CN109583396A - A kind of region prevention method, system and terminal based on CNN two stages human testing - Google Patents
A kind of region prevention method, system and terminal based on CNN two stages human testing Download PDFInfo
- Publication number
- CN109583396A CN109583396A CN201811480634.6A CN201811480634A CN109583396A CN 109583396 A CN109583396 A CN 109583396A CN 201811480634 A CN201811480634 A CN 201811480634A CN 109583396 A CN109583396 A CN 109583396A
- Authority
- CN
- China
- Prior art keywords
- human body
- region
- picture
- detection zone
- human
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000002265 prevention Effects 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012360 testing method Methods 0.000 title claims abstract description 23
- 238000001514 detection method Methods 0.000 claims abstract description 80
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 10
- 230000033001 locomotion Effects 0.000 claims description 9
- 239000006185 dispersion Substances 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 abstract description 5
- 238000012544 monitoring process Methods 0.000 abstract description 5
- 238000011897 real-time detection Methods 0.000 abstract 1
- 238000003860 storage Methods 0.000 description 17
- 238000004422 calculation algorithm Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 6
- 238000004590 computer program Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 235000013399 edible fruits Nutrition 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000002547 anomalous effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- GOLXNESZZPUPJE-UHFFFAOYSA-N spiromesifen Chemical compound CC1=CC(C)=CC(C)=C1C(C(O1)=O)=C(OC(=O)CC(C)(C)C)C11CCCC1 GOLXNESZZPUPJE-UHFFFAOYSA-N 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000032258 transport Effects 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of region prevention method and device based on CNN two stages human testing, is related to computer vision applied technical field.The present invention is based on convolutional neural networks to carry out region prevention to the picture in monitoring video flow, is included in picture and constructs detection zone and identification human body, and judges human body whether in the detection zone of building, if human body in detection zone, executes alarm operation.The prevention region that technical solution provided by the invention may be implemented in setting carries out the real-time detection and alarm of human body entrance, and Detection accuracy is higher than 85% in actual production environment, significantly reduces the working strength of supervisor, improves automatic management level.
Description
Technical field
The present invention relates to computer vision applied technical fields more particularly to a kind of based on CNN two stages human testing
Region prevention method, system and terminal.
Background technique
With artificial intelligence and multimedia development, video monitoring especially keep the safety in production field application in each field of society
It is increasingly wider.
Region prevention is system the most basic in public safety prevention, is the first of anti-illegal-inbreak and anomalous event
Road defence line is also very important one of defence line.Video area prevention is built upon on traditional boundary defence conceptual foundation, is passed through
Using intelligent video analysis technology, not only has intrusion alarm effect, and can also be real-time by the video monitoring equipment of front end
The case where understanding monitoring area issues warning once intrusion behavior occurs at the first time, and informs at Security Personnel in time
Reason.However, since a large amount of report by mistake that the factors such as leaf rocks, light-illuminating, animal travels generate leverages making for user
With experience.
Target detection is the most common problem in machine vision, is a kind of image based on target geometry and statistical nature point
It cuts, the segmentation of target and identification are combined into one by it, and accuracy and real-time are a significant capabilities of whole system, in recent years
Come, for target detection in artificial intelligence, recognition of face is unmanned that fields is waited all to be widely used.Due to illumination variation,
The reasons such as partial occlusion, target scale variation, cause detection difficulty to increase, and the target detection under complex background is theoretical in recent years
With the research hotspot of application.
However, will receive various interference during target detection, for example, angle, block, light intensity etc. because
Element, these factors will lead to target and are distorted, and increase new challenge for target detection.Conventional target detection algorithm has two
A main defect: (1) it is using when sliding window strategy progress regional choice specific aim not strong, improve time complexity and window
Mouth redundancy；(2) there is no good robustness for the diversity of target for the feature of manual designs.
Summary of the invention
The technical problem to be solved by the present invention is to be directed to two defects of conventional target detection algorithm, a kind of letter is provided
Whether single, identification human body that is accurately taking into account detection speed enters detection method, system and the terminal in prevention region.
To solve the above-mentioned problems, the present invention proposes following technical scheme:
In a first aspect, the present invention proposes a kind of region prevention method based on human testing, comprising the following steps:
S1 chooses picture, and detection zone is constructed in the picture, and the detection zone includes prevention region；
S2 carries out region nomination using k different rectangle frames on the convolution characteristic layer of the picture, filters out human body
Candidate region, wherein k is positive integer；
S3 carries out feature extraction to the candidate region automatically using convolutional neural networks；
S4 is classified using feature of the human body sorter model to the extraction, is returned, identifies whether the picture is deposited
In human body.
S5, and if it exists, then judge human body whether in the detection zone of building；
S6, if so, executing alarm operation.
Further technical solution is for it, described to construct detection zone in the picture of selection, comprising:
At least three angle point is marked on the picture；
It is sequentially connected each angle point, forms enclosed region, the enclosed region is detection zone.
Further technical solution is for it, the k value 9.
Further technical solution is for it, before the step S2 further include: the samples pictures comprising human body is collected, to institute
It states samples pictures to be labeled, training obtains the parameter of human body sorter model.
Further technical solution is for it, and the feature includes the textural characteristics, edge feature and motion feature of image；
The textural characteristics include the grey level histogram, edge orientation histogram, gray level co-occurrence matrixes of image；
The edge feature includes perimeter, area, the ratio of width to height, the dispersion degree, tightness of image；
The motion feature includes movement mass center, speed, displacement and gradient.
Whether further technical solution is for it, described to judge human body in the detection zone of building, comprising:
The position of human body is labeled as position of human body point；
A ray is done with position of human body point, judges whether the intersection point sum of ray and enclosed region is odd number；
If the intersection point sum is odd number, determine human body in the detection zone of building.
Further technical solution is for it, it is described a ray is done with position of human body point before, further includes:
Judge the position of human body point whether on the boundary of enclosed region；
If so, determining human body in the detection zone of building；
A ray done with position of human body point if it is not, then executing, judge ray and enclosed region intersection point sum whether be
The step of odd number.
Further technical solution is for it, and the picture is the I frame picture in live video stream.
Second aspect, the present invention propose a kind of region crime prevention system based on human testing, comprising: for executing such as first
The unit of method described in aspect.
The third aspect, the embodiment of the invention provides a kind of terminal, which includes that processor, input equipment, output are set
Standby and memory, the processor, input equipment, output equipment and memory are connected with each other, wherein the memory is for depositing
Storage supports terminal to execute the application code of the above method, and the processor is configured for executing the side of above-mentioned first aspect
Method.
Fourth aspect, the embodiment of the invention provides a kind of computer readable storage medium, the computer storage medium
It is stored with computer program, the computer program includes program instruction, and described program instruction makes institute when being executed by a processor
State the method that processor executes above-mentioned first aspect.
Compared with prior art, the attainable technical effect of present invention institute includes:
It is anti-to carry out region to the picture in monitoring video flow based on convolutional neural networks for present invention application depth learning technology
Model, including constructing detection zone in picture, by carrying out area using k different rectangle frames on the convolution characteristic layer of picture
Domain nomination with accelerate identify human body, and judge human body whether in the detection zone of building, and then execute alarm operation.The present invention
The technical solution of offer has both detection speed and accuracy, it can be achieved that carrying out the real-time inspection of human body entrance in the prevention region of setting
It surveys and alarms, Detection accuracy is higher than 85% in actual production environment, significantly reduces the working strength of supervisor, improves
Automatic management level.
Detailed description of the invention
Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to needed in embodiment description
Attached drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the invention, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is the region prevention method flow chart provided in an embodiment of the present invention based on human testing；
Fig. 2 is the region crime prevention system schematic diagram provided in an embodiment of the present invention based on human testing；
Fig. 3 is a kind of terminal schematic diagram provided in an embodiment of the present invention；
Fig. 4 is the specific flow chart of S101 of embodiment of the present invention step；
Fig. 5 is the detection zone schematic diagram of S101 of embodiment of the present invention step building；
Fig. 6 is the specific flow chart that the embodiment of the present invention identifies human body；
Fig. 7 is the schematic diagram of S105 of embodiment of the present invention step；
Fig. 8 is the specific flow chart of S105 of embodiment of the present invention step.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, is clearly and completely retouched to the technical solution in embodiment
It states, similar reference numerals represent similar component in attached drawing.Obviously, will be described below embodiment is only the present invention one
Divide embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making
Every other embodiment obtained, shall fall within the protection scope of the present invention under the premise of creative work.
It should be appreciated that ought use in this specification and in the appended claims, term " includes " and "comprising" instruction
Described feature, entirety, step, operation, the presence of element and/or component, but one or more of the other feature, whole is not precluded
Body, step, operation, the presence or addition of element, component and/or its set.
It is also understood that in this embodiment of the present invention term used in the description merely for the sake of description particular implementation
Example purpose and be not intended to limit the embodiment of the present invention.Such as the institute in specification and appended book of the embodiment of the present invention
As use, other situations unless the context is clearly specified, otherwise " one " of singular, "one" and "the" are intended to wrap
Include plural form.
In many industries of production and living, all there are some regions, according to safety in production operating provision, needs to forbid personnel
Into, such as the lower zone of construction site derrick car arm, prison enclosure wall neighboring area etc. belong to forbidden zone, need to supervise.Base
In the demand, the embodiment of the present invention proposes a kind of region prevention method and system based on human testing, can automate knowledge
It is horizontal to improve supervision by the personnel of other specific region.See following embodiment:
Embodiment
Referring to Fig. 1, the embodiment of the present invention proposes a kind of region prevention method based on human testing, as seen from the figure, packet
Include following steps:
S101 chooses picture, and detection zone is constructed in the picture, and the detection zone includes prevention region.
It referring to fig. 4, is the specific flow chart of S101 step of the embodiment of the present invention.In specific implementation, containing prevention
I frame picture is chosen in the live video stream in region, constructs detection zone.
In general, there is three kinds of coded frames: intracoded frame (I frame), encoded predicted frame (P frame) and two-way in video streaming
Coded frame (B frame).I frame only carries out video compress using this frame spatial coherence, and P frame is pre- using forward reference frame progress time-domain
Coding is surveyed, B frame carries out time-domain predictive coding using the two-way reference frame of forward and backward.Under normal conditions, I frame compression ratio
Low, picture quality is slightly good, and the embodiment of the present invention chooses the I frame image in live video stream to construct detection zone.
In some embodiments, such as the present embodiment, at least three angle point is marked on the picture chosen；It is sequentially connected each
Angle point forms enclosed region, and the enclosed region is detection zone.
It referring to fig. 4, is the specific flow chart of S101 of embodiment of the present invention step.In some embodiments, such as this implementation
In example, opens and show the picture chosen, angle point is marked on picture.N (N >=3) a angle point is marked, with solid black on picture
Point display.According to clockwise, the abscissa x of all angle points and ordinate y are stored in two-dimensional array.Referring to Fig. 5,
For the detection zone schematic diagram of building.Actual prevention region is elliptic region, takes A, B, C, D and E totally 5 on elliptic region periphery
A angle point is shown on background picture with solid stain, by the transverse and longitudinal coordinate value of 5 angle points according to clockwise direction, that is,
A, the sequence of B, C, D and E are stored in two-dimensional array, judge whether human body uses in the detection zone of building to be subsequent.?
In implementation, after each angle point can be chosen by mouse click, it is moved to new position and records coordinate.
It should be noted that simple and fast for subsequent judgement, this method line segment according to successively connecting clockwise
Each angle point is connect, enclosed region is formed.In an implementation, angle point A connection angle point B forms line segment AB, and so on, successively formed
Line segment BC, CD, DE and EA ultimately form enclosed region ABCDE, that is, the detection zone constructed.It in other implementations, can be with
Related angle point is translated by other methods, and then adjusts the size in building prevention region, it includes practical for making the prevention region of building
Prevention region.Corner location changes the two-dimensional array that update storage transverse and longitudinal coordinate.
In an implementation, technical staff can adjust the size of detection zone by adjusting the quantity of angle point and position, thus
So that detection zone is more bonded actual prevention region, the precision of detection is improved.
S102 carries out region using k different rectangle frames (Anchor Box) on the convolution characteristic layer of the picture
Nomination, filters out human body candidate region, wherein k is positive integer；
Referring to Fig. 6, the flow chart of human body is identified for the embodiment of the present invention.
In specific implementation, the algorithm of target detection for quoting convolutional neural networks carries out region nomination and feature to the picture
It extracts, identifies the picture with the presence or absence of human body.
In certain embodiments, since target possibly is present at any position of image, and the size of target, length-width ratio
Example is not known yet, so the strategy that sliding window can be used traverses entire image, it is different by the way that different scales is arranged
Length-width ratio carry out region nomination.Although the strategy of this exhaustion contains all positions being likely to occur of target, but when
Between complexity it is high, generate that redundancy window is too many, seriously affect subsequent characteristics and extract and the speed of classification, to the performance requirement of GPU
It is high.
In specific implementation, mentioned on the convolution characteristic layer of picture using k different rectangle frames (Anchor Box)
Name screens human body candidate region using friendship and than tool IoU.
It is tested by experiment, is preferably to compromise in terms of time efficiency and accuracy in detection when k takes 9, both guarantees reality
When detect, and can guarantee the accuracy of detection.When k is less than 9, detection speed is improved, but accuracy declines, some human testings
Less than being easy to happen omission.When k is greater than 9, accuracy is improved, but detects speed reduction.
S103 carries out feature extraction to the candidate region automatically using convolutional neural networks；
In specific implementation, the purpose that the description of movement human characteristic is feature extraction step is obtained, characteristic is extracted
Number directly affect in next step identify work speed and accuracy.
The characteristics of image usually extracted specifically includes that the textural characteristics, edge feature and motion feature of image.(1) line
Managing feature mainly includes grey level histogram, edge orientation histogram, gray level co-occurrence matrixes of image etc..(2) edge feature is direct
Show the profile of image, main includes perimeter, area, the ratio of width to height (main shaft rate), dispersion degree, the tightness etc. of image.(3) it transports
Dynamic feature is related to motor behavior, generally comprises movement mass center, speed, displacement, gradient etc..In an implementation, whole picture is inputted,
The feature for obtaining picture automatically using CNN convolutional neural networks forms unified characteristic layer.
S104 is classified using feature of the human body sorter model to the extraction, is returned, whether identifies the picture
There are human bodies.
In specific implementation, object/ is carried out to the corresponding region each Anchor Box using human body sorter model
Non-object bis- classifies, and with k regression model (respectively corresponding to different Anchor Box) fine tuning candidate frame position and greatly
It is small, target classification is finally carried out, identifies the picture with the presence or absence of human body.
It should be noted that technical staff voluntarily can train human body sorter model to carry out identification human body, it is possible to use existing
Some trained human body sorter models are using trained human body classifier mould since the background of picture is different
When type, it is also desirable to be adjusted to parameter, it is ensured that can be accurately identified for the human body of detection zone.
Region nomination, feature extraction, classification, recurrence are shared convolution feature by the present embodiment together, carry out human bioequivalence, both
The speed for improving human testing also ensures the accuracy of detection.
The algorithm of target detection that other maturations can also be used in certain embodiments, is broadly divided into two classes: (1) two-
Stage detection algorithm, the problem of will test are divided into two stages, first generation candidate region (region proposals),
Then to candidate region classification, refine, the Typical Representative of this kind of algorithm is the R-CNN system algorithm based on region proposal,
Such as R-CNN, Fast R-CNN, Faster R-CNN etc.；(2) one-stage detection algorithm does not need region
The proposal stage directly generates the class probability and position coordinate value of object, than more typical algorithm such as YOLO and SSD.It is right
The algorithm that picture carries out human bioequivalence is highly developed now, and those skilled in the art can voluntarily select other algorithms as needed
Human bioequivalence is carried out, the present invention is not specifically limited in this embodiment.
S105, and if it exists, then judge human body whether in the detection zone of building；
Referring to Fig. 7, in specific implementation, since computer does not have the consciousness and thinking of people, for ease of calculation machine machine
Device identification, needs to carry out to judge that human body whether in the detection zone of building, is specifically included using machine language:
The position of human body is labeled as position of human body point；
Judge the position of human body point whether on the boundary of enclosed region；
If so, determining human body in the detection zone of building；
If it is not, then doing a ray with position of human body point, judge whether the intersection point sum of ray and enclosed region is odd number；
If the intersection point sum is odd number, determine human body in the detection zone of building.
For example, with reference to Fig. 7, in one embodiment, to the detection zone ABCDE constructed, human body is detected, by human body
Position is labeled as position of human body point P, next, judging that the position of human body whether in detection zone ABCDE, that is, judges
Whether point P is in polygonal region ABCDE.
It should be noted that in one embodiment, for special case, if point P is on the boundary of polygonal region ABCDE
On, then decision-point P is in polygonal region ABCDE, i.e., the position of human body is in detection zone.
In one embodiment, make a ray from P point, with image right margin compared to point Q (as shown in Figure 7), so
Afterwards according to the number of hits for calculating separately ray Yu Polygonal Boundary line segment clockwise, P is finally judged according to the parity of number of hits
Whether point is in polygon.If the intersection point of ray and polygon is odd number, P point is in polygonal internal；If ray and more
The intersection point number of side shape is even number (including 0), then P point is in outside of polygon.
For some special circumstances: if (1) certain side of polygon is parallel with ray PQ, ignoring this side；(2) such as
Fruit dot P is on certain side of polygon, then point P is in polygon.
As shown in figure 8, in one embodiment, judging detailed process statement of the point P in polygon ABCDE:
Step1: horizontal rays are done to the right using P point as starting point, meet at point Q with image right margin；
Step2: a line Edge of polygon is taken；
Step3: if side Edge is parallel with X-axis, Step2 is jumped to；
Step4: if fruit dot P is on the Edge of side, then point P jumps to Step8 in polygon；
Step5: judging whether ray PQ intersects with side Edge, if intersection, number of hits Count=Count+1；
Step6: judge whether to take all sides, if do not taken, jump to Step2；
Step7: judging the parity of Count, be odd number then point P in polygon, be even number then point P outside polygon,
Jump to Step8；
Step8: terminate.
It should be noted that in the present embodiment, for ease of calculation, ray be using P as origin horizontally to the right；And press up time
The mode of needle successively sentence section polygonal region each side whether with ray intersection, it is ensured that do not omit detection, guarantee detection knot
The accuracy of fruit.
S106, if so, executing alarm operation.
In specific implementation, if it is determined that human body then generates warning message in the detection zone of building, convenient for operation system into
Row safety records and operation.
If human body is not in the polygonal region of building, system does not generate alarm.
It in one embodiment, further include collecting the samples pictures comprising human body before step S102, to the samples pictures
It is labeled, training obtains the parameter of human body sorter model.
In specific implementation, the samples pictures comprising a large amount of human bodies are collected, are labeled using specific purpose tool, training obtains
The parameter of CNN human body sorter model, as shown in fig. 6, the samples pictures collected include positive sample and negative sample.The present embodiment choosing
With 10000 human body pictures, human body is selected using ImgLabel tool box and indicates classification, wherein 8000 are used as training sample
This, remaining picture is as verifying sample.
It referring to fig. 2, is a kind of schematic block of the region crime prevention system based on human testing provided in an embodiment of the present invention
Figure.As shown, the system 200 in the present embodiment can include: selection unit 201, recognition unit 202 sentence segment unit 203, report
Alert unit 204.Wherein,
Selection unit 201 constructs detection zone, institute for choosing the I frame picture in live video stream in the picture
Stating detection zone includes prevention region；
Screening unit 202 is mentioned for carrying out region using k different rectangle frames on the convolution characteristic layer of the picture
Name, filters out human body candidate region, wherein k is positive integer；
Extraction unit 203, for carrying out feature extraction to the candidate region automatically using convolutional neural networks；
Recognition unit 204 is identified for being classified using feature of the human body sorter model to the extraction, being returned
The picture whether there is human body；
Segment unit 205 is sentenced, for if it exists, then judging human body whether in the detection zone of building；
Alarm unit 206 is used for if so, executing alarm operation.
In one embodiment, system 200 further includes training unit 207, and the training unit 207 includes human body for collecting
Samples pictures, the samples pictures are labeled, training obtain the parameter of human body sorter model.
In one embodiment, described to sentence segment unit 205 specifically for the position of human body is labeled as position of human body point；Judgement
Whether the position of human body point is on the boundary of enclosed region；If so, determining human body in the detection zone of building；If it is not,
A ray is done with position of human body point, judges whether the intersection point sum of ray and enclosed region is odd number；If the intersection point sum
For odd number, then determine human body in the detection zone of building.
Referring to Fig. 3, be another embodiment of the present invention provides a kind of 300 schematic block diagram of terminal.This implementation as shown in the figure
Terminal 300 in example may include: one or more processors 301；One or more input equipments 302, it is one or more defeated
Equipment 303 and memory 304 out.Above-mentioned processor 301, input equipment 302, output equipment 303 and memory 304 pass through bus
305 connections.For storing instruction, processor 301 is used to execute the instruction of the storage of memory 302 to memory 302.Wherein, it handles
Device 301 is for executing: choosing picture, detection zone is constructed in the picture, the detection zone includes prevention region；To institute
It states picture and carries out region nomination and feature extraction, identify the picture with the presence or absence of human body；If it exists, then judge human body whether
In the detection zone of building；If so, executing alarm operation.
In one embodiment, the processor 301 is also used to execute: k are utilized on the convolution characteristic layer of the picture
Different rectangle frames carries out region nomination, filters out human body candidate region, wherein k is positive integer；Certainly using convolutional neural networks
It is dynamic that feature extraction is carried out to the candidate region；Classified using feature of the human body sorter model to the extraction, returned,
Identify the picture with the presence or absence of human body.
In one embodiment, the processor 301 is also used to execute: the samples pictures comprising human body is collected, to the sample
This picture is labeled, and training obtains the parameter of human body sorter model.
In one embodiment, the processor 301 is also used to execute: the position of human body is labeled as position of human body point；Sentence
The position of human body point break whether on the boundary of enclosed region；If so, determining human body in the detection zone of building；If
It is no, a ray is done with position of human body point, judges whether the intersection point sum of ray and enclosed region is odd number；If the intersection point is total
Number is odd number, then determines human body in the detection zone of building.
It should be appreciated that in embodiments of the present invention, alleged processor 301 can be central processing unit (Central
Processing Unit, CPU), which can also be other general processors, digital signal processor (Digital
Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit,
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic
Device, discrete gate or transistor logic, discrete hardware components etc..General processor can be microprocessor or this at
Reason device is also possible to any conventional processor etc..
Input equipment 302 may include that Trackpad, fingerprint adopt sensor (for acquiring the finger print information and fingerprint of user
Directional information), microphone etc., output equipment 303 may include display (LCD etc.), loudspeaker etc..
The memory 304 may include read-only memory and random access memory, and to processor 301 provide instruction and
Data.The a part of of memory 304 can also include nonvolatile RAM.For example, memory 304 can also be deposited
Store up the information of device type.
In the specific implementation, processor 301 described in the embodiment of the present invention, input equipment 302, output equipment 303 can
Implementation described in a kind of a embodiment of parameter regulation means provided in an embodiment of the present invention is executed, this also can be performed
The implementation of terminal 300 described in inventive embodiments, details are not described herein.
A kind of computer readable storage medium, the computer-readable storage medium are provided in another embodiment of the invention
Matter is stored with computer program, the realization when computer program is executed by processor:
Picture is chosen, detection zone is constructed in the picture, the detection zone includes prevention region；In the picture
Convolution characteristic layer on using k different rectangle frames carry out region nominations, filter out human body candidate region, wherein k is positive whole
Number；Feature extraction is carried out to the candidate region automatically using convolutional neural networks；It is mentioned using human body sorter model to described
The feature taken is classified, is returned, and identifies the picture with the presence or absence of human body；If it exists, then judge human body whether in building
In detection zone；If so, executing alarm operation.
The samples pictures comprising human body are collected, the samples pictures are labeled, training obtains human body sorter model
Parameter.
The position of human body is labeled as position of human body point；Judge the position of human body point whether on the boundary of enclosed region
On；If so, determining human body in the detection zone of building；If it is not, doing a ray with position of human body point, judges ray and close
Whether the intersection point sum for closing region is odd number；If the intersection point sum is odd number, determine human body in the detection zone of building.
The computer readable storage medium can be the internal storage unit of terminal described in aforementioned any embodiment, example
Such as the hard disk or memory of terminal.The computer readable storage medium is also possible to the External memory equipment of the terminal, such as
The plug-in type hard disk being equipped in the terminal, intelligent memory card (Smart Media Card, SMC), secure digital (Secure
Digital, SD) card, flash card (Flash Card) etc..Further, the computer readable storage medium can also be wrapped both
The internal storage unit for including the terminal also includes External memory equipment.The computer readable storage medium is described for storing
Other programs and data needed for computer program and the terminal.The computer readable storage medium can be also used for temporarily
When store the data that has exported or will export.
Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure
Member and algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware
With the interchangeability of software, each exemplary composition and step are generally described according to function in the above description.This
A little functions are implemented in hardware or software actually, the specific application and design constraint depending on technical solution.Specially
Industry technical staff can use different methods to achieve the described function each specific application, but this realization is not
It is considered as beyond the scope of this invention.
It is apparent to those skilled in the art that for convenience of description and succinctly, the end of foregoing description
The specific work process at end and unit, can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
In several embodiments provided by the present invention, it should be understood that disclosed terminal and method can pass through it
Its mode is realized.For example, the apparatus embodiments described above are merely exemplary, for example, the division of the unit, only
Only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components can be tied
Another system is closed or is desirably integrated into, or some features can be ignored or not executed.In addition, shown or discussed phase
Mutually between coupling, direct-coupling or communication connection can be through some interfaces, the INDIRECT COUPLING or communication of device or unit
Connection is also possible to electricity, mechanical or other form connections.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.Some or all of unit therein can be selected to realize the embodiment of the present invention according to the actual needs
Purpose.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, is also possible to two or more units and is integrated in one unit.It is above-mentioned integrated
Unit both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the present invention
Portion or part steps.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can store journey
The medium of sequence code.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in various equivalent modifications or replace
It changes, these modifications or substitutions should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with right
It is required that protection scope subject to.
Claims (10)
1. a kind of region prevention method based on human testing, which comprises the following steps:
S1 chooses picture, and detection zone is constructed in the picture, and the detection zone includes prevention region；
S2 carries out region nomination using k different rectangle frames on the convolution characteristic layer of the picture, filters out human body candidate
Region, wherein k is positive integer；
S3 carries out feature extraction to the candidate region automatically using convolutional neural networks；
S4 is classified using feature of the human body sorter model to the extraction, is returned, and identifies the picture with the presence or absence of people
Body.
S5, and if it exists, then judge human body whether in the detection zone of building；
S6, if so, executing alarm operation.
2. the region prevention method based on human testing as described in claim 1, which is characterized in that the picture in selection
Middle building detection zone, comprising:
At least three angle point is marked on the picture；
It is sequentially connected each angle point, forms enclosed region, the enclosed region is detection zone.
3. the region prevention method based on human testing as described in claim 1, which is characterized in that the k value 9.
4. the region prevention method based on human testing as claimed in claim 3, which is characterized in that before the step S2 also
Include: to collect the samples pictures comprising human body, the samples pictures are labeled, training obtains the ginseng of human body sorter model
Number.
5. the region prevention method based on human testing as claimed in claim 4, which is characterized in that the feature packet of the extraction
Include the textural characteristics, edge feature and motion feature of image；
The textural characteristics include the grey level histogram, edge orientation histogram, gray level co-occurrence matrixes of image；
The edge feature includes perimeter, area, the ratio of width to height, the dispersion degree, tightness of image；
The motion feature includes movement mass center, speed, displacement and gradient.
6. the region prevention method based on human testing as described in claim 1, which is characterized in that described whether to judge human body
In the detection zone of building, comprising:
The position of human body is labeled as position of human body point；
A ray is done with position of human body point, judges whether the intersection point sum of ray and enclosed region is odd number；
If the intersection point sum is odd number, determine human body in the detection zone of building.
7. the region prevention method based on human testing as claimed in claim 6, which is characterized in that described with position of human body point
Before doing a ray, further includes:
Judge the position of human body point whether on the boundary of enclosed region；
If so, determining human body in the detection zone of building；
A ray is done with position of human body point if it is not, then executing, judges whether the intersection point sum of ray and enclosed region is odd number
The step of.
8. such as the described in any item region prevention methods based on human testing of claim 1-7, which is characterized in that the picture
For the I frame picture in live video stream.
9. a kind of region crime prevention system based on human testing characterized by comprising appoint for executing claim 1-8 such as
The unit of method described in one.
10. a kind of terminal, which includes processor, input equipment, output equipment and memory, and the processor, input are set
Standby, output equipment and memory are connected with each other, which is characterized in that the memory supports terminal to execute as right is wanted for storing
The application code of the described in any item methods of 1-8 is sought, the processor is configured for executing as claim 1-8 is any
Method described in.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811480634.6A CN109583396A (en) | 2018-12-05 | 2018-12-05 | A kind of region prevention method, system and terminal based on CNN two stages human testing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811480634.6A CN109583396A (en) | 2018-12-05 | 2018-12-05 | A kind of region prevention method, system and terminal based on CNN two stages human testing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109583396A true CN109583396A (en) | 2019-04-05 |
Family
ID=65927235
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811480634.6A Pending CN109583396A (en) | 2018-12-05 | 2018-12-05 | A kind of region prevention method, system and terminal based on CNN two stages human testing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109583396A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109911550A (en) * | 2019-04-17 | 2019-06-21 | 华夏天信（北京）智能低碳技术研究院有限公司 | Scratch board conveyor protective device based on infrared thermal imaging and visible light video analysis |
| CN113139427A (en) * | 2021-03-12 | 2021-07-20 | 浙江智慧视频安防创新中心有限公司 | Steam pipe network intelligent monitoring method, system and equipment based on deep learning |
| CN114724256A (en) * | 2022-04-19 | 2022-07-08 | 盐城鸿石智能科技有限公司 | Human body induction control system and method with image analysis function |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101339688A (en) * | 2008-08-27 | 2009-01-07 | 北京中星微电子有限公司 | Intrusion checking method and system |
| CN106355162A (en) * | 2016-09-23 | 2017-01-25 | 江西洪都航空工业集团有限责任公司 | Method for detecting intrusion on basis of video monitoring |
| CN107331097A (en) * | 2017-08-01 | 2017-11-07 | 中科融通物联科技无锡有限公司 | The periphery intrusion preventing apparatus and method merged based on target position information |
| CN108121986A (en) * | 2017-12-29 | 2018-06-05 | 深圳云天励飞技术有限公司 | Object detection method and device, computer installation and computer readable storage medium |
| US20180158189A1 (en) * | 2016-12-07 | 2018-06-07 | Samsung Electronics Co., Ltd. | System and method for a deep learning machine for object detection |
| CN108830280A (en) * | 2018-05-14 | 2018-11-16 | 华南理工大学 | A kind of small target detecting method based on region nomination |
| CN108875577A (en) * | 2018-05-11 | 2018-11-23 | 深圳市易成自动驾驶技术有限公司 | Object detection method, device and computer readable storage medium |
-
2018
- 2018-12-05 CN CN201811480634.6A patent/CN109583396A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101339688A (en) * | 2008-08-27 | 2009-01-07 | 北京中星微电子有限公司 | Intrusion checking method and system |
| CN106355162A (en) * | 2016-09-23 | 2017-01-25 | 江西洪都航空工业集团有限责任公司 | Method for detecting intrusion on basis of video monitoring |
| US20180158189A1 (en) * | 2016-12-07 | 2018-06-07 | Samsung Electronics Co., Ltd. | System and method for a deep learning machine for object detection |
| CN107331097A (en) * | 2017-08-01 | 2017-11-07 | 中科融通物联科技无锡有限公司 | The periphery intrusion preventing apparatus and method merged based on target position information |
| CN108121986A (en) * | 2017-12-29 | 2018-06-05 | 深圳云天励飞技术有限公司 | Object detection method and device, computer installation and computer readable storage medium |
| CN108875577A (en) * | 2018-05-11 | 2018-11-23 | 深圳市易成自动驾驶技术有限公司 | Object detection method, device and computer readable storage medium |
| CN108830280A (en) * | 2018-05-14 | 2018-11-16 | 华南理工大学 | A kind of small target detecting method based on region nomination |
Non-Patent Citations (2)
| Title |
|---|
| 刘鑫昱: ""面向监控图像的行人再识别关键技术研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 董凌凌: ""运动目标的特征提取与行为识别研究"" * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109911550A (en) * | 2019-04-17 | 2019-06-21 | 华夏天信（北京）智能低碳技术研究院有限公司 | Scratch board conveyor protective device based on infrared thermal imaging and visible light video analysis |
| CN113139427A (en) * | 2021-03-12 | 2021-07-20 | 浙江智慧视频安防创新中心有限公司 | Steam pipe network intelligent monitoring method, system and equipment based on deep learning |
| CN114724256A (en) * | 2022-04-19 | 2022-07-08 | 盐城鸿石智能科技有限公司 | Human body induction control system and method with image analysis function |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110390262B (en) | Video analysis method, device, server and storage medium | |
| Medel et al. | Anomaly detection in video using predictive convolutional long short-term memory networks | |
| CN109670441A (en) | A kind of realization safety cap wearing knows method for distinguishing, system, terminal and computer readable storage medium | |
| CN109154976B (en) | System and method for training object classifier through machine learning | |
| US9008365B2 (en) | Systems and methods for pedestrian detection in images | |
| CN108090458A (en) | Tumble detection method for human body and device | |
| CN110188807A (en) | Tunnel pedestrian target detection method based on cascade super-resolution network and improvement Faster R-CNN | |
| CN108009477B (en) | Image people flow number detection method and device, storage medium and electronic equipment | |
| CN111766253A (en) | Solder paste printing quality detection method, data processing device, and computer storage medium | |
| CN112200081A (en) | Abnormal behavior identification method and device, electronic equipment and storage medium | |
| CN111813997A (en) | Intrusion analysis method, device, equipment and storage medium | |
| CN109583396A (en) | A kind of region prevention method, system and terminal based on CNN two stages human testing | |
| CN109360362A (en) | A kind of railway video monitoring recognition methods, system and computer-readable medium | |
| CN111598132B (en) | Portrait recognition algorithm performance evaluation method and device | |
| CN112052815B (en) | Behavior detection method and device and electronic equipment | |
| CN109711322A (en) | A kind of people's vehicle separation method based on RFCN | |
| CN110189355A (en) | Safe escape channel occupies detection method, device, electronic equipment and storage medium | |
| CN110348343A (en) | A kind of act of violence monitoring method, device, storage medium and terminal device | |
| CN110533053A (en) | A kind of event detecting method, device and electronic equipment | |
| CN113706481A (en) | Sperm quality detection method, sperm quality detection device, computer equipment and storage medium | |
| CN113780145A (en) | Sperm morphology detection method, sperm morphology detection device, computer equipment and storage medium | |
| WO2022194069A1 (en) | Saliency map generation method, and abnormal object detection method and device | |
| CN110795975A (en) | Face false detection optimization method and device | |
| Mi et al. | Research on a Fast Human‐Detection Algorithm for Unmanned Surveillance Area in Bulk Ports | |
| CN109754077A (en) | Network model compression method, device and the computer equipment of deep neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190405 |