storehouse calculate Euclidean distance, choose front K nearest sample.In K sample, by the shape of face classifications being designated unknown sample maximum for classification.

Step 5, shape of face analysis result is sent to cell-phone customer terminal by server.

Step 6, mobile client receives shape of face category result, recommends hair style according to shape of face.

Step 7, hair style is shown, supports gesture operation, and comprise hair style and recommend or freely choose hair style, picture amplifies, reduce, rotate, translation, and bandwagon effect is shown in Fig. 3.

Claims (1)

1. the face of a mobile terminal and hair style Matching Model, it is characterized in that: comprise mobile terminal of mobile telephone and server two parts based on the face of mobile terminal and hair style Matching Model, wherein, mobile terminal of mobile telephone gathers facial contour feature vector and sends to server, and server feeds back to mobile terminal of mobile telephone to mobile terminal of mobile telephone.

Mobile terminal of mobile telephone gathers facial image, and image chooses existing photo or take pictures obtain and automatically calculate facial contour, extracts facial contour feature vector; Proper vector is submitted to server, the shape of face classification that reception server returns; Hair style is recommended according to shape of face classification.

Server receives the facial contour feature vector that mobile phone is sent, and judges shape of face classification, shape of face category result is sent to cell-phone customer terminal based on k nearest neighbor algorithm.

Comprise the following steps based on the face of mobile terminal and hair style Matching Model:

Step one, mobile phone mobile client obtains facial image, and image sources has been taken in existing photo or startup camera function.

Step 2, mobile phone mobile client obtains facial contour by ASM (active shape model) algorithm, extracts facial contour feature vector.Owing to having bang, the top of face can be blocked, and therefore facial contour feature vector adopts profile below face, is made up of 20 pixels, [x1, y1, x2, y2, x3, y3 ... x20, y20].

Step 3, facial contour feature vector is sent to server by mobile phone mobile client.

Step 4, server receives face Outline Feature Vector, analyzes shape of face classification.Analytical approach is as follows:

(1) proper vector normalization; Calculate the center (x of proper vector _c, y _c)

$x_c=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_i$

$y_c=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i$

In formula, x _crepresent the transverse center point of face feature vector; y _crepresent longitudinal center's point of face feature vector; I represents each location point of face feature vector,

Each point does translation transformation, calculating reposition (x ' _i, y ' _i)

x′ _i＝x _i-x _c

y′ _i＝y _i-y _c

X ' _irepresent the transverse center point after face feature vector translation transformation; Y ' _irepresent the longitudinal center's point after face feature vector translation transformation;

The scope that vector is normalized to [0-1], method for normalizing is as follows:

$x_i^{\′}=\frac{x_i-x_{\min }}{x_{\max }-x_{\min }}$

$y_i^{\′}=\frac{y_i-y_{\min }}{y_{\max }-y_{\min }}$

In formula, x _minrepresent the normalized horizontal smallest point of face feature vector, x _maxrepresent the normalized horizontal maximum point of face feature vector; y _minrepresent the normalized longitudinal smallest point of face feature vector, y _maxrepresent the normalized longitudinal maximum point of face feature vector;

(2) adopt k nearest neighbor Algorithm Analysis facial contour generic, method is as follows:

Server end store marked classification circle, ellipse, square and triangle face proper vector Sample Storehouse.All proper vectors in sample to be sorted and Sample Storehouse calculate Euclidean distance, choose front K nearest sample.In K sample, by the shape of face classifications being designated unknown sample maximum for classification.

Step 5, shape of face analysis result is sent to cell-phone customer terminal by server.

Step 6, mobile client receives shape of face category result, recommends hair style according to shape of face.

Step 7, hair style is shown, supports gesture operation, and comprise hair style and recommend or freely choose hair style, picture amplifies, reduce, rotate, translation.