CN110163092A - Demographic method, device, equipment and storage medium based on recognition of face - Google Patents
Demographic method, device, equipment and storage medium based on recognition of face Download PDFInfo
- Publication number
- CN110163092A CN110163092A CN201910297454.2A CN201910297454A CN110163092A CN 110163092 A CN110163092 A CN 110163092A CN 201910297454 A CN201910297454 A CN 201910297454A CN 110163092 A CN110163092 A CN 110163092A
- Authority
- CN
- China
- Prior art keywords
- image
- portrait
- cutting
- recognition
- result
- 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
- 238000000034 method Methods 0.000 title claims abstract description 54
- 239000000284 extract Substances 0.000 claims abstract description 12
- 230000002159 abnormal effect Effects 0.000 claims description 32
- 230000006835 compression Effects 0.000 claims description 6
- 238000007906 compression Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 238000013473 artificial intelligence Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 229910002056 binary alloy Inorganic materials 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000001144 postural effect Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000011426 transformation method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of demographic method based on recognition of face, device, equipment and storage mediums, this method comprises: obtaining video image from video flowing when receiving demographics instruction；The video image is subjected to picture cutting, obtains multiple cutting images, and extract the image LBP feature of multiple cutting images；By described image LBP feature input Identification of Images model trained in advance, described image LBP feature is identified by the Identification of Images model, exports recognition result；The number for counting the cutting image that recognition result is portrait, obtains the first number statistical result.The present invention is based on artificial intelligence, are counted using image processing techniques to the number in video, thus greatly improve the efficiency and accuracy of demographics.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of demographic methods based on recognition of face, dress
It sets, equipment and storage medium.
Background technique
At present when needing to carry out demographics to regions such as meeting room, station, markets, generally requires and manually check people
Number, or demographics are obtained indirectly as a result, leading to demographics inefficiency, and statistical result is not by other methods
It is enough accurate.
Summary of the invention
The present invention provides a kind of demographic method based on recognition of face, device, equipment and storage medium, it is intended to improve
The efficiency and accuracy of demographics.
To achieve the above object, the present invention provides a kind of demographic method based on recognition of face, which comprises
When receiving demographics instruction, video image is obtained from video flowing；
The video image is subjected to picture cutting, obtains multiple cutting images, and extract multiple cutting images
Image LBP feature；
By described image LBP feature input Identification of Images model trained in advance, by the Identification of Images model to described
Image LBP feature is identified, recognition result is exported；
The number for counting the cutting image that recognition result is portrait, obtains the first number statistical result.
Preferably, the Identification of Images model that the input of described image LBP feature is trained in advance, by the Identification of Images
Before the step of model identifies described image LBP feature, exports recognition result further include:
The sample image for collecting preset quantity, sets portrait or non-portrait for the label of the sample image；
The sample image is compressed into after 128 × 128 pixels and carries out gray proces and random incomplete processing again, at acquisition
Sample image after reason；
The sample LBP feature of sample image after the processing is extracted, sample LBP feature is obtained；
The sample LBP feature is inputted in the neural network created based on TensorFlow and is trained, described in acquisition
The recognition result of Identification of Images model, the Identification of Images model output is portrait or non-portrait.
Preferably, the cutting image includes the first cutting image and the second cutting image, described by the video image
Picture cutting is carried out, the step of obtaining multiple cutting images includes:
By the video image compression at the compressed video image of 512 × 512 pixels；
The compressed video image is subjected to picture cutting by 64 × 64 pixels, obtains multiple first cutting images；
The overlapping region of cutting image adjacent in the first cutting image is subjected to secondary picture by starting point of 64 pixels
Cutting obtains the second cutting image.
Preferably, the step of image LBP feature for extracting multiple cutting images includes:
The cutting image is divided into multiple regions；
By the center gray value of each of each region pixel 8 neighbor pixels adjacent with the pixel
Gray value be compared, obtain the LBP feature of the pixel；
LBP feature based on the pixel, obtains the histogram in each region；
Acquisition statistic histogram is normalized to the histogram in each region, is based on the statistic histogram
Obtain the image LBP feature of the cutting image.
Preferably, the statistics recognition result is the number of the cutting image of portrait, obtains the first number statistical result
After step further include:
Report interface that the first number statistical result is reported to server according to preset result.
Preferably, described to report interface that the first number statistical result is reported to server according to preset result
Before step further include:
Judge in the first number statistical result whether to include abnormal portrait according to peak count method；
If not including the abnormal portrait in the first number statistical result, then follow the steps: according to preset result
Report interface that the first number statistical result is reported to server；
If including the abnormal portrait in the first number statistical result, in the first number statistical result
After the number of the abnormal portrait, the second demographics are obtained as a result, the second demographics result is reported to described
Server.
Preferably, the Identification of Images model record recognition result is that the portrait of portrait in the cutting image of portrait is sat
Mark, it is described to judge whether include the steps that abnormal portrait includes: in the first number statistical result according to peak count method
Obtain the number that the portrait coordinate occurs within a preset time；
If the number that the portrait coordinate occurs within a preset time is greater than or equal to frequency threshold value, described image is determined
The corresponding portrait of coordinate is not abnormal portrait；
If the number that the portrait image coordinate occurs within a preset time is less than frequency threshold value, determine that described image is sat
Marking corresponding portrait is abnormal portrait, then the corresponding portrait of described image coordinate is labeled as abnormal portrait.
To achieve the above object, the embodiment of the present invention also provides a kind of people counting device based on recognition of face, described
People counting device based on recognition of face includes:
Module is obtained, for obtaining video image from video flowing when receiving demographics instruction；
Extraction module obtains multiple cutting images, and extract described more for the video image to be carried out picture cutting
The image LBP feature of a cutting image；
Identification module is known for the Identification of Images model that the input of described image LBP feature is trained in advance by the portrait
Other model identifies described image LBP feature, exports recognition result；
Statistical module obtains the first demographics for counting the number for the cutting image that recognition result is portrait
As a result.
To achieve the above object, the embodiment of the present invention also provides a kind of people-counting equipment based on recognition of face, described
People-counting equipment based on recognition of face includes processor, memory and storage being known based on face in the memory
Other number statistics program realizes institute as above when the demographics program based on recognition of face is run by the processor
The step of demographic method based on recognition of face stated.
To achieve the above object, the embodiment of the present invention also provides a kind of computer storage medium, and the computer storage is situated between
The demographics program based on recognition of face is stored in matter, the demographics program based on recognition of face is transported by processor
The step of demographic method based on recognition of face as described above is realized when row.
Compared with prior art, it a kind of demographic method based on recognition of face proposed by the present invention, device, equipment and deposits
Storage media, this method comprises: obtaining video image from video flowing when receiving demographics instruction；By the video figure
As carrying out picture cutting, multiple cutting images are obtained, and extract the image LBP feature of the multiple cutting image；By the figure
As LBP feature input Identification of Images model trained in advance, described image LBP feature is known by the Identification of Images model
Not, recognition result is exported；The number for counting the cutting image that recognition result is portrait, obtains the first number statistical result.This hair
It is bright to be based on artificial intelligence, the number in video is counted using image processing techniques, thus greatly improves demographics
Efficiency and accuracy.
Detailed description of the invention
Fig. 1 is the hardware structural diagram for the people-counting equipment based on recognition of face that various embodiments of the present invention are related to；
Fig. 2 is the flow diagram of the demographic method first embodiment the present invention is based on recognition of face；
Fig. 3 is the flow diagram of the demographic method second embodiment the present invention is based on recognition of face；
Fig. 4 is the functional block diagram of the people counting device first embodiment the present invention is based on recognition of face.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
The people-counting equipment based on recognition of face that the embodiment of the present invention relates generally to, which refers to, can be realized network connection
Network access device, the people-counting equipment based on recognition of face can be server, cloud platform etc..
Referring to Fig.1, Fig. 1 is the hardware configuration for the people-counting equipment based on recognition of face that various embodiments of the present invention are related to
Schematic diagram.In the embodiment of the present invention, the people-counting equipment based on recognition of face may include (such as the centre of processor 1001
Manage device Central Processing Unit, CPU), communication bus 1002, input port 1003, output port 1004, storage
Device 1005.Wherein, communication bus 1002 is for realizing the connection communication between these components；Input port 1003 is defeated for data
Enter；Output port 1004 is exported for data, and memory 1005 can be high speed RAM memory, be also possible to stable storage
Device (non-volatile memory), such as magnetic disk storage, memory 1005 optionally can also be independently of aforementioned processing
The storage device of device 1001.It will be understood by those skilled in the art that hardware configuration shown in Fig. 1 is not constituted to of the invention
It limits, may include perhaps combining certain components or different component layouts than illustrating more or fewer components.
With continued reference to Fig. 1, the memory 1005 in Fig. 1 as a kind of readable storage medium storing program for executing may include operating system, net
Network communication module, application program module and the demographics program based on recognition of face.In Fig. 1, network communication module master
It is used to connect server, carries out data communication with server；And processor 1001 can be called and be stored in memory 1005
Demographics program based on recognition of face, and execute the demographics side provided in an embodiment of the present invention based on recognition of face
Method.
The embodiment of the invention provides a kind of demographic methods based on recognition of face.
Referring to Fig. 2, Fig. 2 is the flow diagram of the demographic method first embodiment the present invention is based on recognition of face.
In the present embodiment, the demographic method based on recognition of face is applied to the demographics based on recognition of face
Equipment, which comprises
Step S101 obtains video image when receiving demographics instruction from video flowing；
In the present embodiment, camera is installed in the demographics region for needing to carry out demographics in advance, is taken the photograph by described
As head images the demographics region, acquisition obtains in real time and saves the video flowing.Such as in meeting room certain
A camera, the information such as the indoor scene of shooting meeting, personnel are installed in a position, and save active conference room video flowing.
When receiving the demographics instruction that user is occurred by voice or touch control operation, then view is obtained from video flowing
Frequency image.It is to be appreciated that the demographics instruction includes time point, when the time point can be current time, history
Between and reservation future time.In general, the video flowing has timestamp, obtained and the time according to the timestamp
The corresponding video image of point.
The video image is carried out picture cutting, obtains multiple cutting images, and extract the multiple cut by step S102
The image LBP feature of partial image；
In the present embodiment, needs to carry out secondary cutting to the video image, obtain the first cutting image and the second cutting
Image.Specifically, the described the step of video image is carried out picture cutting, obtains multiple cutting images, includes:
Step S102-1a, by the video image compression at the compressed video image of 512 × 512 pixels；
The video image is compressed, the compressed video image of 512 × 512 pixels is obtained.It is to be appreciated that
In other embodiments, the video image can be compressed by other pixels.
The compressed video image is carried out picture cutting by 64 × 64 pixels, obtains multiple first and cut by step S102-1b
Partial image；
First time cutting is carried out to the compressed video image: the compressed video image is cut by 64 × 64 pixels
Point, obtain multiple first cutting images.
Step S102-1c, by the overlapping region of cutting image adjacent in the first cutting image using 64 pixels as starting point
Secondary picture cutting is carried out, the second cutting image is obtained.
Inevitably, overlapping region is had after image cutting, therefore again by cutting image adjacent in the first cutting image
Overlapping region carries out secondary picture cutting by starting point of 64 pixels, obtains the second cutting image.
In the present embodiment, the cutting image includes the first cutting image and the second cutting image.
In the present embodiment, LBP (Local Binary Patterns, local binary patterns) is a kind of for describing image
The operator of the feature of local grain has the characteristics that gray scale invariance.
Further, the step of image LBP feature for extracting multiple cutting images includes:
Step S102-2a: the cutting image is divided into multiple regions；
The cutting image is divided into the multiple regions of default size, such as is divided into 16 × 16 multiple regions.
Step S102-2b: by the center gray value of each of each region pixel it is adjacent with the pixel 8
The gray value of a neighbor pixel is compared, and obtains the LBP feature of the pixel；
Specifically, if the neighbor grayscale value is greater than the center gray value, the position of the neighbor pixel is marked
It is denoted as 1；If the neighbor grayscale value is less than or equal to the center gray value, the position mark by the neighbor pixel is
0；Being compared in this way with 8 points in the neighborhood of 3*3 can produce 8 bits and (usually can be exchanged into decimal number, i.e.,
LBP feature, the LBP value are the integer between 1-256), thus to obtain the LBP feature of the pixel.
Step S102-2c: the LBP feature based on the pixel obtains the histogram in each region；
After the LBP feature for obtaining each pixel in the region, the LBP feature of each pixel is counted
It can be obtained the histogram in each region.
Step S102-2d: being normalized acquisition statistic histogram to the histogram in each region, is based on institute
State the image LBP feature that statistic histogram obtains the cutting image.
Generally, when expressing image texture using LBP, only focus on Uniform mode, and by other all modes return to
In same class, the image after normalizing as a result, can more embody the texture of each representative region, while desalinate smooth region again
Feature.In the present embodiment, acquisition statistic histogram is normalized to the histogram in each region, is based on the system
Meter histogram obtains the image LBP feature of the cutting image.
Into one, in order to allow the LBP feature that there is rotational invariance, binary system is rotated, such as at the beginning
The initial LBP feature arrived is 10010000, then the initial characteristics can be converted to as after being rotated clockwise
The decimal value of 00001001 minimum value form, the minimum value form described in this way is minimum namely LBP is minimum.No matter described cut
How partial image can rotate, and the LBP is minimum, it is possible thereby to guarantee that LBP has rotational invariance.
Step S103, by described image LBP feature input Identification of Images model trained in advance, by the Identification of Images mould
Type identifies described image LBP feature, exports recognition result；
In the present embodiment, the step S103: the Identification of Images model that the input of described image LBP feature is trained in advance,
Before the step of being identified by the Identification of Images model to described image LBP feature, exporting recognition result further include:
Step S103a collects the sample image of preset quantity, sets portrait or inhuman for the label of the sample image
Picture；
The sample image includes portrait sample image and inhuman picture sample image, and the portrait sample image includes
Face sample image and people's upper part of the body sample image.
In the present embodiment, 100,000 face sample images are collected, collect 50,000 people's upper part of the body sample images,
Portrait is set by the label of 100,000 face sample images and 50,000 people's upper part of the body sample images.Collect 10,000
The non-portrait image, and non-portrait is set by the label of 10,000 non-portrait images.
It is to be appreciated that using people's upper part of the body image as training sample, can in the video image face quilt
When blocking, demographics are carried out according to the feature of people's upper part of the body image, statistical result can be prevented inaccurate, lack number
The generation of thing.Also, by non-portrait image also as training sample, then the Identification of Images model after can making training identifies
Non- portrait keeps statistical result more accurate.
The sample image is compressed into after 128 × 128 pixels and carries out gray proces and random incompleteness again by step S103b
Processing, sample image after being handled；
In the present embodiment, the sample image is compressed into 128 × 128 pixels first, obtains compression samples image.Then
The compression samples image is subjected to gray proces by one of image inversion, logarithmic transformation method, obtains gray scale sample
Image.The gray scale sample image is subjected to random incomplete processing using image repair method again, the sample that obtains that treated
This image.
Step S103c extracts the sample LBP feature of sample image after the processing, obtains sample LBP feature；
Sample image after the processing is divided into multiple sample areas, by each of each sample areas sample picture
The gray value of center of a sample's gray value of vegetarian refreshments 8 sample neighbor pixels adjacent with the sampled pixel point is compared,
Obtain the sample LBP feature of the sampled pixel point；Based on the LBP feature of the sampled pixel point, each sample areas is obtained
Sample histogram；The sample histogram of each sample areas is normalized and obtains statistical sample histogram,
The sample image LBP feature of the sample image is obtained based on the sample statistics histogram.
The sample LBP feature is inputted in the neural network created based on TensorFlow and is instructed by step S103d
Practice, obtains the Identification of Images model, the recognition result of the Identification of Images model output is portrait or non-portrait.
The TensorFlow is a kind of open source code machine learning frame, and TensorFlow is widely used in all kinds of
The programming of machine learning algorithm is realized.Developer can be helped to construct model in extreme code using TensorFlow, and
Should model make required product.
In the present embodiment, the sample LBP feature is inputted in the neural network created based on TensorFlow and is instructed
Practice, after repetition training up to a million times, the sample LBP feature can accurately be divided according to its label for corresponding to sample image
Class, thus to obtain the Identification of Images model, the recognition result of Identification of Images model output is portrait or non-portrait, namely
The recognition result of the sample image of label behaviour picture in the sample image is exported as portrait, by label in the sample image
Recognition result output for the sample image of non-portrait is non-portrait.
Step S104, statistics recognition result are the number of the cutting image of portrait, obtain the first number statistical result.
According to the recognition result that the Identification of Images model exports, statistics recognition result is the cutting image of portrait
Number, using the number of the cutting image as the first number statistical result.
The present embodiment through the above scheme, when receiving demographics instruction, obtains video image from video flowing；It will
The video image carries out picture cutting, obtains multiple cutting images, and the image LBP for extracting the multiple cutting image is special
Sign；By described image LBP feature input Identification of Images model trained in advance, by the Identification of Images model to described image
LBP feature is identified, recognition result is exported；The number for counting the cutting image that recognition result is portrait, obtains the first number
Statistical result.It is based on artificial intelligence as a result, the number in video is counted using image processing techniques, is greatly improved
The efficiency and accuracy of demographics.
As shown in figure 3, second embodiment of the invention proposes a kind of demographic method based on recognition of face, based on above-mentioned
First embodiment shown in Fig. 2, the statistics recognition result are the number of the cutting image of portrait, obtain the first demographics knot
After the step of fruit further include:
Step S106: report interface that the first number statistical result is reported to server according to preset result.
Specifically, it presets and reports interface, it is described to report interface for communicating with the server network.It is understood that
Ground reports interface can also will be in the corresponding camera information of the first number statistical result, area information, temporal information etc.
It reports to the server.
The step S106: report interface that the first number statistical result is reported to server according to preset result
The step of before further include:
Judge in the first number statistical result whether to include abnormal portrait according to peak count method；
It is to be appreciated that since the video image in video flowing is real-time change, therefore obtained from the video flowing
Video image may be not sufficiently stable, and may be walked about personnel, be caused first portrait to count due to postural change etc.
As a result not accurate enough.
Specifically, the number that the portrait coordinate occurs within a preset time is obtained；
The video image is extracted from the video flowing by preset duration, the preset duration can be 100ms,
The video image is carried out picture cutting, obtains multiple cutting images, and extract the multiple cutting image by 200ms etc.
Image LBP feature；By described image LBP feature input Identification of Images model trained in advance, by the Identification of Images model pair
Described image LBP feature is identified, and exports the portrait coordinate of portrait in the cutting image.
If the number that the portrait coordinate occurs within a preset time is greater than or equal to frequency threshold value, described image is determined
The corresponding portrait of coordinate is not abnormal portrait；In the present embodiment, the preset time be can be 1 minute, and the frequency threshold value can
Be 4 times, it is 10 inferior, if such as the portrait coordinate occurs in 1 minute number is 10 times, illustrate described image coordinate pair
The portrait answered is not abnormal portrait.If the number that the portrait image coordinate occurs within a preset time is less than frequency threshold value,
Determine that the corresponding portrait of described image coordinate is abnormal portrait, then the corresponding portrait of described image coordinate is labeled as abnormal people
Picture.If not including the abnormal portrait in the first number statistical result, thens follow the steps: being reported and connect according to preset result
The first number statistical result is reported to server by mouth；If in the first number statistical result including the abnormal people
Picture after the number for then removing the abnormal portrait in the first number statistical result, obtains the second demographics as a result, will
The second demographics result reports to the server.2 abnormal portraits if it exists, then subtract first statistical result
The second demographics result is then obtained after going 2.
The present embodiment through the above scheme, when receiving demographics instruction, obtains video image from video flowing；It will
The video image carries out picture cutting, obtains cutting multiple images, and the image LBP for extracting multiple cutting images is special
Sign；By described image LBP feature input Identification of Images model trained in advance, by the Identification of Images model to described image
LBP feature is identified, recognition result is exported；The number for counting the cutting image that recognition result is portrait, obtains the first number
Statistical result；Report interface that the first number statistical result is reported to server according to preset result.It is based on people as a result,
Work intelligence, counts the number in video using image processing techniques, greatly improves the efficiency of demographics and accurate
Property.
In addition, the present embodiment also provides a kind of people counting device based on recognition of face.It is the present invention referring to Fig. 4, Fig. 4
The functional block diagram of people counting device first embodiment based on recognition of face.
People counting device provided by the invention based on recognition of face is a kind of virtual bench, is stored in shown in FIG. 1
In the memory 1005 of people-counting equipment based on recognition of face, to realize the institute of the demographics program based on recognition of face
It is functional: for obtaining video image from video flowing when receiving demographics instruction；For by the video image into
Row picture cutting obtains multiple cutting images, and extracts the image LBP feature of multiple cutting images；For by the figure
As LBP feature input Identification of Images model trained in advance, described image LBP feature is known by the Identification of Images model
Not, recognition result is exported；For counting the number for the cutting image that recognition result is portrait, the first demographics knot is obtained
Fruit.
Specifically, the people counting device described in the present embodiment based on recognition of face includes:
Module 10 is obtained, for obtaining video image from video flowing when receiving demographics instruction；
Extraction module 20 obtains multiple cutting images, and extract multiple for the video image to be carried out picture cutting
The image LBP feature of the cutting image；
Identification module 30, for the Identification of Images model that the input of described image LBP feature is trained in advance, by the portrait
Identification model identifies described image LBP feature, exports recognition result；
Statistical module 40 obtains the first number system for counting the number for the cutting image that recognition result is portrait
Count result.
Into one, the identification module is also used to:
The sample image for collecting preset quantity, sets portrait or non-portrait for the label of the sample image；
The sample image is compressed into after 128 × 128 pixels and carries out gray proces and random incomplete processing again, at acquisition
Sample image after reason；
The sample LBP feature of sample image after the processing is extracted, sample LBP feature is obtained；
The sample LBP feature is inputted in the neural network created based on TensorFlow and is trained, described in acquisition
The recognition result of Identification of Images model, the Identification of Images model output is portrait or non-portrait.
Into one, the extraction module is also used to:
By the video image compression at the compressed video image of 512 × 512 pixels；
The compressed video image is subjected to picture cutting by 64 × 64 pixels, obtains multiple first cutting images；
The overlapping region of cutting image adjacent in the first cutting image is subjected to secondary picture by starting point of 64 pixels
Cutting obtains the second cutting image.
Into one, the extraction module is also used to:
The cutting image is divided into multiple regions；
By the center gray value of each of each region pixel 8 neighbor pixels adjacent with the pixel
Gray value be compared, obtain the LBP feature of the pixel；
LBP feature based on the pixel, obtains the histogram in each region；
Acquisition statistic histogram is normalized to the histogram in each region, is based on the statistic histogram
Obtain the image LBP feature of the cutting image.
Into one, the statistical module is also used to:
Report interface that the first number statistical result is reported to server according to preset result.
Into one, the statistical module is also used to:
Judge in the first number statistical result whether to include abnormal portrait according to peak count method；
If not including the abnormal portrait in the first number statistical result, then follow the steps: according to preset result
Report interface that the first number statistical result is reported to server；
If including the abnormal portrait in the first number statistical result, in the first number statistical result
After the number of the abnormal portrait, the second demographics are obtained as a result, the second demographics result is reported to described
Server.
Into one, the statistical module is also used to:
Obtain the number that the portrait coordinate occurs within a preset time；
If the number that the portrait coordinate occurs within a preset time is greater than or equal to frequency threshold value, described image is determined
The corresponding portrait of coordinate is not abnormal portrait；
If the number that the portrait image coordinate occurs within a preset time is less than frequency threshold value, determine that described image is sat
Marking corresponding portrait is abnormal portrait, then the corresponding portrait of described image coordinate is labeled as abnormal portrait.
In addition, being stored in the computer storage medium based on people the present invention also provides a kind of computer storage medium
The demographics program of face identification is realized as described above when the demographics program based on recognition of face is run by processor
The step of demographic method based on recognition of face, details are not described herein.
Compared with prior art, it a kind of demographic method based on recognition of face proposed by the present invention, device, equipment and deposits
Storage media, this method comprises: obtaining video image from video flowing when receiving demographics instruction；By the video figure
As carrying out picture cutting, multiple cutting images are obtained, and extract the image LBP feature of multiple cutting images；By the figure
As LBP feature input Identification of Images model trained in advance, described image LBP feature is known by the Identification of Images model
Not, recognition result is exported；The number for counting the cutting image that recognition result is portrait, obtains the first number statistical result.This hair
It is bright to be based on artificial intelligence, the number in video is counted using image processing techniques, thus greatly improves demographics
Efficiency and accuracy.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or the system that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or system institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, method of element, article or system.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that a terminal device executes the present invention respectively
Method described in a embodiment.
The above description is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all utilizations
Equivalent structure made by description of the invention and accompanying drawing content or process transformation, are applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of demographic method based on recognition of face, which is characterized in that the described method includes:
When receiving demographics instruction, video image is obtained from video flowing；
The video image is subjected to picture cutting, obtains multiple cutting images, and extract the image of the multiple cutting image
Local binary patterns LBP feature；
By described image LBP feature input Identification of Images model trained in advance, by the Identification of Images model to described image
LBP feature is identified, recognition result is exported；
The number for counting the cutting image that recognition result is portrait, obtains the first number statistical result.
2. the method according to claim 1, wherein described input training in advance for described image LBP feature
The step of Identification of Images model identifies described image LBP feature by the Identification of Images model, exports recognition result it
Before further include:
The sample image for collecting preset quantity, sets portrait or non-portrait for the label of the sample image；
The sample image is compressed into after 128 × 128 pixels and carries out gray proces and random incomplete processing again, after being handled
Sample image；
The sample LBP feature of sample image after the processing is extracted, sample LBP feature is obtained；
The sample LBP feature is inputted in the neural network created based on TensorFlow and is trained, the portrait is obtained
The recognition result of identification model, the Identification of Images model output is portrait or non-portrait.
3. the method according to claim 1, wherein the cutting image includes that the first cutting image and second is cut
Partial image, described that the video image is carried out picture cutting, the step of obtaining multiple cutting images, includes:
By the video image compression at the compressed video image of 512 × 512 pixels；
The compressed video image is subjected to picture cutting by 64 × 64 pixels, obtains multiple first cutting images；
The overlapping region of cutting image adjacent in the first cutting image is subjected to secondary picture cutting by starting point of 64 pixels,
Obtain the second cutting image.
4. the method according to claim 1, wherein the image LBP for extracting multiple cutting images is special
The step of sign includes:
The cutting image is divided into multiple regions；
By the ash of the center gray value of each of each region pixel 8 neighbor pixel adjacent with the pixel
Angle value is compared, and obtains the LBP feature of the pixel；
LBP feature based on the pixel, obtains the histogram in each region；
Acquisition statistic histogram is normalized to the histogram in each region, is obtained based on the statistic histogram
The image LBP feature of the cutting image.
5. the method according to claim 1, wherein the statistics recognition result is of the cutting image of portrait
After the step of number, the first number statistical result of acquisition further include:
Report interface that the first number statistical result is reported to server according to preset result.
6. according to the method described in claim 5, it is characterized in that, described report interface by described first according to preset result
Before the step of demographics result reports to server further include:
Judge in the first number statistical result whether to include abnormal portrait according to peak count method；
If not including the abnormal portrait in the first number statistical result, thens follow the steps: being reported according to preset result
The first number statistical result is reported to server by interface；
If including the abnormal portrait in the first number statistical result, institute is removed in the first number statistical result
After the number for stating abnormal portrait, the second demographics are obtained as a result, the second demographics result is reported to the service
Device.
7. method according to claim 1 to 6, which is characterized in that the Identification of Images model record identification knot
Fruit is the portrait coordinate of portrait in the cutting image of portrait, described to judge first demographics according to peak count method
As a result whether include the steps that abnormal portrait includes: in
Obtain the number that the portrait coordinate occurs within a preset time；
If the number that the portrait coordinate occurs within a preset time is greater than or equal to frequency threshold value, described image coordinate is determined
Corresponding portrait is not abnormal portrait；
If the number that the portrait image coordinate occurs within a preset time is less than frequency threshold value, described image coordinate pair is determined
The portrait answered is abnormal portrait, then the corresponding portrait of described image coordinate is labeled as abnormal portrait.
8. a kind of people counting device based on recognition of face, which is characterized in that the demographics dress based on recognition of face
It sets and includes:
Module is obtained, for obtaining video image from video flowing when receiving demographics instruction；
Extraction module obtains multiple cutting images, and extract the multiple cut for the video image to be carried out picture cutting
The image LBP feature of partial image；
Identification module, for the Identification of Images model that the input of described image LBP feature is trained in advance, by the Identification of Images mould
Type identifies described image LBP feature, exports recognition result；
Statistical module obtains the first number statistical result for counting the number for the cutting image that recognition result is portrait.
9. a kind of people-counting equipment based on recognition of face, which is characterized in that the demographics based on recognition of face are set
Standby includes processor, the demographics program based on recognition of face of memory and storage in the memory, the base
When the demographics program of recognition of face is run by the processor, such as base of any of claims 1-7 is realized
In the demographic method of recognition of face the step of.
10. a kind of computer storage medium, which is characterized in that be stored in the computer storage medium based on recognition of face
Demographics program is realized when the demographics program based on recognition of face is run by processor as in claim 1-7
The step of demographic method described in any one based on recognition of face.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910297454.2A CN110163092A (en) | 2019-04-12 | 2019-04-12 | Demographic method, device, equipment and storage medium based on recognition of face |
| PCT/CN2019/122079 WO2020207038A1 (en) | 2019-04-12 | 2019-11-29 | People counting method, apparatus, and device based on facial recognition, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910297454.2A CN110163092A (en) | 2019-04-12 | 2019-04-12 | Demographic method, device, equipment and storage medium based on recognition of face |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110163092A true CN110163092A (en) | 2019-08-23 |
Family
ID=67639302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910297454.2A Pending CN110163092A (en) | 2019-04-12 | 2019-04-12 | Demographic method, device, equipment and storage medium based on recognition of face |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110163092A (en) |
| WO (1) | WO2020207038A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111199215A (en) * | 2020-01-06 | 2020-05-26 | 郑红 | People counting method and device based on face recognition |
| WO2020207038A1 (en) * | 2019-04-12 | 2020-10-15 | 深圳壹账通智能科技有限公司 | People counting method, apparatus, and device based on facial recognition, and storage medium |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822111B (en) * | 2021-01-19 | 2024-05-24 | 北京京东振世信息技术有限公司 | Crowd detection model training method and device and crowd counting method and device |
| CN114462653B (en) * | 2022-01-24 | 2022-09-30 | 广东天地和实业控股集团有限公司 | O2O type campus self-service catering management method, system, equipment and medium |
| CN116128883A (en) * | 2023-04-19 | 2023-05-16 | 尚特杰电力科技有限公司 | Photovoltaic panel quantity counting method and device, electronic equipment and storage medium |
| CN117636425A (en) * | 2023-11-23 | 2024-03-01 | 南京石三心网络科技有限公司 | Face recognition system based on information technology |
| CN118379778B (en) * | 2024-06-25 | 2024-08-23 | 贵州聚控云科技有限公司 | Personnel information statistics method and system based on face recognition |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778512A (en) * | 2016-11-25 | 2017-05-31 | 南京蓝泰交通设施有限责任公司 | Face identification method under the conditions of a kind of unrestricted based on LBP and depth school |
| CN109359548A (en) * | 2018-09-19 | 2019-02-19 | 深圳市商汤科技有限公司 | Plurality of human faces identifies monitoring method and device, electronic equipment and storage medium |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| MX2012013865A (en) * | 2012-11-29 | 2014-05-28 | Vision Holdings Mexico S De R L De C V | System and method for paying and counting passengers in transport means. |
| KR101500496B1 (en) * | 2013-12-06 | 2015-03-10 | 주식회사 케이티 | Apparatus for recognizing face and method thereof |
| CN105160313A (en) * | 2014-09-15 | 2015-12-16 | 中国科学院重庆绿色智能技术研究院 | Method and apparatus for crowd behavior analysis in video monitoring |
| CN108563997B (en) * | 2018-03-16 | 2021-10-12 | 新智认知数据服务有限公司 | Method and device for establishing face detection model and face recognition |
| CN108805140B (en) * | 2018-05-23 | 2021-06-29 | 国政通科技股份有限公司 | LBP-based rapid feature extraction method and face recognition system |
| CN109344765A (en) * | 2018-09-28 | 2019-02-15 | 广州云从人工智能技术有限公司 | A kind of intelligent analysis method entering shop personnel analysis for chain shops |
| CN110163092A (en) * | 2019-04-12 | 2019-08-23 | 深圳壹账通智能科技有限公司 | Demographic method, device, equipment and storage medium based on recognition of face |
-
2019
- 2019-04-12 CN CN201910297454.2A patent/CN110163092A/en active Pending
- 2019-11-29 WO PCT/CN2019/122079 patent/WO2020207038A1/en active Application Filing
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778512A (en) * | 2016-11-25 | 2017-05-31 | 南京蓝泰交通设施有限责任公司 | Face identification method under the conditions of a kind of unrestricted based on LBP and depth school |
| CN109359548A (en) * | 2018-09-19 | 2019-02-19 | 深圳市商汤科技有限公司 | Plurality of human faces identifies monitoring method and device, electronic equipment and storage medium |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020207038A1 (en) * | 2019-04-12 | 2020-10-15 | 深圳壹账通智能科技有限公司 | People counting method, apparatus, and device based on facial recognition, and storage medium |
| CN111199215A (en) * | 2020-01-06 | 2020-05-26 | 郑红 | People counting method and device based on face recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020207038A1 (en) | 2020-10-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163092A (en) | Demographic method, device, equipment and storage medium based on recognition of face | |
| CN110826538B (en) | Abnormal off-duty identification system for electric power business hall | |
| CN110119673B (en) | Non-inductive face attendance checking method, device, equipment and storage medium | |
| CN110070101A (en) | Floristic recognition methods and device, storage medium, computer equipment | |
| CN112116582B (en) | Method for detecting and identifying cigarettes in inventory or display scene | |
| CN110807775A (en) | Traditional Chinese medicine tongue image segmentation device and method based on artificial intelligence and storage medium | |
| CN109559362B (en) | Image subject face replacing method and device | |
| CN110675395A (en) | Intelligent on-line monitoring method for power transmission line | |
| CN109815936A (en) | A kind of target object analysis method and device, computer equipment and storage medium | |
| CN111091106B (en) | Image clustering method and device, storage medium and electronic device | |
| CN112926472A (en) | Video classification method, device and equipment | |
| CN113505780A (en) | Two-dimensional code-based intelligent detection maintenance method and equipment | |
| CN110796039B (en) | Face flaw detection method and device, electronic equipment and storage medium | |
| CN114241012B (en) | High-altitude parabolic determination method and device | |
| CN116385918A (en) | Personnel working state evaluation method and system based on intelligent recognition | |
| CN106503747A (en) | A kind of image recognition statistical analysis system | |
| CN111353439A (en) | Method, device, system and equipment for analyzing teaching behaviors | |
| CN107368847B (en) | Crop leaf disease identification method and system | |
| CN113469135A (en) | Method and device for determining object identity information, storage medium and electronic device | |
| CN110414322B (en) | Method, device, equipment and storage medium for extracting picture | |
| CN111800428A (en) | Real-time statistical method and system for digital conference participation | |
| CN112580616A (en) | Crowd quantity determination method, device, equipment and storage medium | |
| CN117541442A (en) | Teaching attendance management method, device, equipment and storage medium | |
| CN116959113A (en) | Gait recognition method and device | |
| CN111797922A (en) | Text image classification method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190823 |
|
| WD01 | Invention patent application deemed withdrawn after publication |