CN109714526A - Intelligent video camera head and control system - Google Patents
Intelligent video camera head and control system Download PDFInfo
- Publication number
- CN109714526A CN109714526A CN201811402218.4A CN201811402218A CN109714526A CN 109714526 A CN109714526 A CN 109714526A CN 201811402218 A CN201811402218 A CN 201811402218A CN 109714526 A CN109714526 A CN 109714526A
- Authority
- CN
- China
- Prior art keywords
- image
- camera head
- video camera
- intelligent video
- data
- 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
Landscapes
- Image Analysis (AREA)
- Studio Devices (AREA)
Abstract
The disclosure provides a kind of intelligent video camera head and the control system comprising it.Wherein intelligent video camera head includes: Image intake device, for shooting or receiving image and/or video containing object；And processing unit, comprising: data preprocessing module is chosen from the image and/or video of shooting and meets the image to impose a condition or video frame；Object extraction module is used for target detection, obtains at least part of target image of the object or object in image；Identification module identifies target image, identifies the mark data for distinguishing object.In the disclosure, by the way that several modules handled using neural network are arranged in intelligent video camera head, processing speed can be improved.
Description
Technical field
The present invention relates to this disclosure relates to technical field of information processing, and in particular to a kind of intelligent video camera head, and comprising
The control system of the intelligent video camera head.
Background technique
There is the occasion of demand to intelligent video camera head, there can be several insoluble problems.For example, in large parking lot,
It often can not find the specific stand of the vehicle of oneself, can only help by key and multidigit staff and find guidance, it can
To handle the parking stall problem found and parked in parking lot with intelligent video camera head help；It is found in the place containing a large amount of personages specific
Personnel, can only also be carried out substantially by means of estimation mode at present；In numerous animals and plants, the animals and plants of particular species are found
Deng can only also be carried out by means of the mode of range estimation.
Summary of the invention
(1) technical problems to be solved
In view of this, the purpose of the present invention is to provide a kind of intelligent video camera heads, and the control comprising the intelligent video camera head
System processed.
(2) technical solution
According to the one side of the disclosure, a kind of intelligent video camera head is provided, comprising:
Image intake device, for shooting or receiving image and/or video containing object；And
Processing unit, comprising: data preprocessing module is chosen from the image and/or video of shooting and meets setting condition
Image or video frame；Object extraction module is used for target detection, and the object or object in acquisition image are at least
Partial target image；Identification module identifies target image, identifies the mark data for distinguishing object.
In a further embodiment, processing unit further include: enhancing module, by the image for meeting setting condition
Either the resolution ratio of video frame be the image of first resolution or video frame handled to obtain the image of second resolution or
Video frame, the second resolution are higher than the first resolution.
In a further embodiment, setting condition includes: from the image and/or video of shooting, and extracting has
The image or video frame of the above difference of given threshold.
In a further embodiment, extracting the video frame with the above difference of given threshold includes: by artificial
Neural network processes video, specifically includes: taking current video frame T, video frame carries out feature extraction through convolution, most passes through afterwards
The output layer of neural network obtains a scoring fT, as the score of the video frame, represent the feature of the frame, fTWith f0Compare, f0
It is initialized as 0, if difference is greater than given threshold, data of the frame as subsequent module, by fTIt is assigned to f0, remove a video
Frame T+1, repeats the above, until completing all video frames.
In a further embodiment, object behaviour, animal, plant, natural object or object is manually manufactured.
In a further embodiment, manually manufacture object is automobile, and at least partly image of the object includes vapour
Vehicle license plate.
In a further embodiment, in object extraction module, be used for target detection, obtain image in object or
At least part of target image of person's object, comprising: by artificial neural network, wherein piece image or video frame are read,
Obtain at least part of target image of object or object.
In a further embodiment, it includes: logical for obtaining object or at least part of target image of object
Artificial neural network is crossed, candidate region is generated using selection searching algorithm, picture is divided into many zonules, passes through level point
Group method merges according to similarity, obtains at least part of boundary candidate frame of object or object；To boundary candidate frame
Using sliding window method, according at least part of scale of object or object as window size in boundary upper slide of frame
It is dynamic, obtain at least part of object region of object or object.
In a further embodiment, mark data includes following at least one: pattern, Chinese character, letter, number and symbol
Number.
In a further embodiment, in identification module, target image is identified, identifies and distinguishes object
Mark data, comprising: by artificial neural network, position picture identification data, identified respectively: all of image are extracted
Candidate frame is sized to each candidate frame and adapts to artificial neural network input size, obtained by convolutional neural networks
Characteristic pattern, in being input to sorter network, the sorter network energy identification feature figure is final to obtain the mark to be obtained in original graph
Data information.
In a further embodiment, processing unit includes neural network processor, integrates the data prediction mould
Block, object extraction module, data processing module and at least one for enhancing module.
In a further embodiment, neural network processor includes: storage unit, for storing the input number
According to neural network parameter and instruction；Control unit for reading special instruction from the storage unit, and is decoded into
Arithmetic element instructs and is input to arithmetic element；Arithmetic element is corresponding for being executed according to arithmetic element instruction to the data
Neural network computing, obtain output neuron.
In a further embodiment, in arithmetic element, executing corresponding neural network computing includes: by input nerve
Member is multiplied with weight data, obtains multiplied result；Add tree operation is executed, for the multiplied result to be passed through add tree step by step
It is added, obtains weighted sum, weighted sum biasing is set or is not processed；The weighted sum set or be not processed to biasing executes activation letter
Number operation, obtains output neuron.
In a further embodiment, processor further include: pretreatment unit, the image for being absorbed to video camera
And/or video data is pre-processed, and face recognition result is converted into, which is to meet neural network to input lattice
The data of formula；And/or direct memory access DMA, the input data for being stored in storage unit, neural network parameter and refer to
It enables, so that control unit and arithmetic element are called.
In a further embodiment, processor further includes following at least one: instruction buffer, is used for from described direct
Memory access DMA cache instruction is called for control unit；Neuron caching is inputted, for slow from the direct memory access DMA
Input neuron is deposited, is called for arithmetic element；Weight caching, for caching weight from the direct memory access DMA, for operation
Cell call；And output neuron caching, the output neuron after operation is obtained for storing from the arithmetic element, with defeated
Out to direct memory access DMA.
In a further embodiment, instruction buffer, input neuron caching, weight caching and output neuron caching
For on piece caching.
According to another aspect of the present disclosure, a kind of control system is provided, comprising: a kind of intelligent video camera head of any of the above, institute
The quantity configuration of intelligent video camera head is stated as one setting place of camera shooting covering；
Control terminal receives the mark data of each intelligent video camera head processing, determines that mark data corresponds to object and setting
Determine the position in place.
In a further embodiment, control terminal further include: display device and/or instantaneous speech power, for exporting
Location information of the object that control terminal determines in setting place.
(3) beneficial effect
By providing the control system containing intelligent video camera head, be conducive to the management in parking lot, or to the unrest on road
The vehicle for stopping leaving about is identified；
By the way that several modules handled using neural network are arranged in intelligent video camera head, processing speed is improved；
By handling several modular concurrents execution of intelligent video camera head, the speed of processing is further speeded up.
Detailed description of the invention
Fig. 1 is the intelligent video camera head schematic illustration of the embodiment of the present invention.
Fig. 2 is the functional-block diagram of processing unit in Fig. 1.
Fig. 3 is the neural network processor functional-block diagram in the processing unit of Fig. 2.
Fig. 4 is the intelligent video camera head application scenarios schematic diagram of the embodiment of the present invention.
Fig. 5 is the control system schematic diagram of the embodiment of the present invention.
Specific embodiment
Below with reference to the attached drawing in the embodiment of the present disclosure, the technical solution in the embodiment of the present disclosure is carried out clear, complete
Ground description, it is clear that described embodiment is only disclosure a part of the embodiment, instead of all the embodiments.Based on this
Disclosed embodiment, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, belongs to the protection scope of the disclosure.
According to the basic conception of the disclosure, a kind of intelligent video camera head, including Image intake device and processing unit, shadow are provided
As capturing apparatus is used for for shooting or receiving image and/or video frame containing object, processing unit from Image intake device
It is chosen in the image and/or video of shooting and meets the image to impose a condition or video frame, then obtained in image or video frame
Object or object at least part of target image；And identify the mark data for distinguishing object.Pass through the intelligence
Can camera by operation, can identify that the unlabeled data of object distinguishes object, (such as unite so that the later period is further processed
Meter analysis, lookup object etc.).
Fig. 1 is 100 schematic illustration of intelligent video camera head of the embodiment of the present invention.As shown in Figure 1, intelligent video camera head 100 wraps
Include Image intake device 120 and processing unit 110.
Wherein, for Image intake device 120, it is used to shoot or receive image and/or video containing object.Its
In, Image intake device 120 can be the existing various electronic equipments for capableing of image recording and/or video in the prior art,
External information is obtained by electromagnetism or optics or other signal sources, structure can refer to the various prior arts of the prior art
The various electronic equipments with image capturing function, including but not limited to video camera, camera, camera, with shooting function
Mobile phone, tablet computer, but the electronic equipment with image capturing function of the prior art does not include to image data progress nerve
The functional unit or module of network operations.
In embodiments of the present invention, object can be people, animal, plant, natural object or manually manufacture object, and individual
Between can have differentiation, can be identified by mark data.Such as object is people, mark data can be people's wearing
Clothes pattern, text and/or number, mark data are also possible to the body part or global feature of people, including but not limited to send out
Type, shape of face or height；It, can be by wear or animal partly or wholly such as object is animal, with the mankind
Feature differentiation comes；Such as object is artificial manufacture object, is further illustrated here by enumerating automobile, for example pass through vapour
Vehicle license plate is distinguished to distinguish automobile, or by vehicle color, model etc..
Fig. 2 is the functional-block diagram of processing unit in Fig. 1.It in some embodiments, can be with for processing unit 200
Including data preprocessing module 201, object extraction module 202 and identification module 204.
Wherein, data preprocessing module 201 from image captured by Image intake device 120 and/or video for selecting
It takes and meets the image to impose a condition or video frame.Here setting condition may include from the image and/or video of shooting
Extract image or video frame with the above difference of given threshold.Further, it can choose as follows, Ke Yitong
Artificial neural network to be crossed, video is processed, is specifically included: taking current video frame T, video frame carries out feature extraction through convolution,
Most the output layer through neural network obtains a scoring f afterwardsT, as the score of the video frame, represent the feature of the frame, fTWith f0
Compare, f0It is initialized as 0, if difference is greater than given threshold, data of the frame as subsequent module, by fTIt is assigned to f0, remove
One video frame T+1, repeats the above, until completing all video frames.
In some embodiments, artificial neural network used in data preprocessing module 201 can be depth nerve net
Network, deep neural network algorithm are divided into training process and use process two parts.In the training process, the figure that usage history is collected
Picture or video content.Optionally, different data sets is used to different target object, for example, in the task of parking lot, with mark
The video in the different parking lots of license plate number utilizes the data as training set training book deep neural network.Here depth mind
It may include convolutional layer, full articulamentum, pond layer and batch normalization (batch norm) layer through network, to video or image
Do bulk processing (selection operation in data preprocessing module 201 namely described above).
Wherein, Objective extraction (attach) module 202 is used for target detection, obtains object or object in image
At least part of target image.Here for the selection of at least part of target image, it should be that this part chosen can
Make the parts of images that differentiation is formed between object, such as the image of the part containing automotive license plate.
In some embodiments, at least part of target image for obtaining object or object may include: with choosing
It selects searching algorithm and generates candidate region, picture is divided into many zonules, is merged by level group technology according to similarity,
Obtain at least part of boundary candidate frame of object or object；Sliding window method is used to boundary candidate frame, according to object
Perhaps at least part of scale of object is slided on bounding box as window size obtains object or object
At least part of object region.Using object as automobile, at least part of target image of object includes automobile for citing
License plate for carrying out the network of target detection for example, can be trained, with selecting searching algorithm to generate candidate region, first
Picture is divided into many zonules by simple region division, is then closed by level group technology according to certain similarity
And the boundary candidate frame of many vehicles is obtained, then, sliding window method is used to the boundary candidate frame of each vehicle, according to license plate
Scale is slided on bounding box as window size, obtains license plate area.
In some embodiments, the artificial neural network that object extraction module 202 uses can be deep neural network, deep
Degree neural network algorithm is divided into training process and use process two parts.In the training process, the satisfaction that usage history is collected is set
The image or video frame of fixed condition.Here deep neural network may include convolutional layer, full articulamentum, pond layer and batch return
One changes (batch norm) layer, obtains object or object extremely to do to the image or video frame that meet setting condition
Least a portion of processing (the acquisition operation in object extraction module 202 namely described above).
Wherein, identification module 204 identifies the mark data for distinguishing object for identifying to target image.This
In mark data, can be the various data that can distinguish object extracted from target image, including but not limited to scheme
Case, Chinese character, letter, numbers and symbols.For people, height, the pattern of wear, text, letter and/or number can be
According to for artificiality, such as automobile, the letter that can be in license plate number is combined with data, it is of course also possible to be part
License plate number, such as latter three of license plate number.
In some embodiments, target image is identified, identifies that the mark data for distinguishing object may include:
By artificial neural network, picture identification data is positioned, is identified respectively: all candidate frames of image is extracted, to each
A candidate frame, is sized and adapts to artificial neural network input size, and the characteristic pattern obtained by convolutional neural networks is inputting
Into sorter network, the sorter network energy identification feature figure is final to obtain the marking data information to be obtained in original graph.Citing
Using object as automobile, at least part of target image of object includes automotive license plate for example, extracting all times of image
Frame is selected, to each candidate frame, is sized the input size for adapting to convolutional neural networks, the feature that convolutional neural networks obtain
Figure, is input in sorter network, the sorter network energy identification feature figure, final to obtain the information to be obtained in original graph, is stopping
In the task of parking lot, that is, obtain the license plate number (combination that identification data here include letter and number) of many vehicles.
In some embodiments, the artificial neural network that identification module 204 uses can be deep neural network, depth mind
It is divided into training process and use process two parts through network algorithm.In the training process, the target image that usage history is collected.This
In deep neural network may include convolutional layer, full articulamentum, pond layer and batch normalization (batch norm) layer, to mesh
Logo image identify the processing (the identification operation in identification module 204 namely described above) of object.
In some embodiments, processing unit 200 can also include enhancing module 203, which passes through to full
The resolution ratio of the image or video frame that impose a condition enough is that the image of first resolution or video frame are handled to obtain second
The image or video frame of resolution ratio, the second resolution are higher than the first resolution.Such as obtained from object extraction module 202
The resolution ratio of the image obtained is lower, and the accuracy rate directly identified by identification module 204 is not high, it is possible to pass through enhancing
Module 203 handles the image of low resolution to obtain high-resolution image, so that clarity improves.This module can be by
Convolution sum deconvolution composition extracts feature by convolution, the characteristic pattern of a higher-dimension is obtained by deconvolution, then through a convolution
The convolutional layer that core size is 1*1 carries out Nonlinear Mapping, the characteristic pattern of higher-dimension is mapped to the characteristic pattern of another higher-dimension, most
Afterwards, then with a convolutional layer it is rebuild, obtains high-resolution image.
In some embodiments, the processing unit includes neural network processor, integrate the data preprocessing module,
Object extraction module 202, data processing module and at least one for enhancing module.That is four moulds in the processing unit
Block can be respectively with four processors, and each neural network processor has identical structure as shown in figure 3, processor part can be with
It is trained and reasoning.
As shown in figure 3, in some embodiments, neural network processor includes storage unit 310,320 and of control unit
Arithmetic element 330, wherein storage unit 310 is for storing input data (can be used as input neuron), neural network parameter
And instruction；Control unit 320 is decoded into arithmetic element 330 for reading special instruction from the storage unit 310
It instructs and is input to arithmetic element 330；Arithmetic element 330 is used to execute the data according to the instruction of arithmetic element 330 corresponding
Neural network computing, obtain output neuron.Wherein, storage unit 310 can also be stored obtains after 330 operation of arithmetic element
The output neuron obtained.Here neural network parameter includes but is not limited to weight, biasing and activation primitive.Preferably,
Initialization weight in parameter is trained weight, can directly carry out artificial neural network operation, save to nerve net
The process that network is trained.
In some embodiments, executed in arithmetic element 330 corresponding neural network computing include: will input neuron and
Weight data is multiplied, and obtains multiplied result；Add tree operation is executed, for the multiplied result to be passed through add tree phase step by step
Add, obtain weighted sum, weighted sum biasing is set or is not processed；
The weighted sum set or be not processed to biasing executes activation primitive operation, obtains output neuron.Preferably,
Activation primitive can be sigmoid function, tanh function, ReLU function or softmax function.
In some embodiments, as shown in figure 3, neural network processor can also include DMA340 (Direct Memory
Access, direct memory access), the input data for being stored in storage unit 310, neural network parameter and instruction, for
Control unit 320 and arithmetic element 330 are called；Further it is also used to after arithmetic element 330 calculates output neuron, to
The output neuron is written in storage unit 310.
In some embodiments, it as shown in figure 3, neural network processor further includes instruction buffer 350, is used for from described straight
Memory access DMA340 cache instruction is connect, is called for control unit 320.The instruction buffer 350 can cache on piece, pass through
Preparation process is integrated on neural network processor, processing speed can be improved when instruction is transferred, when saving integral operation
Between.
In some embodiments, neural network processor further include: input neuron caching 370 is used for from described straight
Memory access DMA340 caching input neuron is connect, is called for arithmetic element 330；Weight caching 360 is used for from described direct
Memory access DMA340 caches weight, calls for arithmetic element 330；Output neuron caching 380 is used to store from the fortune
It calculates unit 330 and obtains the output neuron after operation, with output to direct memory access DMA340.Above-mentioned input neuron caching
370, weight caching 360 and output neuron caching 380 or on piece caching, are integrated in nerve by semiconductor technology
On network processing unit, processing speed can be improved when reading and writing for arithmetic element 330, saves the integral operation time.
Based on the same inventive concept, the embodiment of the present disclosure also provides a kind of control system, comprising: at least one above-mentioned implementation
Intelligent video camera head and control terminal described in example, the quantity configuration of the intelligent video camera head are one setting place of camera shooting covering；Control
End processed is used to receive the mark data of each intelligent video camera head processing, determines that mark data corresponds to object in setting place
Position.
Referring to fig. 4 and shown in Fig. 5, in a setting place 430 (such as various internal or external spaces, including but it is unlimited
In parking lot, square, classroom or office), image capturing and processing are carried out by intelligent video camera head 410, for example, being absorbed
Image and/or video frame in, at least partly include the image of object 420 (such as automobile), be only schematic table in figure
Show can actually there be multiple automobiles (and stand or irregular), it, can be with after shooting and identify by intelligent video camera head
Determine the unlabeled data (such as license plate number or part license plate number) of object.
Due to being deployed to ensure effective monitoring and control of illegal activities by least one camera to setting place 430, it under normal circumstances, can be right
Entire setting place is imaged and is analyzed, and analyzes which intelligent video camera head 410 takes in object 420；Furthermore, it is possible to logical
The image for crossing intelligent video camera head intake determines that object (may be otherwise and install according to intelligent video camera head in the orientation in setting place
Location information carry out referring to acquisition).
Referring to Figure 5, control terminal 440 receives the mark data of object determined by each intelligent video camera head 410, may be used also
It with the azimuth information of combining target object (intelligent video camera head 410), counted, analyzed and is shown, for example analyze setting field
430 where have expired or have also had vacant position, for example analyze the object number that setting place has same or similar mark data
It measures, or determines the azimuth information of object by Search Flags data；Or object is determined by searching for azimuth information
Unlabeled data etc..
In some embodiments, control terminal 440 may include display device and/or instantaneous speech power, for exporting such as
Shown on the object that determines of the control terminal 440 introduced setting place location information.
In embodiment provided by the disclosure, it should be noted that, disclosed relevant apparatus and method can pass through others
Mode is realized.For example, the apparatus embodiments described above are merely exemplary, such as the division of the part or module,
Only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple portions or module can be with
In conjunction with being perhaps desirably integrated into a system or some features can be ignored or does not execute.
In the disclosure, term "and/or" may be had been used.As used herein, term "and/or" means one
Or other or both (for example, A and/or B mean A or B or both A and B).
In the above description, for purpose of explanation, elaborate numerous details in order to provide each reality to the disclosure
Apply the comprehensive understanding of example.However, the skilled person will be apparent that, without certain in these details
Implementable one or more other embodiments.Described specific embodiment be not limited to the disclosure but in order to illustrate.
The scope of the present disclosure is not determined by specific example provide above, is only determined by following claim.At other
In the case of, in form of a block diagram, rather than it is illustrated in detail known circuit, structure, equipment, and operation is so as not to as making to retouching
The understanding stated thickens.In place of thinking to be suitable for, the ending of appended drawing reference or appended drawing reference is weighed in all attached drawings
It is multiple to indicate optionally correspondence or similar element with similar characteristics or same characteristic features, unless otherwise specifying or
Obviously.
Each functional unit/subelement/module/submodule can be hardware in the disclosure, for example the hardware can be electricity
Road, including digital circuit, analog circuit etc..The physics realization of hardware configuration includes but is not limited to physical device, physics device
Part includes but is not limited to transistor, memristor etc..Computing module in the computing device can be any appropriate hard
Part processor, such as CPU, GPU, FPGA, DSP and ASIC etc..The storage unit can be any magnetic storage appropriate and be situated between
Matter or magnetic-optical storage medium, such as RRAM, DRAM, SRAM, EDRAM, HBM, HMC etc..
It is apparent to those skilled in the art that for convenience and simplicity of description, only with above-mentioned each function
The division progress of module can according to need and for example, in practical application by above-mentioned function distribution by different function moulds
Block is completed, i.e., the internal structure of device is divided into different functional modules, to complete all or part of function described above
Energy.
Particular embodiments described above has carried out further in detail the purpose of the present invention, technical scheme and beneficial effects
Describe in detail bright, it should be understood that the above is only a specific embodiment of the present invention, is not intended to restrict the invention, it is all
Within the spirit and principles in the present invention, any modification, equivalent substitution, improvement and etc. done should be included in protection of the invention
Within the scope of.
Claims (18)
1. a kind of intelligent video camera head, characterized by comprising:
Image intake device, for shooting or receiving image and/or video containing object；And processing unit, comprising:
Data preprocessing module is chosen from the image and/or video of shooting and meets the image to impose a condition or video frame；
Object extraction module is used for target detection, obtains at least part of target figure of the object or object in image
Picture；
Identification module identifies target image, identifies the mark data for distinguishing object.
2. intelligent video camera head according to claim 1, which is characterized in that the processing unit further include:
Enhance module, by the image or view that the resolution ratio for meeting the image or video frame that impose a condition are first resolution
Frequency frame is handled to obtain the image of second resolution or video frame, and the second resolution is higher than the first resolution.
3. intelligent video camera head according to claim 1, which is characterized in that the setting condition includes:
From the image and/or video of shooting, image or video frame with the above difference of given threshold are extracted.
4. intelligent video camera head according to claim 3, which is characterized in that extract the view with the above difference of given threshold
Frequency frame includes:
By artificial neural network, video is processed, is specifically included: taking current video frame T, video frame carries out feature through convolution
It extracts, most the output layer through neural network obtains a scoring f afterwardsT, as the score of the video frame, represent the feature of the frame, fT
With f0Compare, f0It is initialized as 0, if difference is greater than given threshold, data of the frame as subsequent module, by fTIt is assigned to f0,
A video frame T+1 is removed, is repeated the above, until completing all video frames.
5. intelligent video camera head according to claim 1, which is characterized in that the object behaviour, animal, plant, nature
Object manually manufactures object.
6. intelligent video camera head according to claim 5, which is characterized in that the artificial manufacture object is automobile, the target
At least partly image of object includes automotive license plate.
7. intelligent video camera head according to claim 1, which is characterized in that in the object extraction module, examined for target
It surveys, obtains at least part of target image of the object or object in image, comprising:
By artificial neural network, wherein piece image or video frame are read, obtains object or object at least partly
Target image.
8. intelligent video camera head according to claim 7, which is characterized in that obtain object or object at least partly
Target image include:
By artificial neural network, candidate region is generated using selection searching algorithm, picture is divided into many zonules, is passed through
Level group technology merges according to similarity, obtains at least part of boundary candidate frame of object or object；To candidate
Bounding box uses sliding window method, according at least part of scale of object or object as window size in bounding box
Upper sliding obtains at least part of object region of object or object.
9. intelligent video camera head according to claim 1, which is characterized in that the mark data includes following at least one:
Pattern, Chinese character, letter, numbers and symbols.
10. intelligent video camera head according to claim 1, which is characterized in that in the identification module, carried out to target image
Identification identifies the mark data for distinguishing object, comprising:
By artificial neural network, picture identification data is positioned, is identified respectively: extracting all candidate frames of image, it is right
Each candidate frame is sized and adapts to artificial neural network input size, the characteristic pattern obtained by convolutional neural networks,
It is input in sorter network, the sorter network energy identification feature figure, it is final to obtain the marking data information to be obtained in original graph.
11. intelligent video camera head according to claim 2, which is characterized in that the processing unit includes Processing with Neural Network
Device integrates the data preprocessing module, object extraction module, data processing module and at least one for enhancing module.
12. intelligent video camera head according to claim 11, which is characterized in that the neural network processor includes:
Storage unit, for storing the input data, neural network parameter and instruction；
Control unit for reading special instruction from the storage unit, and is decoded into arithmetic element and instructs and input
To arithmetic element；
Arithmetic element obtains output mind for executing corresponding neural network computing to the data according to arithmetic element instruction
Through member.
13. intelligent video camera head according to claim 12, which is characterized in that in the arithmetic element, execute corresponding mind
Include: through network operations
Input neuron is multiplied with weight data, obtains multiplied result；
Execute add tree operation, for the multiplied result to be added step by step by add tree, obtain weighted sum, to weighted sum plus
It biases or is not processed；
The weighted sum set or be not processed to biasing executes activation primitive operation, obtains output neuron.
14. intelligent video camera head according to claim 12, which is characterized in that the processor further include:
Pretreatment unit, image and/or video data for absorbing to video camera pre-process, and are converted into recognition of face knot
Fruit, the face recognition result are to meet the data of neural network input format；
And/or direct memory access DMA, the input data for being stored in storage unit, neural network parameter and instruction, for
Control unit and arithmetic element are called.
15. intelligent video camera head according to claim 12, which is characterized in that the processor further includes following at least one
Kind:
Instruction buffer, for being called for control unit from the direct memory access DMA cache instruction；
Neuron caching is inputted, for caching input neuron from the direct memory access DMA, is called for arithmetic element；
Weight caching is called for caching weight from the direct memory access DMA for arithmetic element；And
Output neuron caching, obtains the output neuron after operation for storing from the arithmetic element, to export to direct
Memory access DMA.
16. intelligent video camera head according to claim 15, which is characterized in that described instruction caches, input neuron caches,
Weight caching and output neuron caching are that on piece caches.
17. a kind of control system characterized by comprising
Any intelligent video camera head of an at least claim 1-16, the quantity configuration of the intelligent video camera head are camera shooting covering
One setting place；
Control terminal receives the mark data of each intelligent video camera head processing, determines that mark data corresponds to object in setting field
Position.
18. control system according to claim 17, which is characterized in that the control terminal includes:
Display device and/or instantaneous speech power, location information of the object determining for output control terminal in setting place.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811402218.4A CN109714526B (en) | 2018-11-22 | 2018-11-22 | Intelligent camera and control system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811402218.4A CN109714526B (en) | 2018-11-22 | 2018-11-22 | Intelligent camera and control system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109714526A true CN109714526A (en) | 2019-05-03 |
| CN109714526B CN109714526B (en) | 2021-02-09 |
Family
ID=66254976
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811402218.4A Active CN109714526B (en) | 2018-11-22 | 2018-11-22 | Intelligent camera and control system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109714526B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111739065A (en) * | 2020-06-29 | 2020-10-02 | 上海出版印刷高等专科学校 | Target identification method, system, electronic equipment and medium based on digital printing |
| CN111931692A (en) * | 2020-08-31 | 2020-11-13 | 青岛聚看云科技有限公司 | Display device and image recognition method |
| CN112104840A (en) * | 2020-09-09 | 2020-12-18 | 深圳市有方科技股份有限公司 | Video acquisition method and electronic equipment |
| CN113343891A (en) * | 2021-06-24 | 2021-09-03 | 深圳市起点人工智能科技有限公司 | Detection device and detection method for child kicking quilt |
| WO2021179186A1 (en) * | 2020-03-10 | 2021-09-16 | 华为技术有限公司 | Focusing method and apparatus, and electronic device |
| CN113918679A (en) * | 2021-09-22 | 2022-01-11 | 三一汽车制造有限公司 | Knowledge question and answer method and device and engineering machinery |
| CN114003160A (en) * | 2021-10-29 | 2022-02-01 | 影石创新科技股份有限公司 | Data visualization display method and device, computer equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140300758A1 (en) * | 2013-04-04 | 2014-10-09 | Bao Tran | Video processing systems and methods |
| CN104284150A (en) * | 2014-09-23 | 2015-01-14 | 同济大学 | Smart camera autonomous coordinative tracking method and monitoring system based on road traffic monitoring |
| CN108764468A (en) * | 2018-05-03 | 2018-11-06 | 中国科学院计算技术研究所 | Artificial neural network processor for intelligent recognition |
-
2018
- 2018-11-22 CN CN201811402218.4A patent/CN109714526B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140300758A1 (en) * | 2013-04-04 | 2014-10-09 | Bao Tran | Video processing systems and methods |
| CN104284150A (en) * | 2014-09-23 | 2015-01-14 | 同济大学 | Smart camera autonomous coordinative tracking method and monitoring system based on road traffic monitoring |
| CN108764468A (en) * | 2018-05-03 | 2018-11-06 | 中国科学院计算技术研究所 | Artificial neural network processor for intelligent recognition |
Non-Patent Citations (1)
| Title |
|---|
| CSDN: "《R-CNN论文详解（论文翻译）》", 《HTTPS://BLOG.CSDN.NET/V1_VIVIAN/ARTICLE/DETAILS/78599229》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021179186A1 (en) * | 2020-03-10 | 2021-09-16 | 华为技术有限公司 | Focusing method and apparatus, and electronic device |
| CN113711123A (en) * | 2020-03-10 | 2021-11-26 | 华为技术有限公司 | Focusing method and device and electronic equipment |
| CN111739065A (en) * | 2020-06-29 | 2020-10-02 | 上海出版印刷高等专科学校 | Target identification method, system, electronic equipment and medium based on digital printing |
| CN111931692A (en) * | 2020-08-31 | 2020-11-13 | 青岛聚看云科技有限公司 | Display device and image recognition method |
| CN112104840A (en) * | 2020-09-09 | 2020-12-18 | 深圳市有方科技股份有限公司 | Video acquisition method and electronic equipment |
| CN112104840B (en) * | 2020-09-09 | 2022-10-04 | 深圳市有方科技股份有限公司 | Video acquisition method and mobile baseband workstation |
| CN113343891A (en) * | 2021-06-24 | 2021-09-03 | 深圳市起点人工智能科技有限公司 | Detection device and detection method for child kicking quilt |
| CN113918679A (en) * | 2021-09-22 | 2022-01-11 | 三一汽车制造有限公司 | Knowledge question and answer method and device and engineering machinery |
| CN114003160A (en) * | 2021-10-29 | 2022-02-01 | 影石创新科技股份有限公司 | Data visualization display method and device, computer equipment and storage medium |
| CN114003160B (en) * | 2021-10-29 | 2024-03-29 | 影石创新科技股份有限公司 | Data visual display method, device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109714526B (en) | 2021-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109714526A (en) | Intelligent video camera head and control system | |
| Chang et al. | Batch-normalized maxout network in network | |
| CN104392463B (en) | Image salient region detection method based on joint sparse multi-scale fusion | |
| CN104063719B (en) | Pedestrian detection method and device based on depth convolutional network | |
| CN109523520A (en) | A kind of chromosome automatic counting method based on deep learning | |
| CN103870816B (en) | The method of the plants identification that a kind of discrimination is high | |
| CN107016357A (en) | A kind of video pedestrian detection method based on time-domain convolutional neural networks | |
| CN105354273A (en) | Method for fast retrieving high-similarity image of highway fee evasion vehicle | |
| CN105095866A (en) | Rapid behavior identification method and system | |
| He et al. | A robust method for wheatear detection using UAV in natural scenes | |
| Weideman et al. | Extracting identifying contours for African elephants and humpback whales using a learned appearance model | |
| CN112464730B (en) | Pedestrian re-identification method based on domain-independent foreground feature learning | |
| CN114387499A (en) | Island coastal wetland waterfowl identification method, distribution query system and medium | |
| CN107944403A (en) | Pedestrian's attribute detection method and device in a kind of image | |
| CN105243154A (en) | Remote sensing image retrieval method and system based on significant point characteristics and spare self-encodings | |
| CN109871789A (en) | Vehicle checking method under a kind of complex environment based on lightweight neural network | |
| CN110580446A (en) | Behavior semantic subdivision understanding method, system, computer device and medium | |
| CN105405138A (en) | Water surface target tracking method based on saliency detection | |
| Li et al. | Codemaps-segment, classify and search objects locally | |
| Zhang et al. | A precise apple leaf diseases detection using BCTNet under unconstrained environments | |
| Shen et al. | Yolov5-based model integrating separable convolutions for detection of wheat head images | |
| Qiu et al. | Computer vision technology based on deep learning | |
| Wang et al. | Pig face recognition model based on a cascaded network | |
| Zheng et al. | A large-scale hyperspectral dataset for flower classification | |
| Yang et al. | Urban street tree dataset for image classification and instance segmentation |
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 |