CN104506778A - Flashlight control method and device based on age estimation - Google Patents
Flashlight control method and device based on age estimation Download PDFInfo
- Publication number
- CN104506778A CN104506778A CN201410804921.3A CN201410804921A CN104506778A CN 104506778 A CN104506778 A CN 104506778A CN 201410804921 A CN201410804921 A CN 201410804921A CN 104506778 A CN104506778 A CN 104506778A
- Authority
- CN
- China
- Prior art keywords
- age
- facial image
- image
- facial
- key point
- 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
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a flashlight control method and a flashlight control device based on age estimation. The flashlight control method comprises the following steps: performing face recognition according to the current captured preview camera data, and obtaining a recognized face image; performing age estimation on the face image, judging whether the face image is an infant face image according to age estimation results, automatically switching off the flashlight if the face image is the infant face image. Parents do not need to worry about forgetting to switch off the flashlight during photographing, so that the harms of hard light emitted by the flashlight on eye retina of the infant are avoided.
Description
Technical field
The present invention relates to a kind of shooting technology, particularly a kind of flash lamp control method based on age estimation and device.
Background technology
The shooting of digital video and digital camera and camera function bring people convenient and many fine memories greatly, many father and mother like being born child, one week, ten days, full moon, 100 days, some important dates such as one full year of life take pictures to baby, then get up to record the developmental process of child by collection of photos, but we are when taking pictures to the baby of just birth, often forget closedown photoflash lamp, particularly when indoor shot, due to insufficient light, and photoflash lamp detects insufficient light and automatically opens when causing taking pictures, the eye retina of the high light that photoflash lamp sends to baby causes certain injury, thus affect its visual development.
Summary of the invention
The present invention is for solving the problem, and provide a kind of flash lamp control method based on age estimation and device, it estimated to judge infant by the age, and automatic flash light, forget closedown photoflash lamp when taking pictures without the need to worrying.
For achieving the above object, the technical solution used in the present invention is:
Based on the flash lamp control method that the age is estimated, it is characterized in that, comprise the following steps:
10. catch the camera data of current preview;
Camera data described in 20. pairs carry out recognition of face, and obtain the facial image recognized;
Facial image described in 30. pairs carries out age estimation;
40. according to the age estimated result judge described in facial image whether be infant, if then automatically close photoflash lamp.
Preferably, in described step 30, age estimation is carried out to described facial image, mainly utilize the method for convolutional neural networks to carry out age estimation to described facial image.
Preferably, in described step 30, age estimation is carried out to described facial image, further comprising the steps:
31. collect sample images pedestrian's work of going forward side by side marks the age type of each sample image, and preset standard face figure;
The face characteristic key point of 32. comparison sample images and the face characteristic key point of standard faces figure, key point alignment and adjustment are carried out to sample image, again the facial contour figure that contours extract obtains sample image is carried out to the sample image after adjustment, and the facial contour figure of this sample image is normalized and inputs the training that convolutional neural networks system carries out disaggregated model;
33. obtain the facial image recognized from camera data, the face characteristic key point of this facial image of comparison and the face characteristic key point of standard faces figure, key point alignment and adjustment are carried out to described facial image, again the facial contour figure that contours extract obtains described facial image is carried out to the facial image after adjustment, and adopt described disaggregated model to carry out character classification by age in the facial contour figure of facial image, obtain the age type of described facial image.
Preferably, in the training process carrying out disaggregated model, the sample image of classification error collects and performs step 31 and step 32 successively by convolutional neural networks system, until be decided to be when exceeding expected results described step 32 the disaggregated model of training out be optimal classification model, and in described step 33, adopt the facial contour figure of this optimal classification model to described facial image to carry out character classification by age, obtain the age type of described facial image.
Preferably, described age type comprises: infant, children, teenager, youth, middle age, old age.
Preferably, described age type comprises: two years old with Types Below and more than two years old type.
Preferably, the face characteristic key point of comparison sample image and the face characteristic key point of standard faces figure in described step 32, key point alignment and adjustment are carried out to sample image, or, the face characteristic key point of facial image described in comparison and the face characteristic key point of standard faces figure in described step 33, key point alignment and adjustment are carried out to described facial image, mainly pass through the face characteristic key point of preset standard face figure and correspondence thereof, and according to the sample image obtained or the human face region of facial image and the face characteristic key point of correspondence thereof, the face characteristic key point of affine transformation to sample image or facial image is utilized to align and adjust, again to adjustment after sample image or facial image carry out contours extract be adjusted after facial contour figure, and using the facial contour figure of the facial contour figure after this adjustment as described sample image or facial image.
Preferably, described contours extract mainly carries out wavelet transformation to the sample image after described adjustment or facial image, obtains the frequency domain figure picture of low frequency component, namely described facial contour figure.
Preferably, adopt described disaggregated model to carry out in the facial contour figure of described facial image in described step 33 age type that character classification by age obtains described facial image, mainly the facial contour figure of described described facial image is put into the probability that convolutional neural networks system carries out calculating each age type of described facial image, and the maximum age type of select probability is as the age type of described facial image.
Preferably, described face characteristic key point mainly comprises eye contour, mouth, eyebrow, face mask line, forehead.
In addition, the present invention is based on said method and additionally provide a kind of device for controlling flashlight estimated based on the age, it is characterized in that: it comprises
Camera model, for starting camera and catching the camera data of current preview;
Face recognition module, carries out recognition of face to described camera data, and obtains the facial image recognized;
Character classification by age module, it carries out age estimation to described facial image;
Executive Module, its according to the age estimated result judge described in facial image whether be infant, if then automatically close photoflash lamp, and drive camera to take pictures.
Preferably, described character classification by age module comprises further:
Sampling unit, for collecting the age type of sample image for artificial each sample image of mark;
Edit cell, its face characteristic key point by the described facial image of sample image described in comparison or acquisition and the face characteristic key point of standard faces figure preset, key point alignment and adjustment are carried out to sample image or facial image, then the facial contour figure that contours extract obtains sample image or facial image is carried out to the sample image after adjustment or facial image;
Normalization unit, is normalized the facial contour figure of described sample image or facial image;
Unit, inputs the facial contour figure of the sample image after normalized or facial image the training that convolutional neural networks system carries out disaggregated model;
Judging unit, adopts described disaggregated model to carry out character classification by age in the facial contour figure of described facial image, to judge the age type of described facial image.
The invention has the beneficial effects as follows:
A kind of flash lamp control method based on age estimation of the present invention and device, it is by carrying out recognition of face to the camera data of the current preview of catching, and obtain the facial image recognized, then age estimation is carried out to described facial image, whether the facial image described in judging according to age estimated result is infant, if then automatically close photoflash lamp, father and mother forget closedown photoflash lamp without the need to worrying when taking pictures, thus the injury that the eye retina of the high light avoiding photoflash lamp to send to infant causes.
Accompanying drawing explanation
Accompanying drawing described herein is used to provide a further understanding of the present invention, forms a part of the present invention, and schematic description and description of the present invention, for explaining the present invention, does not form inappropriate limitation of the present invention.In the accompanying drawings:
Fig. 1 is the general flow chart of a kind of flash lamp control method based on age estimation of first embodiment of the invention;
Fig. 2 is the general flow chart of a kind of flash lamp control method based on age estimation of second embodiment of the invention;
Fig. 3 is the schematic block diagram of a kind of device for controlling flashlight based on age estimation of first embodiment of the invention;
Fig. 4 is the schematic block diagram of a kind of device for controlling flashlight based on age estimation of second embodiment of the invention.
Embodiment
In order to make technical problem to be solved by this invention, technical scheme and beneficial effect clearly, understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.
As shown in Figure 1, a kind of flash lamp control method estimated based on the age of the present invention, it comprises the following steps:
10. catch the camera data of current preview;
Camera data described in 20. pairs carry out recognition of face, and obtain the facial image recognized;
Facial image described in 30. pairs carries out age estimation;
40. according to the age estimated result judge described in facial image whether be infant, if then automatically close photoflash lamp.
In described step 30, age estimation is carried out to described facial image, mainly utilize the method for convolutional neural networks to carry out age estimation to described facial image; Specifically, as shown in Figure 2, further comprising the steps:
31. collect sample images pedestrian's work of going forward side by side marks the age type of each sample image, and preset standard face figure;
The face characteristic key point of 32. comparison sample images and the face characteristic key point of standard faces figure, key point alignment and adjustment are carried out to sample image, again the facial contour figure that contours extract obtains sample image is carried out to the sample image after adjustment, and the facial contour figure of this sample image is normalized and inputs the training that convolutional neural networks system carries out disaggregated model;
33. obtain the facial image recognized from camera data, the face characteristic key point of this facial image of comparison and the face characteristic key point of standard faces figure, key point alignment and adjustment are carried out to described facial image, again the facial contour figure that contours extract obtains described facial image is carried out to the facial image after adjustment, and adopt described disaggregated model to carry out character classification by age in the facial contour figure of facial image, obtain the age type of described facial image.
As preferred embodiment, in the training process carrying out disaggregated model, the sample image of classification error collects and performs step 31 and step 32 successively by convolutional neural networks system, until be decided to be when exceeding expected results described step 32 the disaggregated model of training out be optimal classification model, and in described step 33, adopt the facial contour figure of this optimal classification model to described facial image to carry out character classification by age, obtain the age type of described facial image; In described step 32, the facial contour figure of sample image be normalized and input the training that convolutional neural networks system carries out disaggregated model, mainly bringing the facial contour figure of sample image good for manual sort into convolutional neural networks system and learn; And, the sample image of classification error is collected and re-starts mark, namely, during age Type-Inconsistencies for the age type of system automatic classification and manual sort, represent artificial marking error or genealogical classification mistake, need re-start and manually mark and adjust network configuration, again the sample image after mark is again carried out training study again, the process of " training-> adjusts network configuration-> retraining " that so repeats is until classification is correct.
In the present embodiment, network order is input layer->K the full articulamentum of group layer->->SoftMax layer, and wherein K is more than or equal to 1; Group's layer comprises convolutional layer, active coating, down-sampling layer, normalization layer; In convolutional layer, active coating, down-sampling layer, normalization layer each layer core size and export size and can carry out regulating arbitrarily, and each layer has one to input and produces an output, and the output of every one deck is as the input of lower one deck.
Wherein, the input size of input layer is Height x Weight x Channel, and wherein Weight, Height are the wide and high of input layer image, and Channel is the Color Channel of input layer image; Because the present invention uses GPU hardware to realize, Weight=Height; The channel of input picture can only be 1 or 3.
Convolutional layer:
1) size of core must be odd number, and is not more than the wide or high of this layer of input;
2) intermediate representation is wide and high by not changing during convolutional layer, and port number is variable can be constant; Can be any positive integer in theory, because the present invention uses GPU hardware to realize, be the multiple of 16 here.
Active coating:
1) active coating does not change wide, the high or port number that convolutional layer represents;
2) activation primitive that active coating uses includes but not limited to following type function:
f(x)＝1/(1+e
-x)
F (x)=a*tanh (b*x), a, b are any non-zero real
f(x)＝max(0,x)
f(x)＝min(a,max(0,x))
f(x)＝log(1+e
x)
f(x)＝|x|
f(x)＝x
2
f(x)＝ax+b
3) active coating is followed after convolutional layer or full connection.
Down-sampling layer:
1) down-sampling layer does not change the port number of intermediate representation;
2) drawdown ratio of down-sampling layer to image is the size of core: namely core is that the down-sampling layer of m x n can cause intermediate representation to be reduced into last layer (1/m) x (1/n), m and n can be random natural number in theory, because the present invention uses GPU hardware to realize, m=n.Such as, 15x 15x 32, by after the down-sampling of 3x 3, becomes 5x 5x 32; 15x 15x 32, by after the down-sampling of 5x 5, becomes 3x 3x 32; But 15x 15x 32 can not carry out the down-sampling of 2x 2, because 15 can not be divided exactly by 2; Be not that input size must be the power of 2, namely 16,32,64 etc., as long as input size guarantees to be sampled by all down-sampling layers.
Normalization layer:
1) normalization layer does not change any size of intermediate representation;
2) normalization layer is not necessarily, must, add normalization layer and usually can improve precision and increase amount of calculation; Whether add normalization layer, the actual precision of lifting and the speed of loss after adding be seen.
General combination is: convolution-> activates-> down-sampling-> normalization.
Following situation is special:
1) when interpolation normalization layer but increases a lot of operand to precision improvement is less, cancel normalization layer, namely adopt following combination: convolution-> activation-> down-sampling;
2) in advance, effect is substantially identical for normalization layer, namely adopts following combination: convolution-> activates-> normalization-> down-sampling.
3) down-sampling layer is cancelled: convolution-> activates; Or convolution-> activates-> normalization; Down-sampling essence is to increase robustness, has the effect of the operand reducing succeeding layer in passing simultaneously; Usually have which floor down-sampling in a network, but not all " convolution-> activates " all to follow down-sampling below.
Full articulamentum:
1) can become 1 dimension by the intermediate representation after full articulamentum, be no longer 3 dimensions;
2) the full output connected can be any;
3) once enter full connection, just convolution, down-sampling or normalization cannot be carried out;
4) can active coating be connect after full connection, or continue to connect full connection.
SoftMax layer:
After being connected on full articulamentum, effect connects complete the real-valued probability become between [0,1] produced.
The network configuration that the present invention finally uses is as shown in table 1.
Table 1 convolutional neural networks structure
| The number of plies | Type | Core size | Export size | Explain |
| 1 | Input layer | 32x32x3 | ||
| 2 | Convolutional layer | 5x5 | 32x32x32 | |
| 3 | Active coating | 32x32x32 | ||
| 4 | Down-sampling layer | 2x2 | 16x16x32 | f(x)＝x 2 |
| 5 | Normalization layer | 16x16x32 | Use local normalization | |
| 6 | Convolutional layer | 5x5 | 16x16x16 | |
| 7 | Active coating | 16x16x16 | ||
| 8 | Down-sampling layer | 2x2 | 8x8x16 | f(x)＝|x| |
| 9 | Normalization layer | 8x8x16 | Use local normalization |
| 10 | Full articulamentum | 6 data | ||
| 11 | SoftMax layer | 6 data |
Adopt described disaggregated model to carry out in the facial contour figure of described facial image in described step 33 age type that character classification by age obtains described facial image, mainly the facial contour figure of described described facial image is put into the probability that convolutional neural networks system carries out calculating each age type of described facial image, and the maximum age type of select probability is as the age type of described facial image.Concrete mainly carry out key point by the human face region in described facial image and to align and contours extract obtains facial contour figure, put into the input layer of neural net, after entirely connecting, obtain the probability of each label at last SoftMax layer, namely real-valued in interval [0,1]; Be divided into according to age type in the present embodiment: infant, children, teenager, youth, middle age, old age, the age label of totally 6 types, i.e. 6 data, these 6 data and equal 1; The maximum label of select probability is as the label of the age type of this facial image.In step 32, the facial contour figure of this sample image is normalized and inputs the training that convolutional neural networks system carries out disaggregated model, and the determination methods of its age type is similar to the above.
Described age type comprises: infant (0 to 2 years old), children (3 to 6 years old), juvenile (7 to 14 years old), young (15 to 35 years old), middle age (36 to 60 years old), old (more than 61 years old), if judged result is infant, then automatically close photoflash lamp; Or described age type comprises: two years old with Types Below and more than two years old type, if the age type of judged result be two years old with Types Below, then think infant, and automatically close photoflash lamp.
The face characteristic key point of comparison sample image and the face characteristic key point of standard faces figure in described step 32, key point alignment and adjustment are carried out to sample image, or, the face characteristic key point of facial image described in comparison and the face characteristic key point of standard faces figure in described step 33, key point alignment and adjustment are carried out to described facial image, mainly pass through the face characteristic key point of preset standard face figure and correspondence thereof, and according to the sample image obtained or the human face region of facial image and the face characteristic key point of correspondence thereof, the face characteristic key point of affine transformation to sample image or facial image is utilized to align and adjust, again to adjustment after sample image or facial image carry out contours extract be adjusted after facial contour figure, and using the facial contour figure of the facial contour figure after this adjustment as described sample image or facial image.
Described affine transformation affine transformation (Affine Transform) is the one of rectangular space coordinate conversion, it is the linear transformation between a kind of two-dimensional coordinate to two-dimensional coordinate, keep " grazing " (straightness of X-Y scheme, i.e. straight line or straight line can not bend after conversion, circular arc or circular arc) and " collimation " (parallelism, namely keep the relative position relation between X-Y scheme constant, parallel lines or parallel lines, the angle of cut of intersecting straight lines is constant); The face characteristic key point of affine transformation to standard faces figure is utilized to align and adjust, mainly by a series of conversion such as movement, convergent-divergent, upset, rotation, the face characteristic key points such as sample image or facial image and eyes, nose, face in standard faces figure are adjusted to the position corresponding with standard faces figure.
In the present embodiment, described face characteristic key point mainly comprises eye contour, mouth, eyebrow, face mask line, forehead.The detection of described face characteristic key point mainly utilizes ASM (ActiveShape Model) algorithm, and it is divided into training and search two steps: during training, set up the position constraint of each characteristic point, build the local feature of each specified point; During search, the coupling of iteration; This algorithm is that prior art does not repeat at this.Described face characteristic key point mainly comprises eye contour, mouth, eyebrow, face mask line, forehead etc.
The basic thought of face mess generation first designs the standard triangle gridding meeting basic face shape and organ distribution, by defining each vertex of a triangle sequence number, obtains the topological relation between the relative position of grid point and triangle gridding dough sheet; Then with the control point coordinate that human face characteristic point extraction algorithm obtains, calibration distortion is carried out to standard grid, thus realize the personalization face mess generation of different human face photo.
The match point put between curve adopts Lagrange's interpolation to calculate.
The generating algorithm of grid point is described below:
Eye contour: have 16 points about eye contour in 88 characteristic points, and we need to carry out calibration location to 20 points among standard grid.We are according to parabola on dot generation eyes in left eye angle point, right eye angle point and top; By parabola under left eye angle point, right eye angle point and following middle dot generation eyes.All 20 point of acquisition is got four first-class horizontal ranges of parabola.
Mouth: in 88 characteristic points, mouth profile has 22 points, needs in standard grid to carry out calibration location to 34 points.Generate parabola 9 ~ 12 and carry out matching, obtain all 34 points.
Eyebrow: in 88 characteristic points, eyebrow has 16 points, needs in standard grid to carry out calibration location to 20 points.Generate parabola 1,2,3 and 4 and carry out matching, obtain all 20 points.
Face mask line: have 21 points to represent face mask line in 88 characteristic points.And in grid chart, have 33 points represent outline line.Outline line is divided into 4 sections, uses parabola 13 ~ 16 matching respectively.
Forehead: by the forehead trichion of actual face and standard face, both sides cheek peak calculates affine transformation matrix.The effect that forehead part plays in human face expression action is less, and the method that therefore grid of forehead part have employed affine transformation carries out approximate generation.
Other points: as the point at the places such as forehead, cheek, mouth periphery, their coordinate calculates in proportion according to the grid point reserving position.
Described contours extract mainly carries out wavelet transformation to the sample image after described adjustment or facial image, obtains the frequency domain figure picture of low frequency component, namely described facial contour figure.Wherein wavelet transformation is the partial transformation of space (time) and frequency, thus can information extraction from signal effectively.Wavelet transformation is a kind of new transform analysis method, the thought of its inherit and development short time discrete Fourier transform localization, overcome again window size not with shortcomings such as frequency change simultaneously, can provide one with " T/F " window of frequency shift, be the ideal tools of carrying out signal time frequency analysis and process.Its main feature is can the feature of abundant some aspect of outstanding problem by conversion.
In addition, as shown in Figure 3, the present invention is based on said method and additionally provide a kind of device for controlling flashlight estimated based on the age, it is characterized in that: it comprises
Camera model A, for starting camera and catching the camera data of current preview;
Face recognition module B, carries out recognition of face to described camera data, and obtains the facial image recognized;
Character classification by age module C, it carries out age estimation to described facial image;
Executive Module D, its according to the age estimated result judge described in facial image whether be infant, if then automatically close photoflash lamp, and drive camera to take pictures.
As shown in Figure 4, described character classification by age module comprises further:
Sampling unit C1, for collecting the age type of sample image for artificial each sample image of mark;
Edit cell C2, its face characteristic key point by the described facial image of sample image described in comparison or acquisition and the face characteristic key point of standard faces figure preset, key point alignment and adjustment are carried out to sample image or facial image, then the facial contour figure that contours extract obtains sample image or facial image is carried out to the sample image after adjustment or facial image;
Normalization unit C3, is normalized the facial contour figure of described sample image or facial image;
Unit C4, inputs the facial contour figure of the sample image after normalized or facial image the training that convolutional neural networks system carries out disaggregated model;
Judging unit C5, adopts described disaggregated model to carry out character classification by age in the facial contour figure of described facial image, to judge the age type of described facial image.
The present invention proposes a kind of photographic method estimated based on the age, when taking pictures, first age estimation is carried out to the image of preview, and when estimating that the age of current shooting object is for infant, automatic closedown photoflash lamp, and then it is taken, thus the injury that the eye retina of the high light avoiding photoflash lamp to send to infant causes, father and mother forget closedown photoflash lamp without the need to worrying when taking pictures.
It should be noted that, each embodiment in this specification all adopts the mode of going forward one by one to describe, and what each embodiment stressed is the difference with other embodiments, between each embodiment identical similar part mutually see.For the embodiment of categorizing system, due to itself and embodiment of the method basic simlarity, so description is fairly simple, relevant part illustrates see the part of embodiment of the method.Further, one of ordinary skill in the art will appreciate that all or part of step realizing above-described embodiment can have been come by hardware, the hardware that also can carry out instruction relevant by program completes.
Above-mentioned explanation illustrate and describes the preferred embodiments of the present invention, be to be understood that the present invention is not limited to the form disclosed by this paper, should not regard the eliminating to other embodiments as, and can be used for other combinations various, amendment and environment, and can in invention contemplated scope herein, changed by the technology of above-mentioned instruction or association area or knowledge.And the change that those skilled in the art carry out and change do not depart from the spirit and scope of the present invention, then all should in the protection range of claims of the present invention.
Claims (12)
1., based on the flash lamp control method that the age is estimated, it is characterized in that, comprise the following steps:
10. catch the camera data of current preview;
Camera data described in 20. pairs carry out recognition of face, and obtain the facial image recognized;
Facial image described in 30. pairs carries out age estimation;
40. according to the age estimated result judge described in facial image whether be infant, if then automatically close photoflash lamp.
2. a kind of flash lamp control method estimated based on the age according to claim 1, it is characterized in that: in described step 30, age estimation is carried out to described facial image, mainly utilize the method for convolutional neural networks to carry out age estimation to described facial image.
3. a kind of flash lamp control method estimated based on the age according to claim 2, is characterized in that: carry out age estimation to described facial image in described step 30, further comprising the steps:
31. collect sample images pedestrian's work of going forward side by side marks the age type of each sample image, and preset standard face figure;
The face characteristic key point of 32. comparison sample images and the face characteristic key point of standard faces figure, key point alignment and adjustment are carried out to sample image, again the facial contour figure that contours extract obtains sample image is carried out to the sample image after adjustment, and the facial contour figure of this sample image is normalized and inputs the training that convolutional neural networks system carries out disaggregated model;
33. obtain the facial image recognized from camera data, the face characteristic key point of this facial image of comparison and the face characteristic key point of standard faces figure, key point alignment and adjustment are carried out to described facial image, again the facial contour figure that contours extract obtains described facial image is carried out to the facial image after adjustment, and adopt described disaggregated model to carry out character classification by age in the facial contour figure of facial image, obtain the age type of described facial image.
4. a kind of flash lamp control method estimated based on the age according to claim 3, it is characterized in that: in the training process carrying out disaggregated model, the sample image of classification error collects and performs step 31 and step 32 successively by convolutional neural networks system, until be decided to be when exceeding expected results described step 32 the disaggregated model of training out be optimal classification model, and in described step 33, adopt the facial contour figure of this optimal classification model to described facial image to carry out character classification by age, obtain the age type of described facial image.
5. a kind of flash lamp control method estimated based on the age according to claim 3, is characterized in that: described age type comprises: infant, children, teenager, youth, middle age, old age.
6. a kind of flash lamp control method estimated based on the age according to claim 3, is characterized in that: described age type comprises: two years old with Types Below and more than two years old type.
7. a kind of flash lamp control method estimated based on the age according to claim 3, it is characterized in that: the face characteristic key point of comparison sample image and the face characteristic key point of standard faces figure in described step 32, key point alignment and adjustment are carried out to sample image, or, the face characteristic key point of facial image described in comparison and the face characteristic key point of standard faces figure in described step 33, key point alignment and adjustment are carried out to described facial image, mainly pass through the face characteristic key point of preset standard face figure and correspondence thereof, and according to the sample image obtained or the human face region of facial image and the face characteristic key point of correspondence thereof, the face characteristic key point of affine transformation to sample image or facial image is utilized to align and adjust, again to adjustment after sample image or facial image carry out contours extract be adjusted after facial contour figure, and using the facial contour figure of the facial contour figure after this adjustment as described sample image or facial image.
8. a kind of flash lamp control method estimated based on the age according to claim 3, it is characterized in that: described contours extract mainly carries out wavelet transformation to the sample image after described adjustment or facial image, obtain the frequency domain figure picture of low frequency component, namely described facial contour figure.
9. a kind of flash lamp control method estimated based on the age according to claim 3, it is characterized in that: adopt described disaggregated model to carry out in the facial contour figure of described facial image in described step 33 age type that character classification by age obtains described facial image, mainly the facial contour figure of described described facial image is put into the probability that convolutional neural networks system carries out calculating each age type of described facial image, and the maximum age type of select probability is as the age type of described facial image.
10. a kind of flash lamp control method estimated based on the age according to any one of claim 3 to 9, is characterized in that: described face characteristic key point mainly comprises eye contour, mouth, eyebrow, face mask line, forehead.
11. 1 kinds of device for controlling flashlight estimated based on the age, is characterized in that: it comprises
Camera model, for starting camera and catching the camera data of current preview;
Face recognition module, carries out recognition of face to described camera data, and obtains the facial image recognized;
Character classification by age module, it carries out age estimation to described facial image;
Executive Module, its according to the age estimated result judge described in facial image whether be infant, if then automatically close photoflash lamp, and drive camera to take pictures.
12. a kind of device for controlling flashlight estimated based on the age according to claim 11, is characterized in that: described character classification by age module comprises further:
Sampling unit, for collecting the age type of sample image for artificial each sample image of mark;
Edit cell, its face characteristic key point by the described facial image of sample image described in comparison or acquisition and the face characteristic key point of standard faces figure preset, key point alignment and adjustment are carried out to sample image or facial image, then the facial contour figure that contours extract obtains sample image or facial image is carried out to the sample image after adjustment or facial image;
Normalization unit, is normalized the facial contour figure of described sample image or facial image;
Unit, inputs the facial contour figure of the sample image after normalized or facial image the training that convolutional neural networks system carries out disaggregated model;
Judging unit, adopts described disaggregated model to carry out character classification by age in the facial contour figure of described facial image, to judge the age type of described facial image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410804921.3A CN104506778A (en) | 2014-12-22 | 2014-12-22 | Flashlight control method and device based on age estimation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410804921.3A CN104506778A (en) | 2014-12-22 | 2014-12-22 | Flashlight control method and device based on age estimation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104506778A true CN104506778A (en) | 2015-04-08 |
Family
ID=52948492
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410804921.3A Pending CN104506778A (en) | 2014-12-22 | 2014-12-22 | Flashlight control method and device based on age estimation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104506778A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105141854A (en) * | 2015-08-26 | 2015-12-09 | 广东欧珀移动通信有限公司 | Flashlight control method, flashlight control device and terminal equipment |
| CN105187720A (en) * | 2015-08-25 | 2015-12-23 | 广东欧珀移动通信有限公司 | Photographing method and device |
| CN105245787A (en) * | 2015-09-14 | 2016-01-13 | 努比亚技术有限公司 | Photographing control method and mobile terminal thereof |
| CN105791681A (en) * | 2016-02-29 | 2016-07-20 | 广东欧珀移动通信有限公司 | Control method, control device and electronic device |
| CN106131424A (en) * | 2016-07-26 | 2016-11-16 | 维沃移动通信有限公司 | A kind of method and apparatus of shooting |
| CN106203305A (en) * | 2016-06-30 | 2016-12-07 | 北京旷视科技有限公司 | Human face in-vivo detection method and device |
| CN106357989A (en) * | 2016-11-24 | 2017-01-25 | 维沃移动通信有限公司 | Brightness adjustment method and mobile terminal |
| CN106506948A (en) * | 2016-11-02 | 2017-03-15 | 北京小米移动软件有限公司 | Flash lamp control method and device |
| CN106878619A (en) * | 2017-03-15 | 2017-06-20 | 联想(北京)有限公司 | Parameter adjusting method, device and electronic equipment |
| WO2018133202A1 (en) * | 2017-01-22 | 2018-07-26 | 华为技术有限公司 | Flash light prompt method and terminal device |
| CN110179672A (en) * | 2019-05-30 | 2019-08-30 | 苏州大学附属儿童医院 | A kind of intelligent medicine box equipment and its method for carrying out medication managing |
| WO2021008214A1 (en) * | 2019-07-17 | 2021-01-21 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Intelligent flash intensity control systems and methods |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007208922A (en) * | 2006-02-06 | 2007-08-16 | Fujifilm Corp | Imaging apparatus |
| US20070268370A1 (en) * | 2001-09-18 | 2007-11-22 | Sanno Masato | Image pickup device, automatic focusing method, automatic exposure method, electronic flash control method and computer program |
| US20100007746A1 (en) * | 2008-07-09 | 2010-01-14 | Samsung Digital Imaging Co., Ltd. | Photography control method and apparatus for prohibiting use of flash |
| CN103824052A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Multilevel semantic feature-based face feature extraction method and recognition method |
| CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
-
2014
- 2014-12-22 CN CN201410804921.3A patent/CN104506778A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070268370A1 (en) * | 2001-09-18 | 2007-11-22 | Sanno Masato | Image pickup device, automatic focusing method, automatic exposure method, electronic flash control method and computer program |
| JP2007208922A (en) * | 2006-02-06 | 2007-08-16 | Fujifilm Corp | Imaging apparatus |
| US20100007746A1 (en) * | 2008-07-09 | 2010-01-14 | Samsung Digital Imaging Co., Ltd. | Photography control method and apparatus for prohibiting use of flash |
| CN103824052A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Multilevel semantic feature-based face feature extraction method and recognition method |
| CN103824054A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Cascaded depth neural network-based face attribute recognition method |
Non-Patent Citations (1)
| Title |
|---|
| JUNJUN0595: "基于仿射变换的人脸对齐的实现方法", 《HTTPS://WENKU.BAIDU.COM/VIEW/1F467329BD64783E09122BE1.HTML》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105187720A (en) * | 2015-08-25 | 2015-12-23 | 广东欧珀移动通信有限公司 | Photographing method and device |
| CN105141854A (en) * | 2015-08-26 | 2015-12-09 | 广东欧珀移动通信有限公司 | Flashlight control method, flashlight control device and terminal equipment |
| CN105245787A (en) * | 2015-09-14 | 2016-01-13 | 努比亚技术有限公司 | Photographing control method and mobile terminal thereof |
| CN105245787B (en) * | 2015-09-14 | 2019-04-02 | 努比亚技术有限公司 | Filming control method and its mobile terminal |
| CN105791681A (en) * | 2016-02-29 | 2016-07-20 | 广东欧珀移动通信有限公司 | Control method, control device and electronic device |
| CN105791681B (en) * | 2016-02-29 | 2019-05-03 | Oppo广东移动通信有限公司 | Control method, control device and electronic device |
| CN106203305B (en) * | 2016-06-30 | 2020-02-04 | 北京旷视科技有限公司 | Face living body detection method and device |
| CN106203305A (en) * | 2016-06-30 | 2016-12-07 | 北京旷视科技有限公司 | Human face in-vivo detection method and device |
| CN106131424A (en) * | 2016-07-26 | 2016-11-16 | 维沃移动通信有限公司 | A kind of method and apparatus of shooting |
| CN106506948A (en) * | 2016-11-02 | 2017-03-15 | 北京小米移动软件有限公司 | Flash lamp control method and device |
| CN106357989A (en) * | 2016-11-24 | 2017-01-25 | 维沃移动通信有限公司 | Brightness adjustment method and mobile terminal |
| WO2018133202A1 (en) * | 2017-01-22 | 2018-07-26 | 华为技术有限公司 | Flash light prompt method and terminal device |
| CN106878619A (en) * | 2017-03-15 | 2017-06-20 | 联想(北京)有限公司 | Parameter adjusting method, device and electronic equipment |
| CN106878619B (en) * | 2017-03-15 | 2019-10-29 | 联想(北京)有限公司 | Parameter adjusting method, device and electronic equipment |
| CN110179672A (en) * | 2019-05-30 | 2019-08-30 | 苏州大学附属儿童医院 | A kind of intelligent medicine box equipment and its method for carrying out medication managing |
| WO2021008214A1 (en) * | 2019-07-17 | 2021-01-21 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Intelligent flash intensity control systems and methods |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104506778A (en) | Flashlight control method and device based on age estimation | |
| CN104537630A (en) | Method and device for image beautifying based on age estimation | |
| CN109934115B (en) | Face recognition model construction method, face recognition method and electronic equipment | |
| CN103824050B (en) | A kind of face key independent positioning method returned based on cascade | |
| CN109815826B (en) | Method and device for generating face attribute model | |
| CN105740909B (en) | Text recognition method under a kind of natural scene based on spatial alternation | |
| CN110738161A (en) | face image correction method based on improved generation type confrontation network | |
| US9978119B2 (en) | Method for automatic facial impression transformation, recording medium and device for performing the method | |
| JP2019032821A (en) | Data augmentation techniques using style transformation with neural network | |
| WO2020007307A1 (en) | Sky filter method for panoramic images and portable terminal | |
| CN107146198A (en) | A kind of intelligent method of cutting out of photo and device | |
| CN104299263B (en) | A kind of method that cloud scene is modeled based on single image | |
| CN107871101A (en) | A kind of method for detecting human face and device | |
| CN111723654A (en) | High-altitude parabolic detection method and device based on background modeling, YOLOv3 and self-optimization | |
| CN111709290B (en) | Crowd counting method based on coding and decoding-jump connection scale pyramid network | |
| CN103942822A (en) | Facial feature point tracking and facial animation method based on single video vidicon | |
| CN103971112B (en) | Image characteristic extracting method and device | |
| CN109086768A (en) | The semantic image dividing method of convolutional neural networks | |
| CN109714519A (en) | A kind of method and system of adjust automatically image frame | |
| CN103413352A (en) | Scene three-dimensional reconstruction method based on RGBD multi-sensor fusion | |
| CN110263768A (en) | A kind of face identification method based on depth residual error network | |
| CN110472495B (en) | Deep learning face recognition method based on graphic reasoning global features | |
| CN106228575A (en) | Merge convolutional neural networks and the tracking of Bayesian filter and system | |
| CN109360179A (en) | A kind of image interfusion method, device and readable storage medium storing program for executing | |
| CN108257194B (en) | Face simple stroke generation method based on convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150408 |
|
| WD01 | Invention patent application deemed withdrawn after publication |