CN109359530A - Intelligent video monitoring method and device - Google Patents
Intelligent video monitoring method and device Download PDFInfo
- Publication number
- CN109359530A CN109359530A CN201811061880.8A CN201811061880A CN109359530A CN 109359530 A CN109359530 A CN 109359530A CN 201811061880 A CN201811061880 A CN 201811061880A CN 109359530 A CN109359530 A CN 109359530A
- Authority
- CN
- China
- Prior art keywords
- bgp
- face
- image
- coding
- obtains
- 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.)
- Granted
Links
Classifications
-
- 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
- G06V40/161—Detection; Localisation; Normalisation
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an intelligent video monitoring method and a device, wherein the method comprises the following steps: s1, monitoring a target video in real time, and detecting a face image from a video image in real time and outputting the face image; s2, coding the face image detected in the step S1 to obtain a code of the image to be detected; s3, respectively matching and comparing the codes of the images to be detected with database codes obtained from the face images of the database in advance to obtain and output identification results; the device comprises a human face real-time monitoring module, an image intelligent coding module and a human face rapid retrieval module. The invention can realize intelligent video monitoring, improve the intelligent level of video monitoring and greatly reduce the occupied space of video data storage and transmission.
Description
Technical field
The present invention relates to technical field of video monitoring more particularly to a kind of intelligent video monitoring methods and device.
Background technique
With currently increasingly improving to security requirement, video monitoring system effect is more obvious.Traditional video surveillance is more
For carrying out video capture, storage and playback, it is difficult to play the role of early warning and alarming, and monitor in real time and need artificially not stop to supervise
Video causes a large amount of manpowers and waste of time, while with the explosive growth of data volume, traditional monitoring mode can not yet
Meet the demand stored and transmitted to massive video data.Above-mentioned traditional video monitoring uses manual operation, post-mordem forensics
Outmoded mode, intelligence degree is low, and can occupy there are massive video data and largely store and transmit the various problems such as resource.
For the above problem of traditional monitoring mode, intelligent video monitoring comes into being, and intelligent video monitoring is also referred to as
Video Analysis Technology as carries out automatic understanding and analysis to video by artificial intelligence and machine vision.For video monitoring
Technology presently mainly uses following a few class methods: 1, carrying out target detection from background modeling angle;2, from single camera and more
Target following is carried out in terms of Camera location;3, Classification and Identification is carried out to target;4, it using Activity recognition algorithm, and is supervised in video
It is all usually at present to compare picture to be retrieved and all pictures in database one by one when in control for face retrieval, such
Method can generate biggish time loss, and actual retrieval efficiency is slower, required to be processed especially for massive video data
Image data amount is huge, and above-mentioned face retrieval method can not quickly identify target facial image from video monitoring, simultaneously should
Class method needs all pictures in storing data library, it is still desirable to biggish to store and transmit space.
Realize that method for quickly retrieving mainly includes following a few classes at present: 1) based on the search method of bag of words;2) it is based on
The search method of KD tree;3) search method based on vector clusters and quantization;4) based on the search method of Hash.Above-mentioned a few classes are fast
Fast search method usually realizes complexity, and the object of quick-searching is usually CIFAR-10, INRIA Holidays etc. with scene
Based on database picture, be generally unsuitable for realizing the quick-searching of facial image, and since face has variable property, it is right
The feature extraction and description of face are influenced vulnerable to expression, illumination etc., and human face posture, expression become in especially large-scale face database
Change is more violent, and it is difficult to extract stable description to carry out quick-searching, is applicable in current method for quickly retrieving directly to carry out people
Face quick-searching can not achieve accurate retrieval.Therefore, it is urgent to provide a kind of intelligent video monitoring method, reality is enabled to
Existing face quick-searching, it is horizontal to improve Intellectualized monitoring, while can reduce video data and store and transmit occupied space.
Summary of the invention
The technical problem to be solved in the present invention is that, for technical problem of the existing technology, the present invention provides one
Kind implementation method is simple, intelligence degree is high, monitoring precision is high and stores and transmits the intelligent video monitoring to occupy little space
Method and device.
In order to solve the above technical problems, technical solution proposed by the present invention are as follows:
A kind of intelligent video monitoring method, step include:
S1. face real-time monitoring: real-time monitoring is carried out to target video, real-time detection goes out facial image from video image
Output;
S2. image intelligent encodes: encoding to the step S1 facial image detected, obtains image to be checked and compile
Code;
S3. face quick-searching: by image to be checked coding respectively with the database that is obtained in advance by database facial image
Coding carries out matching comparison, obtains recognition result output.
As the further improvement of the method for the present invention, video image is used in the step S1 and is believed based on heuristic priori
The BGP face identification method of breath detects facial image, and step includes:
BGP is extracted: being based on BGP algorithm to image to be checked and is carried out BGP feature extraction, obtains being divided into after BGP characteristic image more
A sub-block counts the BGP histogram of each sub-block, and the statistic histogram of each sub-block of obtained each sub-block is spelled
It connects, obtains corresponding BGP feature vector;
Characteristic weighing: will be in the BGP feature vector according to the heuristic prior information containing human face structure characteristic information
The statistic histogram of part sub-block is weighted processing, the BGP feature vector that obtains that treated;
Recognition of face: treated that BGP feature vector is identified using described, exports face recognition result.
Further improvement as the method for the present invention: the heuristic prior information includes the position model where human face five-sense-organ
Information is enclosed, the position range information where the human face five-sense-organ is found out and face position from the BGP feature vector
The statistic histogram of corresponding target sub-block is weighted processing, according to the raising pair of the significance level of human face structure characteristic information
Answer the weight of sub-block statistic histogram.
Further improvement as the method for the present invention: the specific steps of the BGP feature vector that obtains that treated are as follows:
In the BGP characteristic image, the sub-block ordinal number in longitudinal M/6 to M/2, lateral N/5 to N*2/5 is counted, left eye position is obtained
Shared BGP block ordinal number is set, longitudinal M/6 to M/2, the sub-block ordinal number in the lateral region N*3/5 to N*4/5 is counted, obtains the right side
BGP block ordinal number shared by eye position counts longitudinal M/3 to M, the sub-block ordinal number in the lateral region N*2/5 to N*3/5, obtains
BGP block ordinal number shared by nose and mouth position, wherein M*N is the piecemeal number when BGP characteristic image carries out piecemeal, obtains institute
It states and finds out statistic histogram corresponding with BGP piecemeal ordinal number occupied by the face position after BGP feature vector and added
Power processing, the BGP feature vector that obtains that treated.
As the further improvement of the method for the present invention, the step of step S1, includes:
S11. video monitoring: the video information of real-time monitoring target video obtains the video information of real-time monitoring;
S12. Face datection: detect that facial image is exported from the video information of the real-time monitoring;
S13. face grabs: identifying face part from the facial image that the step S12 is exported and extracts, obtains
It is exported to human face target image;
S14. image enhancement: the human face target image is carried out include enhancing, denoising processing after export final people
Face image.
Further improvement as the method for the present invention: in the step S2, BGP algorithm is based on to image to be checked and carries out multistage
BGP feature extraction, input of the BGP characteristic image obtained by upper level BPG feature extraction as next stage BPG feature extraction,
And the feature vector obtained to every grade of BGP feature extraction encodes, and obtains the multi-layer coding for corresponding to BGP at different levels, each layer coding
Length shorten step by step form pyramid coding structure in order, obtain image to be checked coding.
As the further improvement of the method for the present invention, the step of step S3 includes: in advance to face each in database
Image is based respectively on BGP algorithm and successively carries out multistage BGP feature extraction, the BGP characteristic pattern obtained by upper level BPG feature extraction
As the input as next stage BPG feature extraction, and the feature vector obtained to every grade of BGP feature extraction encodes, and obtains
The multi-layer coding of corresponding BGP at different levels, the length of each layer coding shortens step by step in order forms pyramid coding structure, is counted
It is encoded according to library image;When carrying out face retrieval, the image coding to be checked is successively compared with database images coding
Compared with identifying target face by comparison result.
A kind of intelligent video monitoring apparatus, comprising:
Face real-time monitoring module, for carrying out real-time monitoring to target video, real-time detection goes out people from video image
Face image output;
Image intelligent coding module, the facial image for detecting to the face real-time monitoring module encode,
Obtain image coding to be checked;
Face quick-searching module, for by image to be checked coding respectively with the number that is obtained in advance by database facial image
Matching comparison is carried out according to library coding, obtains recognition result output.
Further improvement as apparatus of the present invention: the face real-time monitoring module includes face recognition module, is used for
Facial image, the recognition of face mould are detected using the BGP face identification method based on heuristic prior information to video image
Block includes:
BGP extraction unit carries out BGP feature extraction for being based on BGP algorithm to image to be checked, obtains BGP characteristic image
After be divided into multiple sub-blocks, count the BGP histogram of each sub-block, and by the statistics histogram of each sub-block of obtained each sub-block
Figure is spliced, and corresponding BGP feature vector is obtained;
Characteristic weighing unit, for according to the heuristic prior information containing human face structure characteristic information that the BGP is special
The statistic histogram of molecule block is weighted processing in the middle part of sign vector, the BGP feature vector that obtains that treated;
Face identification unit, for use it is described treated that BGP feature vector is identified, export recognition of face knot
Fruit.
Further improvement as apparatus of the present invention: in described image intelligently encoding module, image to be checked is calculated based on BGP
Method carries out multistage BGP feature extraction, and the BGP characteristic image obtained by upper level BPG feature extraction is mentioned as next stage BPG feature
The input taken, and the feature vector obtained to every grade of BGP feature extraction encodes, and obtains the multi-layer coding for corresponding to BGP at different levels,
The length of each layer coding shortens step by step in order forms pyramid coding structure, obtains image coding to be checked;
In the face quick-searching module, in advance to facial image each in database be based respectively on BGP algorithm successively into
Row multistage BGP feature extraction, the BGP characteristic image obtained by upper level BPG feature extraction is as next stage BPG feature extraction
Input, and the feature vector obtained to every grade of BGP feature extraction encodes, and obtains the multi-layer coding for corresponding to BGP at different levels, each layer
The length of coding shortens step by step form pyramid coding structure in order, obtains database images coding;Carry out face retrieval
When, compared with the image coding to be checked is carried out successively with database images coding, target person is identified by comparison result
Face.
Compared with the prior art, the advantages of the present invention are as follows:
1, intelligent video monitoring method and device of the present invention, by successively carry out face real-time monitoring, image intelligent coding,
Face quick-searching realizes automatic detection face, realizes intelligent monitoring, while by carrying out intelligently encoding to facial image, will be to
Inspection image coding is compared with database coding can be realized quick-searching, can not only improve face retrieval efficiency, moreover it is possible to
Enough downscaled video data significantly store and transmit occupied space, effectively promote video monitoring level of intelligence and efficiency, can satisfy
All kinds of real-time application requirements.
2, intelligent video monitoring method and device of the present invention further consider the structure feature of face, realize people in BGP
On the basis of face identification, carried out in conjunction with statistic histogram of the heuristic prior information of face to molecule block in the middle part of BGP feature vector
Weighting processing can be sufficiently sharp to improve the weight of corresponding sub-block statistic histogram according to the significance level of face characteristic information
Face intelligent recognition is realized with the heuristic information of face and priori knowledge, it can be in conjunction with BGP algorithm and heuristic information
The identification and robustness of next code are promoted under the premise of not increasing complexity, to improve Intellectualized monitoring level.
3, intelligent video monitoring method and device of the present invention, further by application layer union II system gradient mode to face
Image carries out multistage BGP, based on the mode of multistage BGP, can sufficiently excavate deeper marginal information, the texture letter of picture
More robust character representation is more stablized in the useful informations such as breath, gradient information, formation, so as to extract richer texture letter
Breath improves recognition of face precision, while being encoded by the feature vector to multistage BGP, realizes level coding, forms golden word
Tower coding is successively retrieved based on pyramid coding according to coding level when retrieval, may be implemented by fuzzy matching to smart
It really matches, by while capable of guaranteeing retrieval precision, effectively improving recall precision slightly to the retrieval of essence, realizes that face is quickly examined
Rope.
Detailed description of the invention
Fig. 1 is the implementation process schematic diagram of the present embodiment intelligent video monitoring method.
Fig. 2 is the implementation process schematic diagram based on heuristic prior information detection facial image in the present embodiment.
Fig. 3 is the implementation process schematic diagram that BGP feature extraction is carried out in the present embodiment.
Fig. 4 is the realization principle schematic diagram for dividing BGP image in the present embodiment based on heuristic prior information.
Fig. 5 is the facial image effect diagram detected in concrete application embodiment of the present invention.
Fig. 6 is the implementation process schematic diagram of intelligently encoding in the present embodiment.
Fig. 7 is the schematic illustration for the pyramid coding that the present embodiment constructs.
Fig. 8 is the principle schematic diagram of concrete application of the present invention intelligent video monitoring apparatus used in the examples.
Fig. 9 is the realization principle schematic diagram of the present embodiment intelligent video device.
Figure 10 is the realization principle schematic diagram that intelligent monitoring is realized in concrete application embodiment of the present invention.
Specific embodiment
Below in conjunction with Figure of description and specific preferred embodiment, the invention will be further described, but not therefore and
It limits the scope of the invention.
As shown in Figure 1, the present embodiment intelligent video monitoring method step includes:
S1. face real-time monitoring: real-time monitoring is carried out to target video, real-time detection goes out facial image from video image
Output;
S2. image intelligent encodes: encoding to the step S1 facial image detected, obtains image coding to be checked;
S3. face quick-searching: by image to be checked coding respectively with the database that is obtained in advance by database facial image
Coding carries out matching comparison, obtains recognition result output.
The present embodiment realizes automatic inspection by successively carrying out face real-time monitoring, image intelligent coding, face quick-searching
Face is surveyed, intelligent monitoring may be implemented, so that " ex-post analysis " is changed into " analysis in progress ", it is artificial to can solve traditional video surveillance
Operation, the outmoded mode of post-mordem forensics and massive video data occupy the problem of largely storing and transmitting resource, at the same by pair
Facial image carries out intelligently encoding, and image to be checked coding is encoded with database and is compared, it can be achieved that quick-searching goes out target,
Face retrieval efficiency can not only be improved, additionally it is possible to which downscaled video data significantly store and transmit occupied space, effectively promote view
Frequency monitoring intelligent level and efficiency.
As shown in Fig. 2, being known to video image using the BGP face based on heuristic prior information in the present embodiment step S1
Other method detects facial image, and step includes:
BGP is extracted: being based on BGP algorithm to image to be checked and is carried out BGP feature extraction, obtains being divided into after BGP characteristic image more
A sub-block counts the BGP histogram of each sub-block, and the statistic histogram of each sub-block of obtained each sub-block is spelled
It connects, obtains corresponding BGP feature vector;
Characteristic weighing: according to the heuristic prior information containing human face structure characteristic information by part in BGP feature vector
The statistic histogram of sub-block is weighted processing, the BGP feature vector that obtains that treated;
Recognition of face: using treated, BGP feature vector is identified, exports face recognition result.
The it is proposed of binary system gradient mode (BGP) is thought based on image gradient direction (IGO) and binary system describing mode
It is derived from the novel descriptor of Gradientfaces presumably, which replaces image pixel intensities come to people with image gradient direction (IGO)
Face is described, to realize that the robustness to illumination change, i.e., the aspect ratio that gradient field is extracted have more from the feature of intensity domain
There are identification and robustness.By measuring the relationship between the local pixel in image gradient domain in BGP algorithm, and by bottom office
Portion's structure efficient coding is one group of string of binary characters, not only increases judgement index, even more enormously simplifies computation complexity.For
Discovery gradient field potential structure, BGP are and to be encoded to a series of binary strings, energy from multi-directional computing image gradient
It enough indicates the variation of small boundary and texture information, therefore there is very strong identification, even if in face of blocking, illumination, expression shape change
Deng can also obtain preferable accuracy of identification, and when due to encoding using BGP to image, each encoded radio has contained neighbour
The information of domain pixel relationship, rather than just the strength information of pixel itself, therefore the image after BGP coding becomes various environment
Change more robust, especially there is stronger illumination invariant.
Although can obtain higher accuracy of identification in recognition of face based on BGP, BGP operator is as a kind of general
Operator, the BGP feature vector for being directly based upon extraction are identified, do not consider the special of face during identifying face
Property, cannot efficiently use the particularity information of face, and eyes in a such as face picture, nose, mouth face information is big
Body position be it is fixed, i.e., there are face specific some structural informations and heuristic information, such priori knowledge can make
It is applied in recognition of face for heuristic information with Statistical error performance.
The present embodiment considers the structure feature of face, heuristic in conjunction with face on the basis of BGP realizes recognition of face
Prior information is weighted processing to the statistic histogram of molecule block in the middle part of BGP feature vector, according to face characteristic information
Significance level improves the weight of corresponding sub-block statistic histogram, and the heuristic information and priori knowledge that can make full use of face are come
It realizes face intelligent recognition, can effectively improve accuracy of identification, efficiency and robust in conjunction with BGP algorithm and heuristic information
Property.
As shown in figure 3, the present embodiment carries out the step of BGP feature extraction are as follows:
Feature extraction is carried out to input image data based on BGP algorithm, obtains BGP characteristic image;
BGP characteristic image is divided into the sub-block not overlapped;
Count the BGP histogram of obtained each sub-block;
Obtained all sub-block histograms are spliced in order and obtain final BGP feature vector.
In concrete application embodiment, it is assumed that take radius of neighbourhood R=1, then BGP neighborhood territory pixel number P=8, piecemeal number
M*N (enables M=N=5), since 8 neighborhood territory pixels determine 8 directions (4 main, 4 auxiliary), then any one center
The BGP encoded radio of pixel should be tetrad, and being converted into decimal value should be 0 to 15, there is 8 kinds of structures in 16 kinds of modes
Mode ignores non-structural mode, then the BGP histogram dimension d=8 of each sub-block, and facial image is straight after single-stage BGP coding
Square figure dimension should be d1=200 (5*5*8).
After one width gray level image is carried out BGP feature extraction, a width BGP characteristic image is obtained.The present embodiment is specifically by BGP
Characteristic image carries out 5*5 piecemeal, and 25 sub-blocks put in order as from left to right, from top to bottom as shown in table 1.
Table 1:BGP characteristic image piecemeal sequence
A1 | A2 | A3 | A4 | A5 |
A6 | A7 | A8 | A9 | A10 |
A11 | A12 | A13 | A14 | A15 |
A16 | A17 | A18 | A19 | A20 |
A21 | A22 | A23 | A24 | A25 |
Statistic histogram can 8 kinds of tactic patterns occur in effecting reaction this sub-block the frequency, use XA1Indicate A1Son
The frequency that the statistic histogram of block, i.e. 8 kinds of tactic patterns occur, is expressed as 8 dimension row vectors, uses XAkIndicate AkThe statistics of sub-block
Histogram forms final vector P after being spliced the histogram vectors of all sub-blocks, as former gray level image is through binary system ladder
The feature vector that degree mode computation obtains, i.e., are as follows:
P=[XA1,XA2,...XAk,...XA25]
(1)
After obtaining features described above vector P, recycle heuristic prior information straight to the statistics of molecule block in the middle part of feature vector P
Fang Tu, to improve the weight of corresponding sub-block statistic histogram according to the significance level of human face structure characteristic information, after composition processing
Feature vector P, the corresponding power of human face structure characteristic item is improved or reduced according to significance level in treated feature vector P
Weight, characterization face characteristic that can be more accurate are subsequent that highly efficient identification may be implemented based on this feature vector.
In the present embodiment, heuristic prior information includes the position range information where human face five-sense-organ, according to human face five-sense-organ
The position range information at place finds out the statistic histogram of target sub-block corresponding with face position from BGP feature vector
It is weighted processing, to improve the weight of corresponding sub-block statistic histogram according to the significance level of human face structure characteristic information.
Face position is substantially stationary in face and is the key that recognition of face, and the heuristic prior information of the present embodiment is specifically wrapped
The position range information where human face five-sense-organ is included, is known to make full use of face information as heuristic prior information to improve face
Other performance.It is understood that heuristic prior information can also use or increase other structures characteristic information with further
Recognition performance is improved, weighting is then configured to improve or reduce weight with specific reference to significance level.Face is identified by using for reference the mankind
Shi Wuguan information is important and the relatively-stationary priori knowledge of face position and heuristic prior information, can make full use of people
The location information and structural information of face face are added five in conjunction with the BGP blocked histogram feature that BGP method is face position
Official's weight coefficient enhances face characterization ability, can effectively promote the intelligent level of next code, not increase complexity
Under the premise of be able to ascend the identification and robustness of next code.
It, can be in BGP feature using heuristic prior information specifically after BGP characteristic image is multiple sub-blocks according to M*N points
The approximate region of corresponding face position is found out in image, is counted in BGP characteristic image after BGP piecemeal occupied by face position
The BGP piecemeal ordinal number that corresponding face position can be obtained, the statistic histogram of the BGP piecemeal of the correspondence face position is added
Power processing, the BGP feature vector that can be obtained that treated is for subsequent identification.
After facial image longitudinal direction trisection, lateral five equal parts, usual eyes are at longitudinal intermediate one third, in cross
To second and the 4th 1/5th at;Nose and mouth are at longitudinal second and third one third, in transverse direction
Centre 1/5th at.After the present embodiment obtains BGP characteristic image, the experience according to above-mentioned face in face relative position is known
Knowledge and enlightening information, determine BGP piecemeal occupied by face position in conjunction with the piecemeal rule of BGP method, to corresponding
Statistic histogram is weighted processing.
At eyes position relatively first one third of longitudinal direction, in order to enhance face position determination
Robustness, the present embodiment finely tune the determination of face position, as shown in figure 4, wherein M=N=5, i.e. BGP characteristic image
When carrying out 5*5 piecemeal, by longitudinal six equal parts of BGP characteristic image, laterally five equal parts, eyes are six points a in longitudinal second and third
One of in region, in lateral second and the 4th 1/5th regions;Nose and mouth are in longitudinal third to the 6th six
In/mono- region, in lateral 1/5th region of centre, heuristic information is made of above- mentioned information to determine face position
Set corresponding sub-block.
In the present embodiment, the specific steps for the BGP feature vector that obtains that treated are as follows: in BGP characteristic image, statistics is vertical
To M/6 to M/2, lateral N/5 to N*2/5 in sub-block ordinal number, obtain BGP block ordinal number shared by left eye position, statistics is longitudinal
M/6 to M/2, the sub-block ordinal number in the lateral region N*3/5 to N*4/5, obtain BGP block ordinal number shared by right eye position, count
Sub-block ordinal number in the region N*2/5 to N*3/5 of longitudinal M/3 to M, transverse direction, obtains BGP block sequence shared by nose and mouth position
Number, wherein M*N is piecemeal numbers when BGP characteristic image carries out piecemeal, is found out after obtaining BGP feature vector and face position
The occupied corresponding statistic histogram of BGP piecemeal ordinal number is weighted processing, the BGP feature vector that obtains that treated.
In concrete application embodiment, M=N=5 can obtain BGP piecemeal ordinal number occupied by face position according to above-mentioned steps
It is above-mentioned face position in BGP block statistics histogram feature for A2, A7, A12, A4, A9, A14, A8, A13, A18, A23
The histogram feature of occupied BGP block is weighted, and weight coefficient size characterization computer thinks face information relative to face
The significance level of other positions, the BGP feature vector obtained after processing are as follows:
P=[XA1,w*XA2,XA3,...,XA6,w*XA7...
w*XA12,...,w*XA23,XA24,XA25] (2)
Wherein, w is weight coefficient.
Through the above steps, face information is even more important the present embodiment when identifying face using the mankind and relative position is fixed
Priori knowledge and heuristic prior information, be first depending on human experience and determine that face are distributed in the general location of face, then
Feature extraction is carried out to face using BGP operator, face is then divided into several sub-blocks and is counted to obtain the straight of each sub-block
Square figure feature is finally weighted the corresponding feature in face position, carries out people by weighting treated BGP feature vector
Face identification can effectively improve accuracy of identification and robustness.
In the present embodiment, the step of step S1, includes:
S11. video monitoring: the video information of real-time monitoring target video obtains the video information of real-time monitoring;
S12. Face datection: detect that facial image is exported from the video information of real-time monitoring;
S13. face grabs: identifying face part from the facial image that step S12 is exported and extracts, obtains people
The output of face target image;
S14. image enhancement: human face target image is carried out include enhancing, denoising processing after export final face figure
Picture.
By above-mentioned steps can real-time detection to clearly facial image, after the facial image that real-time detection is arrived carries out
It may recognize that target after continuous intelligently encoding, quick-searching.In concrete application embodiment, the people that is detected using the above method
Face image with accurate detection as shown in figure 5, as can be known from Fig. 5, the face image of face can be gone out by the above method.
In the present embodiment, in step S2, BGP algorithm is based on to image to be checked and carries out multistage BGP feature extraction, by upper level
Input of the BGP characteristic image that BPG feature extraction obtains as next stage BPG feature extraction, and every grade of BGP feature extraction is obtained
To feature vector encoded, obtain the multi-layer coding for corresponding to BGP at different levels, the length of each layer coding shortens shape step by step in order
At pyramid coding structure, image coding to be checked is obtained.
In the present embodiment, the step of step S3 include: in advance to facial image each in database be based respectively on BGP algorithm according to
Secondary progress multistage BGP feature extraction, the BGP characteristic image obtained by upper level BPG feature extraction are mentioned as next stage BPG feature
The input taken, and the feature vector obtained to every grade of BGP feature extraction encodes, and obtains the multi-layer coding for corresponding to BGP at different levels,
The length of each layer coding shortens step by step in order forms pyramid coding structure, obtains database images coding;Carry out face
When retrieval, compared with image to be checked coding is carried out successively with database images coding, target face is identified by comparison result.
The purpose of Face datection is realization accurate detection face in real time, for human face detection tech, real-time and standard
True property is a pair of amount to intercouple, it is difficult to while achieve the effect that ideal, consider in practical application, real-time should be the
One Consideration, while accuracy is easy impacted mainly since human face posture variation is different, the present embodiment consideration only detection
Positive face detects program to simplify, and the mode by cascading BGP is identified, can obtain higher accuracy rate.
The present embodiment carries out multistage BGP to facial image by application layer union II system gradient mode, based on multistage BGP's
Mode can sufficiently excavate the useful informations such as the deeper marginal information of picture, texture information, gradient information, be formed more
Add and stablize more robust character representation, so as to extract richer texture information, improves recognition of face precision, pass through simultaneously
The feature vector of multistage BGP is encoded, realizes level coding, pyramid coding is formed, the pyramid is based on when retrieval
Formula coding is successively retrieved according to coding level, may be implemented by fuzzy matching to accurate matching, by slightly to the retrieval of essence, Neng Goubao
While demonstrate,proving retrieval precision, recall precision is effectively improved, realizes face quick-searching.
When the present embodiment step S2 intelligently encoding, multistage BGP feature extraction is carried out to facial image, is by upper level BPG
Input of the BGP characteristic image that feature extraction obtains as next stage BPG feature extraction, it is final after multistage BGP feature extraction to obtain
The multiple feature vectors at different levels to correspondence, every grade of BPG feature extraction are as described above.It is specific first to extract BGP on the basis of original image
Feature obtains a width BGP characteristic pattern, which still has grayscale information, texture information, gradient information etc., again to BGP spy
Sign figure carries out a BGP feature extraction, can further extract more depth, more implicit, richer texture information and edge
The useful informations such as information, after successively carrying out feature extraction to coding characteristic image in this mode, available multistage BGP feature
Image;Every grade of BGP feature coding image is all used as to the input of single-stage BGP algorithm, then available multiple groups feature vector, finally
Multiple groups feature vector is encoded respectively, to form pyramid coding.
Facial image each in input database is subjected to a BGP feature extraction according to above-mentioned steps respectively first, is obtained
1st grade of BGP characteristic image and the 1st grade of BGP feature vector, then successively by (i-1)-th grade of BGP characteristic image according to above-mentioned steps into
BGP feature extraction of row obtains i-stage characteristic image and i-stage BGP feature vector;Image to be checked is pressed in step S2
Successively BGP feature extraction is carried out according to above-mentioned steps, obtains the 1st grade of BGP characteristic image and the 1st grade of BGP feature vector, then successively
(i-1)-th grade of BGP characteristic image is subjected to a BGP feature extraction according to above-mentioned steps, obtains i-stage characteristic image and i-th
Grade BGP feature vector, wherein i=2,3,4 ... n, n are the BGP feature extraction series of required execution.
When the feature vector that the present embodiment obtains every grade of BGP feature extraction encodes, by adjusting the BGP radius of neighbourhood
Or BGP piecemeal number shortens the BGP code length of each layer step by step, forms pyramid coding structure.Building coding pyramid
Purpose be formation length it is different level coding so that it is subsequent based on the different length level coding may be implemented by thick
To the quick-searching of essence.
The present embodiment coding pyramid construction process is as shown in fig. 6, by all regarding every grade of BGP characteristic image as single-stage
The input of BGP algorithm obtains multiple groups feature vector according to above-mentioned steps, multiple groups coding is obtained after being encoded respectively, with level
It is several litres high, by adjusting the field BGP radius or piecemeal number, shorten BGP code length successively, to realize coding pyramid
Building.Formed pyramid coding structure specifically predefine pyramidal number of plies n, according to image size select n group M, N and
R parameter, wherein M*N is BGP piecemeal quantity, and R is the BGP radius of neighbourhood, according to selected n group { Mi,Ni,RiCalculate n-layer coding
Pi, and the length of each layer coding is made to meet D1>D2>D3….>Dn, wherein i=1 ... n.
When the present embodiment constructs pyramid, code length reduces as coding number of levels increases, and code length is specifically pressed
Formula Di=Mi*Ni*di is calculated, wherein wherein Di is i-th layer of code length, Mi*Ni is BGP block count when i-stage encodes
Amount, di is the statistic histogram dimension of BGP each sub-block when i-stage encodes, and di is only related with Ri, therefore code length depends on
Mi, Ni, Ri tri- amounts, i.e. BGP piecemeal number and field radius.The present embodiment specifically takes Mi=Ni, Ri to take 1 or 2, in practical behaviour
In work, Ri=1 is specifically taken, and changes the value of Mi and Ni, both still keeps equal, realizes different levels coding in this way
Extraction, to complete to encode pyramidal building, the pyramid constructed was as shown in fig. 7, every layer of width indication of pyramid should
Layer code length, it can be seen that number of levels is higher, BGP code length is smaller.
It completes after encoding pyramidal building, quick-searching can be carried out based on pyramid coding.The present embodiment
In step S3, by image to be checked coding with database images coding successively from it is top to first layer progress successively compared with, every time
By the coding progress of the current layer of the coding of current layer and the last each image retrieved in image to be checked coding when comparing
With comparing, exported after retrieving multiple matched images.By encoding database images coding with image to be checked from top
It carries out successively coding to first layer to compare, by the comparison result that encodes every time come rapid drop range of search, so that background information
It is excluded rapidly, it is ensured that more calculation amounts are all applied in possible human face region, recall precision is effectively improved.
The specific steps of the present embodiment step S3 include:
S31. n-th layer coding Pn in image to be checked coding is compared with n-th layer coding Pn in database images coding,
Several pictures nearest with image distance to be checked in database are retrieved according to coding result, n is pyramid coding structure
The number of plies;
S32. by (n-1)th layer of coding of (n-1)th layer of coding in image to be checked coding and the last each image retrieved
Be compared, according to coding compare result retrieval go out in each image of last retrieval with nearest multiple of image distance to be checked
Figure, circulation execute step S32, and the comparison until completing first layer coding retrieves multiple most similar target images;
S33. image to be checked is compared with multiple most similar target images respectively, obtains final recognition result.
Through the above steps, it can be realized based on pyramid coding by obscuring accurate matching, quickly contracting is encoded by matching
Small seeking scope, then by accurately being matched in determining small range, face may be implemented and fast, accurately retrieve.
The present embodiment uses off-line operation for the pyramidal building of image coding in database, obtains database images volume
It is stored after code, only constructs coding pyramid online to image to be retrieved, the database images coding for recalling storage carries out
Matching, can be significantly by regarding the coding of the database pyramid containing great amount of images as off-line operation to realize final identification
Promote effectiveness of retrieval and real-time.
The present embodiment intelligent video monitoring apparatus includes:
Face real-time monitoring module, for carrying out real-time monitoring to target video, real-time detection goes out people from video image
Face image output;
Image intelligent coding module, for face real-time monitoring module monitors to facial image encode, obtain
Image coding to be checked;
Face quick-searching module, for by image to be checked coding respectively with the number that is obtained in advance by database facial image
Matching comparison is carried out according to library coding, obtains recognition result output.
In the present embodiment, face real-time monitoring module includes:
Video monitoring module obtains the video information of real-time monitoring for the video information of real-time monitoring target video;
Face recognition module, for identifying that facial image is exported from the video information of real-time monitoring;
Face handling module is obtained for identifying face part from the facial image that step S12 is exported and extracting
It is exported to human face target image;
Image enhancement module, for human face target image carry out include enhancing, denoising processing after export final people
Face image
The present embodiment face recognition module specifically uses the BGP recognition of face based on heuristic prior information to video image
Method detects facial image, comprising:
BGP extraction unit carries out BGP feature extraction for being based on BGP algorithm to image to be checked, obtains BGP characteristic image
After be divided into multiple sub-blocks, count the BGP histogram of each sub-block, and by the statistics histogram of each sub-block of obtained each sub-block
Figure is spliced, and corresponding BGP feature vector is obtained;
Characteristic weighing unit, for according to the heuristic prior information containing human face structure characteristic information by BGP feature to
The statistic histogram of amount middle part molecule block is weighted processing, the BGP feature vector that obtains that treated;
Face identification unit, for using that treated, BGP feature vector is identified, exports face recognition result.
In the present embodiment, in image intelligent coding module, BGP algorithm progress multistage BGP feature is based on to image to be checked and is mentioned
It takes, input of the BGP characteristic image obtained by upper level BPG feature extraction as next stage BPG feature extraction, and to every grade
The feature vector that BGP feature extraction obtains is encoded, and the multi-layer coding for corresponding to BGP at different levels is obtained, and the length of each layer coding is pressed
Sequence shortens form pyramid coding structure step by step, obtains image coding to be checked.
In the present embodiment, in face quick-searching module, BGP algorithm is based respectively on to facial image each in database in advance
Multistage BGP feature extraction is successively carried out, the BGP characteristic image obtained by upper level BPG feature extraction is as next stage BPG feature
The input of extraction, and the feature vector obtained to every grade of BGP feature extraction encodes, the multilayer for obtaining corresponding to BGP at different levels is compiled
Code, the length of each layer coding shortens step by step in order forms pyramid coding structure, obtains database images coding;Carry out people
When face is retrieved, compared with image to be checked coding is carried out successively with database images coding, target face is identified by comparison result.
The present embodiment intelligent video monitoring apparatus and above-mentioned intelligent video monitoring method correspond, and no longer go to live in the household of one's in-laws on getting married one by one again
It states.
The present invention realizes intelligent video device used by above-mentioned intelligent video monitoring method in concrete application embodiment
As shown in figure 8, include front-end and back-end, front end includes that face real-time monitoring module, image intelligent coding module, face are quickly examined
Rope module, face real-time monitoring module specifically include video reception module, face detection module, face handling module and image
Enhance module, rear end is database management module, and front-end and back-end data are transmitted by cloud network, and mass data can pass through
Cloud network storage, wherein video reception module is responsible for acquiring video from camera, face detection module by face in video into
Automatic detection, the face that face handling module will test individually extract row rapidly, and image enhancement module is responsible for face
Target Photo such as is enhanced, is denoised at the processing;Image intelligent coding module encodes facial image, traditional hard to solve
The excessively high problem of part resource cost;Face quick-searching module encodes for realizing target face coding with database face fast
Speed ratio pair.
By above structure, it can be achieved that face detect automatically, the functions such as image intelligent coding, face quick-searching, realize
The intelligence of video monitoring, and the level of intelligence of video monitoring can be promoted, save the hardware money of mass data storage and transmitting
Source.
As shown in figure 9, the present embodiment above-mentioned intelligent video device includes front-end operations, consistency operation and cloud behaviour when working
Make, specifically:
Front-end operations part: including widely distributed camera real-time monitoring multitude of video information, the video received is believed
Breath is handled in real time in intelligent chip, first progress face real-time monitoring, and the face frame occurred in video is extracted as to be checked
Ask picture;Then image enhancement processing is carried out to the face picture by image processing methods such as interpolation, denoisings;Again to obtaining
Target Photo carries out intelligently encoding, obtains succinct robust and the object code with identification;Object code is transmitted to by cloud network
Cloud.
Consistency operation part: storing the facial image of magnanimity perpetual object in face database, by all images with equally
Method carries out the processing such as image enhancement intelligently encoding, obtains coded data library;Again by all codings in entire coded data library
Cloud is transmitted to by cloud network.
Cloud operation part: after cloud receives the coding and object code of face database, according to object code from coded number
According to quick-searching in library, most like object to learn whether the target object belongs to face database realizes early warning therewith
The purpose of alarm.
Above-mentioned apparatus concrete application of the present invention is as shown in Figure 10, including four kinds of typical case scenes, and escaped criminal arrests and cruelly
Probably arrest in application, after face that front end surveillance device will test carries out image procossing and intelligently encoding, with backstage runaway convict or
Face database encoding ratio pair is feared cruelly, finds suspect and alarm;In hotel security protection and community security defence application, monitoring device
It is passed to cloud after the face coding that moment will test automatically, and is compared with the face coding being registered in database, sends out
Existing nonnative personnel generates early warning.
Above-mentioned only presently preferred embodiments of the present invention, is not intended to limit the present invention in any form.Although of the invention
It has been disclosed in a preferred embodiment above, however, it is not intended to limit the invention.Therefore, all without departing from technical solution of the present invention
Content, technical spirit any simple modifications, equivalents, and modifications made to the above embodiment, should all fall according to the present invention
In the range of technical solution of the present invention protection.
Claims (10)
1. a kind of intelligent video monitoring method, which is characterized in that step includes:
S1. face real-time monitoring: real-time monitoring is carried out to target video, it is defeated to go out facial image for real-time detection from video image
Out;
S2. image intelligent encodes: encoding to the step S1 facial image detected, obtains image coding to be checked;
S3. face quick-searching: image to be checked coding is encoded by the database that database facial image obtains respectively in advance
Matching comparison is carried out, recognition result output is obtained.
2. intelligent video monitoring method according to claim 1, which is characterized in that adopted in the step S1 to video image
Facial image is detected with the BGP face identification method based on heuristic prior information, step includes:
BGP is extracted: being based on BGP algorithm to image to be checked and is carried out BGP feature extraction, is divided into multiple sons after obtaining BGP characteristic image
Block counts the BGP histogram of each sub-block, and the statistic histogram of each sub-block of obtained each sub-block is spliced,
Obtain corresponding BGP feature vector;
Characteristic weighing: according to the heuristic prior information containing human face structure characteristic information by part in the BGP feature vector
The statistic histogram of sub-block is weighted processing, the BGP feature vector that obtains that treated;
Recognition of face: treated that BGP feature vector is identified using described, exports face recognition result.
3. intelligent video monitoring method according to claim 2, which is characterized in that the heuristic prior information includes people
Position range information where face face, the position range information where the human face five-sense-organ, from the BGP feature vector
In find out the statistic histogram of target sub-block corresponding with face position and be weighted processing, to be believed according to human face structure feature
The significance level of breath improves the weight of corresponding sub-block statistic histogram.
4. intelligent video monitoring method according to claim 3, which is characterized in that the BGP feature that obtains that treated
The specific steps of vector are as follows: in the BGP characteristic image, count in longitudinal M/6 to M/2, lateral N/5 to N*2/5
Sub-block ordinal number obtains BGP block ordinal number shared by left eye position, counts longitudinal M/6 to M/2, the lateral region N*3/5 to N*4/5
Interior sub-block ordinal number obtains BGP block ordinal number shared by right eye position, counts longitudinal M/3 to M, the lateral area N*2/5 to N*3/5
Sub-block ordinal number in domain obtains BGP block ordinal number shared by nose and mouth position, and wherein M*N is that the BGP characteristic image carries out piecemeal
When piecemeal number, obtain finding out after the BGP feature vector corresponding with BGP piecemeal ordinal number occupied by the face position
Statistic histogram be weighted processing, the BGP feature vector that obtains that treated.
5. intelligent video monitoring method described according to claim 1~any one of 4, which is characterized in that the step S1
The step of include:
S11. video monitoring: the video information of real-time monitoring target video obtains the video information of real-time monitoring;
S12. Face datection: detect that facial image is exported from the video information of the real-time monitoring;
S13. face grabs: identifying face part from the facial image that the step S12 is exported and extracts, obtains people
The output of face target image;
S14. image enhancement: the human face target image is carried out include enhancing, denoising processing after export final face figure
Picture.
6. intelligent video monitoring method described according to claim 1~any one of 4, which is characterized in that the step S2
In, BGP algorithm is based on to image to be checked and carries out multistage BGP feature extraction, the BGP feature obtained by upper level BPG feature extraction
Input of the image as next stage BPG feature extraction, and the feature vector obtained to every grade of BGP feature extraction encodes, and obtains
To the multi-layer coding of correspondence BGP at different levels, the length of each layer coding shortens step by step in order forms pyramid coding structure, obtains
Image coding to be checked.
7. intelligent video monitoring method according to claim 6, which is characterized in that the step of step S3 includes: pre-
BGP algorithm first is based respectively on to facial image each in database and successively carries out multistage BGP feature extraction, by upper level BPG feature
Input of the obtained BGP characteristic image as next stage BPG feature extraction is extracted, and to the spy that every grade of BGP feature extraction obtains
Sign vector is encoded, and the multi-layer coding for corresponding to BGP at different levels is obtained, and the length of each layer coding shortens step by step in order forms golden word
Tower coding structure obtains database images coding;When carrying out face retrieval, by the image coding to be checked and the database
Image coding is successively compared, and identifies target face by comparison result.
8. a kind of intelligent video monitoring apparatus characterized by comprising
Face real-time monitoring module, for carrying out real-time monitoring to target video, real-time detection goes out face figure from video image
As output;
Image intelligent coding module, the facial image for detecting to the face real-time monitoring module are encoded, are obtained
Image coding to be checked;
Face quick-searching module, for by image to be checked coding respectively with the database that is obtained in advance by database facial image
Coding carries out matching comparison, obtains recognition result output.
9. intelligent video monitoring apparatus according to claim 8, which is characterized in that the face real-time monitoring module includes
Face recognition module, for detecting face figure using the BGP face identification method based on heuristic prior information to video image
Picture, the face recognition module include:
BGP extraction unit carries out BGP feature extraction for being based on BGP algorithm to image to be checked, divides after obtaining BGP characteristic image
For multiple sub-blocks, count the BGP histogram of each sub-block, and by the statistic histogram of each sub-block of obtained each sub-block into
Row splicing, obtains corresponding BGP feature vector;
Characteristic weighing unit, for according to the heuristic prior information containing human face structure characteristic information by the BGP feature to
The statistic histogram of amount middle part molecule block is weighted processing, the BGP feature vector that obtains that treated;
Face identification unit, for use it is described treated that BGP feature vector is identified, export face recognition result.
10. intelligent video monitoring apparatus according to claim 8 or claim 9, which is characterized in that described image intelligently encoding module
In, BGP algorithm is based on to image to be checked and carries out multistage BGP feature extraction, the BGP feature obtained by upper level BPG feature extraction
Input of the image as next stage BPG feature extraction, and the feature vector obtained to every grade of BGP feature extraction encodes, and obtains
To the multi-layer coding of correspondence BGP at different levels, the length of each layer coding shortens step by step in order forms pyramid coding structure, obtains
Image coding to be checked;
In the face quick-searching module, in advance to facial image each in database be based respectively on BGP algorithm successively carry out it is more
Grade BGP feature extraction, the BGP characteristic image obtained by upper level BPG feature extraction is as the defeated of next stage BPG feature extraction
Enter, and the feature vector obtained to every grade of BGP feature extraction encodes, obtain the multi-layer coding for corresponding to BGP at different levels, each layer is compiled
The length of code shortens step by step form pyramid coding structure in order, obtains database images coding;When carrying out face retrieval,
Compared with the image coding to be checked is carried out successively with database images coding, target face is identified by comparison result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811061880.8A CN109359530B (en) | 2018-09-12 | 2018-09-12 | Intelligent video monitoring method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811061880.8A CN109359530B (en) | 2018-09-12 | 2018-09-12 | Intelligent video monitoring method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109359530A true CN109359530A (en) | 2019-02-19 |
CN109359530B CN109359530B (en) | 2022-01-25 |
Family
ID=65350968
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811061880.8A Active CN109359530B (en) | 2018-09-12 | 2018-09-12 | Intelligent video monitoring method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109359530B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109948700A (en) * | 2019-03-19 | 2019-06-28 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating characteristic pattern |
CN111898416A (en) * | 2020-06-17 | 2020-11-06 | 绍兴埃瓦科技有限公司 | Video stream processing method and device, computer equipment and storage medium |
CN117062288A (en) * | 2023-08-10 | 2023-11-14 | 广东正力通用电气有限公司 | Single-lamp monitoring system and method based on Internet of things |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103839033A (en) * | 2012-11-20 | 2014-06-04 | 广东工业大学 | Face identification method based on fuzzy rule |
CN104008395A (en) * | 2014-05-20 | 2014-08-27 | 中国科学技术大学 | Intelligent bad video detection method based on face retrieval |
CN105046224A (en) * | 2015-07-16 | 2015-11-11 | 东华大学 | Block self-adaptive weighted histogram of orientation gradient feature based face recognition method |
CN106777349A (en) * | 2017-01-16 | 2017-05-31 | 广东工业大学 | Face retrieval system and method based on deep learning |
-
2018
- 2018-09-12 CN CN201811061880.8A patent/CN109359530B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103839033A (en) * | 2012-11-20 | 2014-06-04 | 广东工业大学 | Face identification method based on fuzzy rule |
CN104008395A (en) * | 2014-05-20 | 2014-08-27 | 中国科学技术大学 | Intelligent bad video detection method based on face retrieval |
CN105046224A (en) * | 2015-07-16 | 2015-11-11 | 东华大学 | Block self-adaptive weighted histogram of orientation gradient feature based face recognition method |
CN106777349A (en) * | 2017-01-16 | 2017-05-31 | 广东工业大学 | Face retrieval system and method based on deep learning |
Non-Patent Citations (3)
Title |
---|
LIWEI LI等: ""Face Recognition Algorithm Based on Cascading BGP Feature Fusion"", 《2018 CHINESE CONTROL AND DECISION CONFERENCE (CCDC)》 * |
周霞等: "《Shearlet 多方向特征融合与加权直方图的人脸识别算法》", 《光电工程》 * |
黄凯奇等: ""图像物体分类与检测算法综述"", 《计算机学报》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109948700A (en) * | 2019-03-19 | 2019-06-28 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating characteristic pattern |
CN111898416A (en) * | 2020-06-17 | 2020-11-06 | 绍兴埃瓦科技有限公司 | Video stream processing method and device, computer equipment and storage medium |
CN117062288A (en) * | 2023-08-10 | 2023-11-14 | 广东正力通用电气有限公司 | Single-lamp monitoring system and method based on Internet of things |
Also Published As
Publication number | Publication date |
---|---|
CN109359530B (en) | 2022-01-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Huang et al. | Detection algorithm of safety helmet wearing based on deep learning | |
CN104778481B (en) | A kind of construction method and device of extensive face pattern analysis sample storehouse | |
CN104408429B (en) | A kind of video represents frame extracting method and device | |
CN112491796B (en) | Intrusion detection and semantic decision tree quantitative interpretation method based on convolutional neural network | |
CN113011319A (en) | Multi-scale fire target identification method and system | |
CN113537099B (en) | Dynamic detection method for fire smoke in highway tunnel | |
CN103839065A (en) | Extraction method for dynamic crowd gathering characteristics | |
CN111241343A (en) | Road information monitoring and analyzing detection method and intelligent traffic control system | |
CN105426903A (en) | Cloud determination method and system for remote sensing satellite images | |
CN107133569A (en) | The many granularity mask methods of monitor video based on extensive Multi-label learning | |
CN109359530A (en) | Intelligent video monitoring method and device | |
CN107169106A (en) | Video retrieval method, device, storage medium and processor | |
CN107222660A (en) | A kind of distributed network visual monitor system | |
CN111476160A (en) | Loss function optimization method, model training method, target detection method, and medium | |
CN111191531A (en) | Rapid pedestrian detection method and system | |
CN106886771A (en) | The main information extracting method of image and face identification method based on modularization PCA | |
CN117119253B (en) | High-quality video frame extraction method for target object | |
CN111881803A (en) | Livestock face recognition method based on improved YOLOv3 | |
CN114581769A (en) | Method for identifying houses under construction based on unsupervised clustering | |
CN112115794A (en) | Method for detecting wearing condition of safety helmet based on deep learning | |
Yi et al. | Research on fruit recognition method based on improved yolov4 algorithm | |
Cheng et al. | Research on Fast Target Detection And Classification Algorithm for Passive Millimeter Wave Imaging | |
Anandhi | Edge Computing-Based Crime Scene Object Detection from Surveillance Video Using Deep Learning Algorithms | |
He et al. | A Deep Learning Method for Detecting Phone Call Behaviors of Bidding Evaluation Expert | |
CN112052336B (en) | Traffic emergency identification method and system based on social network platform information |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |