CN103208099A - Human body recognition method and human body recognition system for internet - Google Patents
Human body recognition method and human body recognition system for internet Download PDFInfo
- Publication number
- CN103208099A CN103208099A CN2013100864574A CN201310086457A CN103208099A CN 103208099 A CN103208099 A CN 103208099A CN 2013100864574 A CN2013100864574 A CN 2013100864574A CN 201310086457 A CN201310086457 A CN 201310086457A CN 103208099 A CN103208099 A CN 103208099A
- Authority
- CN
- China
- Prior art keywords
- image
- human body
- body recognition
- human
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention provides a human body recognition method and a human body recognition system for internet. The human body recognition method for the internet comprises the following steps of A acquiring and obtaining an image; B performing space model transformation of acquired image; C inputting the image to a mathematical model and obtaining human body region parameters in the image through mathematical model treatment; and D establishing a model sample corresponding to the human body region parameters according to the human body region parameters. The human body recognition method and the human body recognition system for the internet have the advantages that the human body region parameters in the image can be quickly and accurately detected, and a guarantee is provided for follow-up treatment.
Description
Technical field
The present invention relates to recognition methods, relate in particular to human body recognition methods and system for the internet.
Background technology
In the current fitting software, change one's clothes to such an extent that carry out the coupling of corresponding big or small clothes according to model's sample of making in advance.Make like this and cannot carry out the selection of corresponding clothes according to the size of actual human body.Thereby the feasible software that tries on a dress is experienced the user and is not reached desirable effect.
Summary of the invention
In order to solve the problems of the prior art, the invention provides a kind of human body recognition methods for the internet.
The invention provides a kind of human body recognition methods for the internet, comprise the steps:
A. image acquisition is obtained image;
B. the image that collects is carried out the spatial model conversion;
C. with image inputting mathematical model, handle the human region parameter of obtaining in the image through mathematical model;
D. set up corresponding with it model's sample according to the human region parameter.
As a further improvement on the present invention, in described step B, the coloured image that collects is converted into gray level image.
As a further improvement on the present invention, in described step C, comprise the steps:
C1. detect and keep the marginal information of human body in the image, reject the smooth region in the image;
C2. scan image obtains the human region parameter in the image.
As a further improvement on the present invention, in described step C2, to each line scanning of advancing in the image, from left to right scanning when finding left margin, is noted current location earlier; The right from image begins to scan toward the left side then, when finding right margin, notes current location; After entire image had been finished in scanning, the zone according to the border, the left and right sides of every row is formed got access to the human region parameter.
As a further improvement on the present invention, between described steps A and described step B, also comprise the image pre-treatment step, in described image pre-treatment step, the image that obtains is removed noise or reduced noise processed.
As a further improvement on the present invention, in described pre-treatment step, by median filtering method the image that obtains is removed noise or reduced noise processed.
As a further improvement on the present invention, in described steps A, the image that obtains is for extracting single-frame images from camera video stream.
The present invention also provides a kind of human body recognition system for the internet, comprising:
Image acquisition units is used for image acquisition, obtains image;
The spatial model converter unit is used for the image that collects is carried out the spatial model conversion;
The mathematical model unit is used for image inputting mathematical model, handles the human region parameter of obtaining in the image through mathematical model;
Model's sample is set up the unit, is used for setting up corresponding with it model's sample according to the human region parameter.
As a further improvement on the present invention, in described spatial model converter unit, the coloured image that collects is converted into gray level image.
As a further improvement on the present invention, described mathematical model unit also comprises:
Detection module, for detection of and keep the marginal information of human body in the image, reject the smooth region in the image;
Scan module is used for scan image, obtains the human region parameter in the image.
The invention has the beneficial effects as follows: the present invention can detect human region parameter in the image fast and accurately, for follow-up processing provides safeguard.
Description of drawings
Fig. 1 is the human body recognition methods process flow diagram for the internet of the present invention.
Fig. 2 is the specific implementation method process flow diagram of step S3 of the present invention.
Embodiment
As shown in Figure 1, the invention discloses a kind of human body recognition methods for the internet, comprise step S1 to step S4, in step S1, image acquisition is obtained image; In step S2, the image that collects is carried out the spatial model conversion; In step S3, with image inputting mathematical model, handle the human region parameter of obtaining in the image through mathematical model; In step S4, set up corresponding with it model's sample according to the human region parameter.
As one embodiment of the present of invention, in step S1, the image that obtains is for extracting single-frame images from camera video stream; Perhaps in step S1, the image that obtains also can be the photo of having taken.
To reduce in the actual camera shooting process surrounding environment and need carry out the pre-service of image to current frame image to the influence of photographic images quality in order to eliminate, remove or reduce the noise of image, thereby be conducive to the processing of successive image; So between described step S1 and described step S2, also comprise the image pre-treatment step, in described image pre-treatment step, the image that obtains removed noise or reduced noise processed; In described pre-treatment step, by median filtering method the image that obtains is removed noise or reduced noise processed.
Medium filtering is based on a kind of nonlinear signal processing technology that can effectively suppress noise of sequencing statistical theory, the ultimate principle of medium filtering is that the value of any in digital picture or the Serial No. is replaced with the Mesophyticum of each point value in the neighborhood of this point, the actual value that pixel value around allowing approaches, thus isolated noise spot eliminated.
If the size of image is the width of image for W*H(W, H is the height of image), Cm (x, y) m passage (x, value y) on the current coordinate of (x ∈ [0, W-1], y ∈ [0, H-1], m ∈ [0,2]) sign.Coloured image has three passages, is respectively R, G, B passage.
If the size of masterplate Tem is TemW*TemW (TemW is odd number), then the masterplate the inside adds up to TemT=TemW*TemW value.TemT value sorted from small to large, can get access to an intermediate value temM.
If Tem be currently located at channel C m position, the upper left corner (x0, y0), then with (x0, y0) TemT value in the Tem centered by can calculate Tem at the intermediate value temM0 of current location, with the value Cm of the value replacement current location of temM0 (x0, y0)=temM0.
Successively each pixel in three channel image is done processing like this, can obtain through the image after the filtering.
Can be transformed to different spatial models in the image channel, in order to be conducive to the effect of follow-up modeling, need earlier image to be carried out the spatial model conversion.Gathering the image of coming in is coloured image, have three color spaces of RGB, consider the information that following model uses and take into account processing speed, the RGB image need be converted into gray level image and show, so in step S2, the coloured image that collects is converted into gray level image.
Following formula is gray processing formula: GrayValue=R*0.299+G*0.587+B*0.114.If the image behind the gray processing be f (x, y), (x ∈ [0, W-1], y ∈ [0, H-1]). then f (x, y)=(x, y) (x, y) (x, y) * 0.114 for * 0.587+c3 for * 0.299+c2 for c1.
At the characteristics of human body in the real process, set up corresponding mathematical model by mathematical way, characteristics of human body and corresponding priori are input in this mathematical model.This mathematical model totally be divided into the input with output two parts, be most important in the flow process also be the step of most critical.Image and corresponding priori are used as the input that parameter is carried out mathematical model, conversion process and understanding by a sequence image in the mathematical model, obtain out corresponding human region parameter, so in step S3, with image inputting mathematical model, handle the human region parameter of obtaining in the image through mathematical model.
As shown in Figure 2, in step S3, also comprise following two steps: in step W1, detect and keep the marginal information of human body in the image, reject the smooth region in the image; In step W2, scan image obtains the human region parameter in the image.
In step W1, use Suo Beier operator (Sobel operator) to detect the marginal information of human body in the image.The Suo Beier operator is a discreteness difference operator, is used for the approximate value of gradient of arithmograph image brightness function.Use this operator in any point of image, will produce the matrix that corresponding gradient vector or its this operator of method vector comprise two groups of 3x3, be respectively horizontal and vertical, it and image are made the plane convolution, can draw respectively laterally and the approximate value of brightness difference longitudinally.
After handling through the Suo Beier operator, the edge of the human body in the image will embody, and smooth region will be disallowable falls, thereby the zone of being left is the zone that comprises human body information.
In step W2, to each line scanning of advancing in the image, from left to right scanning when finding left margin, is noted current location earlier; The right from image begins to scan toward the left side then, when finding right margin, notes current location; After entire image had been finished in scanning, the zone according to the border, the left and right sides of every row is formed got access to the human region parameter.
Formula (1) and formula (2) are respectively horizontal direction and vertical direction sobel detection algorithm.
(x, y) 1 M0(x0 in y0), can calculate corresponding value Dx0, Dy0 respectively through formula (1), (2) for f.
Then for
To f (x, y) each point in is done similar processing, then can obtain through the numerical value fs after the sobel conversion (x, y).In order to make image processing method just, need y) carry out binaryzation to fs(x.
If the threshold value of binaryzation be Th. then, when fs (x, y) 〉=Th the time, Fsb (x, y)=255; When fs (x, y)＜Th the time, Fsb (x, y)=0.
(x y) has been the edge of human body in the image to Fsb like this.Smooth region is disallowable to be fallen, thereby remaining zone is the zone that comprises human body information.
Parameters such as the human body shoulder that arranges when taking pictures according to the mobile phone end, waist, pin, we tentatively can set out human body border, the left and right sides roughly and be respectively LeftBody, RightBody.(x scans about y) going, and then can find the border, the left and right sides of every row, is made as Bl (y), Br (y) respectively, y ∈ [0, H-1] to Fsb.
Namely the LefetBody from horizontal direction begins to scan to the right, runs into Fsb(x, y)=255, then current x position is composed to Bl (y)=x, stops to scan to the right; From the RgihtBody scanning that begins to turn left, (x y)=255, then gives Br (y)=x the position of current x to run into Fsb; Stop to scan left, simultaneously, this end of scan continues the scanning of next line.
Successively after every line scanning, can tentatively extract human body marginal information position Fbody (x, y).If (x y)=255 is the border to Fbody, and (x y)=0 is non-borderline region to Fbody.
The subsequent processing steps that between step S3 and step S4, also comprises image, through the zone of model transferring after coming out owing to mathematical model is set up and actual environment light causes reason, might cause zone problem not too accurately, for head it off, to carry out the subsequent processing steps of image according to the priori of image again, namely carry out the correction of image.
According to coordinate parameters such as the eyes of the human body of setting, shoulder, waist, pin, (x, y) boundary position in carries out the fraction correction to Fbody.To Fbody(x, y) zone is divided into four part A 0, A1, and A2, A2 is corresponding to the zone of head, shoulder side, waist, pin.Then can get access to A0L, A0R, A0B, A0T identify left margin and the right margin of head respectively, and the rest may be inferred.
Then (x y) the y ∈ in [A0B, A0T] time, judges (whether xR belongs to interval [A0L, A0R] to FBody for x, y)=255 left and right sides boundary coordinate xL of correspondence as Fbody; In the time of xL＜A0L, xL=A0L; As XR〉A0R the time, xR=A0R; And the like, do similar operation, thereby finish the correction of image.
Human region parameter of the present invention comprises the human parameters that shoulder breadth, waistline, height etc. reflect by image.
The invention also discloses a kind of human body recognition system for the internet, comprising: image acquisition units, be used for image acquisition, obtain image; The spatial model converter unit is used for the image that collects is carried out the spatial model conversion; The mathematical model unit is used for image inputting mathematical model, handles the human region parameter of obtaining in the image through mathematical model; Model's sample is set up the unit, is used for setting up corresponding with it model's sample according to the human region parameter.
In described spatial model converter unit, the coloured image that collects is converted into gray level image.
Described mathematical model unit also comprises: detection module, for detection of and keep the marginal information of human body in the image, reject the smooth region in the image; Scan module is used for scan image, obtains the human region parameter in the image.
The present invention can detect human region parameter in the image, for follow-up processing provides safeguard fast and accurately by carrying out above each step.
The present invention is used for detecting the zone of human body automatically to be used for a finding width of cloth of suitable size to try on when trying on a dress accurately from a frame picture; Advantage: detection and localization goes out human region automatically, simple aspect.
In order to make fitting software reach the comparatively ideal effect that tries on a dress, need make the user select corresponding clothes to try on according to the actual size of own health, and this need solve the problem that how to get access to the health actual size, the present invention is by the recognition technology based on image, get access to the actual size of health, thereby make the fitting software users experience do better.
Purpose of the present invention is exactly the shortcoming that overcomes prior art, a kind of human body size automatic identification technology based on image is provided, thereby provides more friendly user to experience.Human body actual size in the software tries on a dress.The application layer program gets access to the human body actual size, preserves all size parameter, carries out the coupling of clothes when making things convenient for the user to change one's clothes.
During use, for example aiming at Zhang by mobile phone takes pictures, picture shot is handled by method of the present invention, thereby obtain Zhang's human region parameter, this human region parameter comprises Zhang's data such as height, shoulder breadth and measurements of the chest, waist and hips, sets up the model's sample that meets Zhang's human region parameter, so just on model's sample that clothes or the trousers of different size can be through Zhang, thereby can observe directly habited effect, make things convenient for the user.
Above content be in conjunction with concrete preferred implementation to further describing that the present invention does, can not assert that concrete enforcement of the present invention is confined to these explanations.For the general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, can also make some simple deduction or replace, all should be considered as belonging to protection scope of the present invention.
Claims (10)
1. a human body recognition methods that is used for the internet is characterized in that, comprises the steps:
A. image acquisition is obtained image;
B. the image that collects is carried out the spatial model conversion;
C. with image inputting mathematical model, handle the human region parameter of obtaining in the image through mathematical model;
D. set up corresponding with it model's sample according to the human region parameter.
2. human body recognition methods according to claim 1 is characterized in that: in described step B, the coloured image that collects is converted into gray level image.
3. human body recognition methods according to claim 2 is characterized in that, comprises the steps: in described step C
C1. detect and keep the marginal information of human body in the image, reject the smooth region in the image;
C2. scan image obtains the human region parameter in the image.
4. human body recognition methods according to claim 3 is characterized in that: in described step C2, to each line scanning of advancing in the image, from left to right scanning when finding left margin, is noted current location earlier; The right from image begins to scan toward the left side then, when finding right margin, notes current location; After entire image had been finished in scanning, the zone according to the border, the left and right sides of every row is formed got access to the human region parameter.
5. human body recognition methods according to claim 4 is characterized in that: also comprise the image pre-treatment step between described steps A and described step B, in described image pre-treatment step, the image that obtains is removed noise or reduced noise processed.
6. human body recognition methods according to claim 5 is characterized in that: in described pre-treatment step, by median filtering method the image that obtains is removed noise or reduced noise processed.
7. according to each described human body recognition methods of claim 1 to 6, it is characterized in that: in described steps A, the image that obtains is for extracting single-frame images from camera video stream.
8. a human body recognition system that is used for the internet is characterized in that, comprising:
Image acquisition units is used for image acquisition, obtains image;
The spatial model converter unit is used for the image that collects is carried out the spatial model conversion;
The mathematical model unit is used for image inputting mathematical model, handles the human region parameter of obtaining in the image through mathematical model;
Model's sample is set up the unit, is used for setting up corresponding with it model's sample according to the human region parameter.
9. human body recognition system according to claim 8 is characterized in that: in described spatial model converter unit, the coloured image that collects is converted into gray level image.
10. according to Claim 8 or 9 described human body recognition systems, it is characterized in that described mathematical model unit also comprises:
Detection module, for detection of and keep the marginal information of human body in the image, reject the smooth region in the image;
Scan module is used for scan image, obtains the human region parameter in the image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013100864574A CN103208099A (en) | 2013-03-18 | 2013-03-18 | Human body recognition method and human body recognition system for internet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013100864574A CN103208099A (en) | 2013-03-18 | 2013-03-18 | Human body recognition method and human body recognition system for internet |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103208099A true CN103208099A (en) | 2013-07-17 |
Family
ID=48755316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013100864574A Pending CN103208099A (en) | 2013-03-18 | 2013-03-18 | Human body recognition method and human body recognition system for internet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103208099A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102298797A (en) * | 2011-08-31 | 2011-12-28 | 深圳市美丽同盟科技有限公司 | Three-dimensional virtual fitting method, device and system |
| CN102346913A (en) * | 2011-09-20 | 2012-02-08 | 宁波大学 | Simplification method of polygon models of image |
| CN102682211A (en) * | 2012-05-09 | 2012-09-19 | 晨星软件研发(深圳)有限公司 | Three-dimensional fitting method and device |
| CN102945530A (en) * | 2012-10-18 | 2013-02-27 | 贵州宝森科技有限公司 | 3D intelligent apparel fitting system and method |
-
2013
- 2013-03-18 CN CN2013100864574A patent/CN103208099A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102298797A (en) * | 2011-08-31 | 2011-12-28 | 深圳市美丽同盟科技有限公司 | Three-dimensional virtual fitting method, device and system |
| CN102346913A (en) * | 2011-09-20 | 2012-02-08 | 宁波大学 | Simplification method of polygon models of image |
| CN102682211A (en) * | 2012-05-09 | 2012-09-19 | 晨星软件研发(深圳)有限公司 | Three-dimensional fitting method and device |
| CN102945530A (en) * | 2012-10-18 | 2013-02-27 | 贵州宝森科技有限公司 | 3D intelligent apparel fitting system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103927719B (en) | Picture processing method and device | |
| CN102663354B (en) | Face calibration method and system thereof | |
| CN109190522B (en) | Living body detection method based on infrared camera | |
| CN106548185B (en) | A kind of foreground area determines method and apparatus | |
| CN104599284B (en) | Three-dimensional facial reconstruction method based on various visual angles mobile phone auto heterodyne image | |
| CN108765278A (en) | A kind of image processing method, mobile terminal and computer readable storage medium | |
| CN104574285B (en) | One kind dispels the black-eyed method of image automatically | |
| CN107977650B (en) | Method for detecting human face and device | |
| CN109670430A (en) | A kind of face vivo identification method of the multiple Classifiers Combination based on deep learning | |
| CN103714345B (en) | A kind of method and system of binocular stereo vision detection finger fingertip locus | |
| CN103679759A (en) | Methods for enhancing images and apparatuses using the same | |
| CN104318603A (en) | Method and system for generating 3D model by calling picture from mobile phone photo album | |
| CN110264493A (en) | A kind of multiple target object tracking method and device under motion state | |
| WO2016029399A1 (en) | Object selection based on region of interest fusion | |
| WO2008111550A1 (en) | Image analysis system and image analysis program | |
| CN106412420B (en) | It is a kind of to interact implementation method of taking pictures | |
| CN106067167A (en) | Image processing method and device | |
| CN110276831A (en) | Constructing method and device, equipment, the computer readable storage medium of threedimensional model | |
| CN106097261A (en) | Image processing method and device | |
| Hong et al. | Near-infrared image guided reflection removal | |
| KR20100121817A (en) | Method for tracking region of eye | |
| CN105809085B (en) | Human-eye positioning method and device | |
| CN111669492A (en) | Method for processing shot digital image by terminal and terminal | |
| CN107944424A (en) | Front end human image collecting and Multi-angle human are distributed as comparison method | |
| CN111222432A (en) | Face living body detection method, system, equipment and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130717 |