CN105141924B - Wireless image monitoring system based on 4G technologies - Google Patents
Wireless image monitoring system based on 4G technologies Download PDFInfo
- Publication number
- CN105141924B CN105141924B CN201510594468.2A CN201510594468A CN105141924B CN 105141924 B CN105141924 B CN 105141924B CN 201510594468 A CN201510594468 A CN 201510594468A CN 105141924 B CN105141924 B CN 105141924B
- Authority
- CN
- China
- Prior art keywords
- msub
- mrow
- image
- video
- wireless communication
- 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.)
- Active
Links
Landscapes
- Closed-Circuit Television Systems (AREA)
Abstract
The invention discloses a kind of wireless image monitoring system based on 4G technologies, the system includes：Image monitor in unmanned plane and the video frequency transmitter installed in ground central station；Wherein, image monitor includes：Monitoring processor, satellite navigation unit, high definition motion cameras, generator terminal 4G wireless communication modules on unmanned plane；The monitoring processor is using SOC single-chips multimedia processor as core, with video input and output interface, audio input output interface, video A/D/digital analog converter, storage and network communication interface；Video frequency transmitter includes：Stand end 4G wireless communication modules, central site image processing module and display terminal；The station end 4G wireless communication modules, receive the picture signal of the generator terminal 4G wireless communication modules.System that employs 4G cordless communication networks to realize transmission of video, enhances the freedom and flexibility ratio of video monitoring, improves operating efficiency, system cost is low, and has higher security.
Description
Technical field
The present invention relates to picture control field, and in particular to a kind of wireless image monitoring system based on 4G technologies.
Background technology
The video of unmanned plane shooting at present is typically all to be transmitted the video to by graphic transmission equipment on earth station system,
Then observer can be on ground base station, the video of real time inspection unmanned plane shooting, but due to graphic transmission equipment and day
The limitation of line so that the distance between position and unmanned plane where ground base station must be in certain scopes, so as to cause
Observer also must can not carry out real time inspection unmanned plane bat with ground base station within this range if it have left this scope
The video taken the photograph, it is applied to that there is very big limitation.
The key that UAV Video transmission application is realized is wireless transmission link means.Current Radio Transmission Technology master
Have including following technology：It is 3G network (CDMA2000, WCDMA, TD-SCDMA), 4G (TD-LTE and FDD-LTE) network, wireless
LAN (WIFI), satellite, microwave etc..
Satellite and microwave technology are the traditional means of wireless video transmission, and the great advantage of communication technology of satellite is service
Scope is wide, powerful, using flexible, is not influenceed by geographical environment and other external environment conditions, especially not by external electromagnetic
The influence of environment.But both technical costs remain high, its expensive initial expenditure of construction and communication fee often make one to hope and
Step back, can not be widely applied.
The technique construction such as WIMAX/WIFI wireless MAN needs Construction Party to build come the Video Applications covered on a large scale
If a large amount of base stations, one side base station construction cost is huge, and non-general user can bear；On the other hand even if a certain unit is built up
Wireless MAN, because its initial construction cost is huge is reluctant to be shared with other users, so as to be made to social resources
Into larger waste.
Fourth generation mobile phone mobile communication standard, refer to fourth generation mobile communication technology, foreign language abbreviation：4G；The skill
Art includes two kinds of standards of TD-LTE and FDD-LTE, and (from stricti jurise, LTE is 3.9G, is wirelessly marked for 4G although being promoted
Standard, but it is not recognized as by 3GPP the wireless communication standard IMT- of future generation described by International Telecommunication Union in fact
Advanced, therefore it is also not up to 4G standard in a strict sense.The LTE Advanced of only upgrade version just meet state
Requirement of the border telecommunication union to 4G)；4G is to integrate 3G and WLAN, and can quickly transmit data, high quality, audio, regard
Frequency and image etc.；4G can be downloaded with more than 100Mbps speed, than fast 25 times current of home broadband AD SL (4,000,000), and energy
Enough meet requirement of nearly all user for wireless service；In addition, 4G can not have in DSL and Cable Modem
The place deployment of covering, then expands to whole distract again；It is obvious that 4G has incomparable superiority.
The content of the invention
The present invention provides a kind of wireless image monitoring system based on 4G technologies, and the system supports vision guided navigation, image to know
Not and avoidance, 4G cordless communication networks are employed and realize transmission of video, enhance the freedom and flexibility ratio of video monitoring, and greatly
It is big to reduce installation wiring work, operating efficiency is improved, reduces system cost, the high speed that can solve large capacity image data is handed over
Change, and there is higher security.
To achieve these goals, the present invention provides a kind of wireless image monitoring system based on 4G technologies, the system bag
Include：
Image monitor in unmanned plane and the video frequency transmitter installed in ground central station；
Wherein, image monitor includes：
Monitoring processor, satellite navigation unit, high definition motion cameras, generator terminal 4G radio communications on unmanned plane
Module；
The monitoring processor is using SOC single-chips multimedia processor as core, with video input and output interface, audio
Input/output interface, video A/D/digital analog converter, storage and network communication interface；
Video frequency transmitter includes：
Stand end 4G wireless communication modules, central site image processing module and display terminal；
The station end 4G wireless communication modules, receive the picture signal of the generator terminal 4G wireless communication modules.
Preferably, SOC single-chips multimedia processor is connected completion with generator terminal 4G wireless communication modules by usb bus and regarded
The 4G wireless communication transmissions of frequency image.
Preferably, the monitoring processor using SOC single-chip multimedia processor i.MX27 as core processor, its
Using ARM926 as core I P, processing operation real time operating system Linux.
Preferably, SOC single-chips multimedia processor passes through SDR bus external data memory storage SDRAMs；It is total by EMI
Line Add-In memory NANDFlash；High definition motion cameras is connected by CSI interface；Pass through I2S bus external audio frequencies AD
Converter.
Preferably, the central site image processing module includes：
Acquiring unit, the acquiring unit are used to obtain the frame transmitted from station end 4G wireless communication modules,
Obtain the image of frame expression；
Denoising unit, the denoising unit are used to remove the noise number in rule removal described image according to predetermined noise
According to；
Recognition unit, the recognition unit are used for the image in the removal noise data according to predetermined object recognition rule
Middle identification destination object；
Adding device, the adding device are used to add label for the frame, and the label can be based on semantic meaning representation target
The predetermined characteristic of object；
Memory cell, the memory cell are used to store label corresponding to the frame.
The present invention has advantages below and beneficial effect：(1) support high-definition digital image to pass ground back in real time, meet high definition
Digital Transmission requirement, vision guided navigation, obstacle avoidance and images steganalysis tracking are supported, meets development of new techniques requirement；(2) in
The pre-defined algorithm of center station dot image processing module, it is easy to the high-layer semantic information of people's intuitivism apprehension, and realization pair on this basis
The classification of video monitoring image data and mark, realize the fast and efficiently acquisition of video monitoring image.
Brief description of the drawings
Fig. 1 shows a kind of block diagram of wireless image monitoring system based on 4G technologies of the present invention.
Fig. 2 shows a kind of wireless image monitoring method based on 4G technologies of the present invention.
Embodiment
Fig. 1 is a kind of wireless image monitoring system based on 4G technologies for showing the present invention.The system includes：It is arranged on
Image monitor 1 in unmanned plane and the video frequency transmitter 2 installed in ground central station.
Wherein, image monitor 1 includes：Monitoring processor 11, satellite navigation unit 13, height on unmanned plane
Clear motion cameras 12, generator terminal 4G wireless communication modules 14 and vision computer 15.
The monitoring processor 11 is also embedded with Ethernet switching chip (LANswitch), the Ethernet switching chip
(LANswitch) it is connected with flight control computer 15 (ARM) by LAN (LAN).
The monitoring processor 11 is using SOC single-chips multimedia processor as core, with video input and output interface, sound
Frequency input/output interface, video A/D/digital analog converter, storage and network communication interface；
Video frequency transmitter 2 includes：End 4G wireless communication modules 21, central site image processing module 22 and display stand eventually
End 23；The station end 4G wireless communication modules 21, receive the picture signal of the generator terminal 4G wireless communication modules 14.
Preferably, SOC single-chips multimedia processor is connected completion with generator terminal 4G wireless communication modules by usb bus and regarded
The 4G wireless communication transmissions of frequency image.
Preferably, the monitoring processor using SOC single-chip multimedia processor i.MX27 as core processor, its
Using ARM926 as core I P, processing operation real time operating system Linux.
Preferably, SOC single-chips multimedia processor passes through SDR bus external data memory storage SDRAMs；It is total by EMI
Line Add-In memory NANDFlash；High definition motion cameras is connected by CSI interface；Pass through I2S bus external audio frequencies AD
Converter.
Have DSP Processor, arm processor inside the vision computer 15, run (SuSE) Linux OS, with 100,000,000 with
Too network interface is connected with the flight control computer, is expanded by the Ethernet switching chip (LANswitch) of the monitoring processor
The Ethernet exchanging formula bus of exhibition receives the picture that high definition Flying Camera is passed back, and the analysis for carrying out image resolves, and is passed with light stream
Sensor, ultrasonic sensor, Inertial Measurement Unit data are merged, and carry out vision guided navigation, obstacle avoidance, images steganalysis
Tracking.
The Ethernet exchanging formula that the high definition motion cameras 12 is directly extended by Ethernet interface and monitoring processor 11
Bus is attached, and supports the forwarding of multiple video flowings, by Ethernet switching chip (LANswitch) by HD video data
It is transmitted to vision computer (DSP+ARM) and carries out image calculating.
The satellite navigation unit 13 is GPS/ Big Dippeves reception chip, magnetic compass, single-chip microcomputer, goes out CAN and is counted with flying control
Calculation machine (ARM) connects, and supports GPS and Beidou navigation positioning, supports resolving of the magnetometer to attitude of flight vehicle, and survey with inertia
Measure unit (IMU) and carry out data fusion, attitude of flight vehicle and position of aircraft are finally resolved by monitoring processor 11.
Video frequency transmitter 2 includes：Stand end 4G wireless communication modules 21, multichannel distribution module 22, central site image
Processing module 23 and display terminal 24.The station end 4G wireless communication modules 21, are received through satellite network or mobile communications network
The picture signal of described image transmitter module transmitting 14；Described multichannel distribution module 22 is by video compression encoder, more letters
Road communications distribution equipment, communication equipment, gateway device composition, described communication equipment include wired transmission equipment, short distance without
Line communication equipment, mobile communication equipment, satellite communication equipment, described center image processing system are shown by decoding device, image
Show that equipment forms.
Preferably, the central site vision computer 15 includes：
Acquiring unit, the acquiring unit are used to obtain the frame transmitted from station end 4G wireless communication modules,
Obtain the image of frame expression.
Denoising unit, the denoising unit are used to remove the noise number in rule removal described image according to predetermined noise
According to；Image is obtaining, usually can made image deterioration by interference and the influence of various noises in transmission and storing process.In order to
Obtain the digital picture of high quality, it is necessary to noise reduction process is carried out to image, is keeping the same of raw information integrality as far as possible
When, and can enough removes information useless in signal.In view of video monitoring system is the spy of the monitoring to movable destination object mostly
Different property, in the embodiment of the application, the immovable background that will be not required to monitoring or key monitoring is divided with movable prospect
From the background parts for the monitor video that will be obtained remove as a part for noise data.
Recognition unit, the recognition unit are used for the image in the removal noise data according to predetermined object recognition rule
Middle identification destination object.The purpose retrieved to image is to identify destination object therein, first has to extract destination object
Feature, and according to this feature identification object.Therefore one of subject matter of image retrieval is exactly the extraction of characteristics of the underlying image.
The application embodiment is based on extracting the clarification of objective in the image after denoising to realize the identification of destination object.
Adding device, the adding device are used to add label for the frame, and the label can be based on semantic meaning representation target
The predetermined characteristic of object.After the identification for completing destination object, you can fill label, the label of filling to the destination object identified
Being capable of the expression based on the high-layer semantic information of the intuitivism apprehension of people.
Memory cell, the memory cell are used to store label corresponding to the frame.
Fig. 2 shows a kind of wireless image monitoring method based on 4G technologies of the present invention.This method specifically includes as follows
Step：
S1. monitoring processor, which starts, flies control program, and the satellite navigation unit starts GPS navigation program；
S2. high definition motion cameras gathers video image according to the track of winged control program, and vision computer is carried out to image
Processing；
S3. generator terminal 4G wireless communication modules, and station end 4G wireless communication modules, coordinate the wireless transmission for completing picture signal
And reception；
S4. central site image processing module is handled the picture signal received, and is shown on display terminal.
In addition to following navigator fix step preferably, in step sl,：
Monitoring processor 11 transmits the location data come to satellite navigation unit 13 and judged：
If location data is in normal range (NR)：Then the location data received is stored in memory by monitoring processor 11；
The location data in normal range (NR) refers to：By the longitude of two neighboring sampled point, latitude in location data
Value, height value are compared two-by-two, if the difference of the longitude of two neighboring sampled point is no more than 0.0002 degree, and two neighboring are adopted
The difference of the latitude of sampling point is no more than 0.00018 degree, and the difference of the height of two neighboring sampled point is no more than 20 meters, and it is fixed to judge
Position data are normal range (NR)；
If location data occurs abnormal：Then monitoring processor 11 recalls the location data stored in memory, according to
Historical track returns to homeposition；
Abnormal refer to occurs for the location data：By the longitude of two neighboring sampled point, latitude value, height in location data
Angle value is compared two-by-two, if the difference of longitude more than 0.0002 degree, or latitude difference more than 0.00018 degree, or height
It is abnormal then to judge that location data occurs more than 20 meters for difference.
Preferably, the location data is longitude information x, latitude information y, elevation information of the unmanned plane at each time point
Z set, it is designated as { xt yt zt }；Wherein,
(x1 y1 z1) is longitude, latitude, elevation information of the unmanned plane the 1st time point；
(x2 y2 z2) is longitude, latitude, elevation information of the unmanned plane the 2nd time point；
By that analogy, (xt-1 yt-1 zt-1) is that unmanned plane is believed in the longitude at the t-1 time point, latitude, height
Breath；(xt yt zt) is longitude, latitude, elevation information of the unmanned plane t-th of time point；
The interval at two neighboring time point takes 0.5 to 5.0 second；Each historical location data is stored in monitoring processor 11
Memory in；
By the location data at t-th of time point compared with the location data at the t-1 time point：
If xt-xt-1 ＜ 0.0002, and yt-yt-1 ＜ 0.00018, and 20 meters of zt-zt-1 ＜,
I.e. the difference of longitude is no more than 0.0002 degree, and the difference of latitude is no more than 0.00018 degree, and the difference of height does not surpass
When crossing 20 meters, judge that the location data at t-th of time point belongs to normal range (NR), and the location data at t-th of time point is deposited
Enter the memory of monitoring processor 11；
If xt-xt-1 >=0.0002, or yt-yt-1 >=0.00018, or zt-zt-1 >=20 meter；That is the difference of longitude, latitude
Any one in the difference of degree, the difference of height exceeds normal range (NR), and judging the location data at t-th of time point, there occurs different
Often, namely think the flight of unmanned plane there occurs exception；
By monitoring processor 11 by the location data at the t-1 time point in memory, the positioning at the t-2 time point
Data ... the location data at the 2nd time point, the location data at the 1st time point gradually read, and control unmanned vehicle
The departure place returned according to original track.
Preferably, in step sl, flying control program is included at application layer program, real-time task scheduler and external interrupt
Manage program, hardware initialization program, hardware drive program, CAN communication protocol procedure, LAN (TCP/IP) communication protocol program, institute
State application layer program to be connected with real-time task scheduler and external interrupt processor, the real-time task scheduler and outer
Portion's interrupt handling routine is connected with hardware initialization program, and the hardware initialization program is connected with hardware drive program.
Preferably, the application layer program includes Applied layer interface program, power management and electric quantity monitoring program, flight refers to
Show lamp control program, security control program, visual spatial attention program, flight tracking control program, augmentation control program, remote control decoding journey
Sequence, communication processing program.
Including following sub-step preferably, in step s 2,：
S21. the video file dispenser of vision computer 15 is split to video file；
S22. the file that the video compression encoder of vision computer 15 is completed to segmentation is compressed；
S23. operation is encrypted to the video file compressed in the encryption device of vision computer 15.
Preferably, in step s 4, video image can be handled with the following method：
S41：A frame in video is obtained, that is, obtains the image of frame expression.
S42：The noise data in rule removal described image is removed according to predetermined noise.
Image is obtaining, usually can drop image by interference and the influence of various noises in transmission and storing process
Matter.In order to obtain the digital picture of high quality, it is necessary to carry out noise reduction process to image, keeping raw information complete as far as possible
While property, and information useless in signal can be removed.
The final purpose of video image denoising is to improve given image, and solve real image is caused due to noise jamming
The problem of image quality decrease.Picture quality is effectively improved by noise-removed technology, increases signal to noise ratio, preferably embodies original figure
As entrained information.
The method for carrying out denoising to image at present is divided into two classes substantially：Space domain method and transpositions domain.The former is
Data operation is directly carried out on original image, the gray value of pixel is handled；Common spatial domain Image denoising algorithm has
Neighborhood averaging, medium filtering, LPF etc..The latter is to carry out computing in the relevant spatial domain in processing pixel field,
Certain computing is carried out to image, image is changed into transform domain from transform of spatial domain, then the conversion coefficient in transform domain is handled,
Carry out again inverse transformation by image from transform domain be transformed into spatial domain reach remove picture noise purpose.Wherein, Fourier transformation
It is the transform method that is commonly used in image denoising with wavelet transformation.Because denoising method is more ripe technology, therefore this
Application embodiment can not form the limitation to application according to the actual conditions unrestricted choice above method.
In view of video monitoring system is the particularity of the monitoring to movable destination object mostly, the embodiment party of the application
In formula, the immovable background for being not required to monitoring or key monitoring is separated with movable prospect, the monitor video that will be obtained
Background parts remove as a part for noise data.
S43：According to predetermined object recognition rule destination object is identified in the image of the removal noise data.
The purpose retrieved to image is to identify destination object therein, first has to extract the feature of destination object,
And according to this feature identification object.Therefore one of subject matter of image retrieval is exactly the extraction of characteristics of the underlying image.
The application, which can extract characteristics of the underlying image, can include color, texture, shape and the depth of field.
1st, color
Color is a kind of very important visual property of body surface, is one of the main Perception Features that people identifies image；With retouching
State the shape of image, Texture eigenvalue is compared, color characteristic is the most basic Image Visual Features of CBIR, is graphical representation and inspection
The most straightforward approach used in rope, main reason is that color characteristic calculate it is simple, in its information and image specific object and
Scene type is highly dependent.In addition, color characteristic is relatively small to the dependence at image size in itself, direction, visual angle.It is but real
In border, the different cameras caused by the difference meetings such as environment illumination intensity, shooting angle, imaging characteristic, object distance gathers
To the color of same target there is difference.In order to solve this problem, obtain stably, with uniquely target signature table
Reach, color transfer method or color changeover method can be utilized to eliminate color distortion, improve the robustness of color characteristic.
, can be first to collecting before eliminate color distortion using color transfer method or color changeover method
Video monitoring image carries out enhancing pretreatment.
Research shows that human visual system perceives the intensity of illumination of object in a manner of nonlinearity.But video camera
It is then relatively simple etc. imaging process.Generally, it is different, this feelings that video camera imaging directly perceives from the mankind
Condition is more obvious when the dynamic range of object is larger.Dynamic range refers to the ratio between most bright and most dark object brightness in scene
Value.As a result of the method for region adaptivity, human visual system can be perceived more than 1000：1 dynamic range, and it is common
Display can only show 100：1 dynamic range., need to be to figure when object dynamic range is more than the scope that display can be shown
As carrying out dynamic range compression, to be adapted to display.The global logarithmic function of simple tone mapping method use,
Gamma is corrected or Sigmoid functions compress the dynamic range of image, easily causes local loss in detail.Higher color
The method for adjusting mapping to use region adaptivity, the method based on Retinex is one type.
Retinex theories are color and the brightness that object is perceived based on human vision regulation that Land proposed in 1963
Model, its basic thought is that people perceives illumination of certain point and is not dependent on the absolute illumination value of the point, also and around it
Illumination value is relevant.Retinex enhancings processing can improve color of image shape constancy, compress dynamic range of images, improve contrast,
Effectively display is submerged in the details in shadow region.It is right first that Retinex methods, which were applied in the step of the application embodiment,
The video monitoring image collected carries out illumination estimation, and illumination is then subtracted from video monitoring image in log-domain, suppresses
Influence of the illumination variation to image, obtains enhanced image.
After carrying out enhancing processing to the video monitoring image collected using Retinex algorithm, using color transfer or
Color robin carries out color difference eliminating processing to the video monitoring image of the enhancing, improves the robustness of color characteristic.Video is supervised
It is to remove the important component of picture noise that control image, which carries out color difference eliminating processing,.
2nd, texture
Textural characteristics are the shared intrinsic characteristic in all objects surface and the reflection to imaging surface space structure and attribute,
It is a kind of visual signature independent of color or the reflection image homogeneity phenomenon of brightness.Textural characteristics contain body surface knot
The important information of structure tissue line, it shows as the regularity of gray scale or distribution of color on image, therefore is normally thought of as image
Certain local property, or localized region between pixel relation a kind of measurement.
Conventional image texture characteristic has co-occurrence matrix, wavelet texture, Tamura textural characteristics etc..Wherein, Haralick
Deng the method that image texture characteristic is described using co-occurrence matrix, the spatial dependence of gray level in image is have studied from mathematical angle
And the statistical information of this dependence is recorded in the form of matrix.What gray level co-occurrence matrixes counted is the space point of color intensity
Cloth information, co-occurrence matrix (co-occurrence matrix) is constructed according to the azimuth-range relation between image pixel, therefrom
Extract description of the significant statistical nature as textural characteristics.
The characteristics of video monitoring image is that target is often in mobile status.In non-rigid object tracking or long-time mesh
In mark tracking, the global characteristics such as objective contour may change greatly, and local feature then has good consistency, therefore local special
It is preferably to select to levy point methods.
The application embodiment can use local binary (Local Binary Pattern, LBP) description right
Face is detected, to improve retrieval precision and retrieval rate to pedestrian.
3rd, shape
Shape facility is important in image expression and image understanding based on the segmentation to objects in images and region
Feature.Intuitively, conversion of the people to body form, rotation and scaling are insensitive, so the shape facility of extraction also has pair
The consistency answered, it is a kind of effective image-region shape descriptor.
4th, the depth of field
For visual attention, the depth of field extracted from single image be one in general manner, top-down ground feature,
Defocusing blurring just occurs when target is placed on beyond the focal zone of video camera.
The extracting method of the application embodiment depth of field feature can include two key steps.First, single width figure is estimated
As the fog-level at edge.Then, Gauss weighting is carried out so as to obtain the relative scape of each fundamental region to the fuzziness at edge
It is deep.The circular of single image depth map is as follows：
First, carried out with the Gauss collecting image that standard deviation is σ 1 secondary fuzzy.Then, calculated in the edge of image former
The ratio T of the gradient of beginning image and the gradient of secondary blurred picture.Fog-level σ at image border can be according to equation below
Calculate：
The application edge of Canny rim detections son extraction image, the standard deviation for setting secondary Gaussian Blur is σ 1=
1.Then, the fog-level σ of all edges of image is normalized to [0,1] section.
Then, fundamental region m relative depth of field Wm is defined as the Gauss weighted average of all edge blurry degree of image：
(i, j) is the coordinate of fundamental region m pixel, and σ (i, j) is edge pixel point M (i, j) fog-level, high
This weight definition is：
WhereinNeighborhood Vij, the σ W for belonging to (i, j) are the secondary Gaussian Blur standard deviation of the relative depth of field, and it is used
The suppression depth of field of sensitiveness in to(for) distance between edge pixel point M (i, j) and fundamental region m.σ W values have to depth of field feature
Considerable influence, the excessive then depth of field of value tend to identical, and value is too small, strengthens On Local Fuzzy.σ W are set in the application embodiment
It is set to σ W=0.15.
The low-level image features such as the above-mentioned color referred to, texture, shape, the depth of field are global characteristics.Certainly, global characteristics are not
It is limited to above-mentioned four kinds of features, such as it can also include GIST features and fractal characteristic, will not be repeated here.In addition, image
Local feature, such as SIFT feature can also be included.
The purpose of CBIR is on the basis of image vision low-level image feature is extracted, and is looked for from image library
Go out associated picture.The content characteristic of image includes low-level image feature and high-level semantics features, with the color of extraction, texture, shape, scape
The low-level image feature of deep character representation image, low-level image feature is selected and combined preferably to simulate people to image high level language by training
The direct feel of adopted feature, it is convenient to map image vision low-level image feature to obtain the high-level semantics features of image.
In order to which the later stage is easy to retrieve, according to the low-level image feature of extraction, first the video monitoring image of acquisition can be classified.
The identification of each semantic category is considered as a two independent classification problems.Assuming that all video monitoring images share m classes, L is designated as
=A1, A2 ... Am }, the amount of images for belonging to semantic category Ai is Ni, and the classification problem of m classes is converted into two class classification problems：
For any one class Ai, training positive example is all images that such is included, and counter-example is to be not belonging to this in training set
The positive example sum of the image of the every other class of class, i.e. Ai classes is Ni, and counter-example sum is
To a given semantic category A ∈ L, the training set T=of its two classes classification problem (x1, y1), (x2, y2) ...,
(xl, yl) }；(xi, yi) represents given in advance and passes through one group of image of semantic filling label, and wherein xi ∈ Rn are a figure
As vector, the image for belonging to the features such as same or analogous color, texture, shape and the depth of field is represented.Yi ∈ {+1, -1 }, if yi
=+1 represents xi ∈ A, i.e. the image that vector x i is represented belongs to semantic classes A.Similarly, yi=-1 is represented
S44：Label is added for the frame, the label can the predetermined characteristic based on semantic meaning representation destination object.
After the identification for completing destination object, you can fill label to the destination object identified, the label of filling being capable of base
In the expression of the high-layer semantic information of the intuitivism apprehension of people.Feature expressed by these labels of filling is easy for intuitivism apprehension
High-layer semantic information.
S45：The corresponding frame stores the label, and later stage acquisition tag library is easy in formation.
The application embodiment is by the vision low-level image feature to extraction, according to predetermined Algorithm mapping to being easy to people directly perceived
The high-layer semantic information of understanding, and the classification to video monitoring image data and mark are realized on this basis, preferably expression
The semanteme of video monitoring image data, reduce " the semanteme even eliminated between characteristics of the underlying image and the abundant semantic content of the mankind
Wide gap ", realize the fast and efficiently acquisition of video monitoring image.
S46：Inquiry request is received, the inquiry request is accompanied with keyword.
When needing to inquire about destination object, inquiry request is received, is accompanied with advance to target pair in inquiry request
As the keyword being defined.
S47：The keyword is searched in the label of the storage, is obtained corresponding with the keyword identical label
Frame.
S48：The frame obtained sequentially in time described in arrangement.
Obtained all frames with the destination object are arranged sequentially in time；Further, by Time Continuous
Frame combines to form video, using discontinuous frame as single image.So destination object can be eliminated to a certain extent to exist
Time and interruption spatially, direct objective information is provided to study and judge destination object motion track.
As described above, although the embodiment and accompanying drawing that are limited according to embodiment are illustrated, to the art
Various modification and variation can be carried out for technical staff with general knowledge from above-mentioned record.For example, according to explanation
Technology in illustrated method mutually different order carry out, and/or according to the system with explanation, structure, device, circuit etc.
The mutually different form of method illustrated by inscape is combined or combined, or is carried out according to other inscapes or equipollent
Replace or displacement also may achieve appropriate effect.For general technical staff of the technical field of the invention, do not taking off
On the premise of from present inventive concept, some equivalent substitutes or obvious modification are made, and performance or purposes are identical, should all be considered as
Belong to protection scope of the present invention.
Claims (4)
1. a kind of wireless image monitoring system based on 4G technologies, the system include：
Image monitor in unmanned plane and the video frequency transmitter installed in ground central station；
Wherein, image monitor includes：
Monitoring processor, satellite navigation unit, high definition motion cameras, generator terminal 4G radio communication molds on unmanned plane
Block；
The monitoring processor is using SOC single-chips multimedia processor as core, with video input and output interface, audio input
Output interface, video A/D/digital analog converter, storage and network communication interface；
Video frequency transmitter includes：
Stand end 4G wireless communication modules, central site image processing module and display terminal；
The station end 4G wireless communication modules, receive the picture signal of the generator terminal 4G wireless communication modules；
The central site image processing module includes：
Acquiring unit, the acquiring unit are used to obtain the frame transmitted from station end 4G wireless communication modules, that is, obtained
Obtain the image of frame expression；
Denoising unit, the denoising unit are used to remove the noise data in rule removal described image according to predetermined noise；
Recognition unit, the recognition unit are used in the image of the removal noise data know according to predetermined object recognition rule
Other destination object；
Adding device, the adding device are used to add label for the frame, and the label can be based on semantic meaning representation destination object
Predetermined characteristic；
Memory cell, the memory cell are used to store label corresponding to the frame；
Described image processing module is handled video image in the following way：
(1) frame in video is obtained, that is, obtains the image of frame expression；
(2) noise data in rule removal described image is removed according to predetermined noise；
(3) destination object is identified in the image of the removal noise data according to predetermined object recognition rule, first has to extract
The feature of destination object, and include color, texture, shape and the depth of field, scape according to this feature identification object, the feature of destination object
The extracting method of deep feature can include two key steps：
First, estimate the fog-level at single image edge, then, Gauss weighting is carried out so as to obtain often to the fuzziness at edge
The relative depth of field of individual fundamental region, the circular of single image depth map are as follows：
First, carried out with the Gauss collecting image that standard deviation is σ 1 secondary fuzzy, then, original graph is calculated in the edge of image
The ratio T of the gradient of picture and the gradient of secondary blurred picture, the fog-level σ at image border can be according to equation below meter
Calculate：
<mrow>
<mi>&sigma;</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<msqrt>
<mrow>
<msup>
<mi>T</mi>
<mn>2</mn>
</msup>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msqrt>
</mfrac>
<msub>
<mi>&sigma;</mi>
<mn>1</mn>
</msub>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
With the edge of Canny rim detections son extraction image, the standard deviation for setting secondary Gaussian Blur is σ 1=1, then, figure
As the fog-level σ of all edges normalizes to [0,1] section；
Then, fundamental region m relative depth of field Wm is defined as the Gauss weighted average of all edge blurry degree of image：
<mrow>
<msub>
<mi>W</mi>
<mi>m</mi>
</msub>
<mo>=</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>M</mi>
</mrow>
</munder>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>-</mo>
<mi>&sigma;</mi>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<msub>
<mi>i</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>j</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
(i, j) is the coordinate of fundamental region m pixel, and σ (i, j) is edge pixel point M (i, j) fog-level, Gao Siquan
Redefine for：
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<msub>
<mi>i</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>j</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mn>2</mn>
<msubsup>
<mi>&sigma;</mi>
<mi>W</mi>
<mn>2</mn>
</msubsup>
</mrow>
</mfrac>
<mo>|</mo>
<mo>|</mo>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
<mo>)</mo>
<mo>-</mo>
<mo>(</mo>
<mrow>
<msub>
<mi>i</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>j</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
</mrow>
<mo>)</mo>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>i</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>j</mi>
<msub>
<mi>p</mi>
<mi>m</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>&Element;</mo>
<msub>
<mi>V</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
WhereinNeighborhood Vij, the σ W for belonging to (i, j) are the secondary Gaussian Blur standard deviation of the relative depth of field, and it is used to suppress
For the depth of field for the sensitiveness of distance between edge pixel point M (i, j) and fundamental region m, σ W values have larger shadow to depth of field feature
Ring, the excessive then depth of field of value tends to identical, and value is too small, strengthens On Local Fuzzy, and σ W are arranged into σ W in the application embodiment
=0.15；
(4) label is added for the frame, the label can the predetermined characteristic based on semantic meaning representation destination object；
(5) the corresponding frame stores the label, and later stage acquisition tag library is easy in formation；
(6) inquiry request is received, the inquiry request is accompanied with keyword；
(7) keyword is searched in the label of the storage, obtains frame corresponding with the keyword identical label；
(8) frame obtained sequentially in time described in arrangement.
2. the system as claimed in claim 1, it is characterised in that the SOC single-chips multimedia processor by usb bus with
The 4G wireless communication transmissions of video image are completed in the connection of generator terminal 4G wireless communication modules.
3. system as claimed in claim 2, it is characterised in that the monitoring processor uses SOC single-chip multi-media processings
For device i.MX27 as core processor, it uses ARM926 as core I P, processing operation real time operating system Linux.
4. system as claimed in claim 3, it is characterised in that SOC single-chips multimedia processor passes through the external number of SDR buses
According to memory storage SDRAM；Pass through EMI bus Add-In memories NANDFlash；High definition Flying Camera is connected by CSI interface
Machine；Pass through I2S bus external audio frequency a/d converters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510594468.2A CN105141924B (en) | 2015-09-17 | 2015-09-17 | Wireless image monitoring system based on 4G technologies |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510594468.2A CN105141924B (en) | 2015-09-17 | 2015-09-17 | Wireless image monitoring system based on 4G technologies |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105141924A CN105141924A (en) | 2015-12-09 |
| CN105141924B true CN105141924B (en) | 2018-03-30 |
Family
ID=54727120
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510594468.2A Active CN105141924B (en) | 2015-09-17 | 2015-09-17 | Wireless image monitoring system based on 4G technologies |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105141924B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105578143A (en) * | 2015-12-29 | 2016-05-11 | 成都移动魔方科技有限公司 | Intelligent remote monitoring system |
| CN106452556A (en) * | 2016-09-14 | 2017-02-22 | 芜湖扬展新材料科技服务有限公司 | Data transmission system for aircraft based on 4G network |
| CN106412500A (en) * | 2016-09-14 | 2017-02-15 | 芜湖扬展新材料科技服务有限公司 | Embedded Web network monitoring system based on RTSP protocol |
| CN106645147A (en) * | 2016-11-16 | 2017-05-10 | 王金鹏 | Method for pest and disease damage monitoring |
| CN106568777A (en) * | 2016-11-16 | 2017-04-19 | 王金鹏 | Pest and disease monitoring system |
| CN106843251A (en) * | 2017-02-20 | 2017-06-13 | 上海大学 | Crowded crowd promptly dredges unmanned plane |
| CN107831784A (en) * | 2017-11-13 | 2018-03-23 | 广州纳飞智能技术有限公司 | A kind of UAV Flight Control device based on CPU sizes framework and Linux system |
| CN107888884A (en) * | 2017-11-23 | 2018-04-06 | 深圳市智璟科技有限公司 | Unmanned plane command dispatching system platform and method based on 4G networks |
| CN108391047A (en) * | 2018-01-23 | 2018-08-10 | 倪惠芳 | A kind of indoor unmanned plane multi-angled shooting control system of electronic information technical field |
| CN108447237A (en) * | 2018-03-01 | 2018-08-24 | 四川智慧鹰航空科技有限公司 | A kind of signal transmitting apparatus for unmanned plane remote control |
| CN108280981B (en) * | 2018-03-01 | 2020-12-04 | 四川智慧鹰航空科技有限公司 | Wireless control system of micro unmanned aerial vehicle |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102183955A (en) * | 2011-03-09 | 2011-09-14 | 南京航空航天大学 | Transmission line inspection system based on multi-rotor unmanned aircraft |
| CN102238367A (en) * | 2010-04-23 | 2011-11-09 | 上海东嵌科技发展有限公司 | Video monitoring system based on 3G mobile communication technology |
| CN101778260B (en) * | 2009-12-29 | 2012-01-04 | 公安部第三研究所 | Method and system for monitoring and managing videos on basis of structured description |
| CN102620736A (en) * | 2012-03-31 | 2012-08-01 | 贵州贵航无人机有限责任公司 | Navigation method for unmanned aerial vehicle |
| CN203773717U (en) * | 2013-11-12 | 2014-08-13 | 武汉大学 | Remote visual touch screen control system for unmanned plane |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2961601B1 (en) * | 2010-06-22 | 2012-07-27 | Parrot | METHOD FOR EVALUATING THE HORIZONTAL SPEED OF A DRONE, IN PARTICULAR A DRONE SUITABLE FOR AUTOPILOT STATIONARY FLIGHT |
| CN102999926B (en) * | 2012-11-12 | 2016-06-29 | 北京交通大学 | A kind of image vision significance computational methods merged based on low-level image feature |
-
2015
- 2015-09-17 CN CN201510594468.2A patent/CN105141924B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101778260B (en) * | 2009-12-29 | 2012-01-04 | 公安部第三研究所 | Method and system for monitoring and managing videos on basis of structured description |
| CN102238367A (en) * | 2010-04-23 | 2011-11-09 | 上海东嵌科技发展有限公司 | Video monitoring system based on 3G mobile communication technology |
| CN102183955A (en) * | 2011-03-09 | 2011-09-14 | 南京航空航天大学 | Transmission line inspection system based on multi-rotor unmanned aircraft |
| CN102620736A (en) * | 2012-03-31 | 2012-08-01 | 贵州贵航无人机有限责任公司 | Navigation method for unmanned aerial vehicle |
| CN203773717U (en) * | 2013-11-12 | 2014-08-13 | 武汉大学 | Remote visual touch screen control system for unmanned plane |
Non-Patent Citations (1)
| Title |
|---|
| GPS失效下的无人机组合导航***;柳明;《济南大学学报(自然科学版)》;20150331;第29卷(第2期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105141924A (en) | 2015-12-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105141924B (en) | Wireless image monitoring system based on 4G technologies | |
| CN105120237B (en) | Wireless image monitoring method based on 4G technologies | |
| US11393212B2 (en) | System for tracking and visualizing objects and a method therefor | |
| US20190286938A1 (en) | Real-to-synthetic image domain transfer | |
| CN110400363A (en) | Map constructing method and device based on laser point cloud | |
| CN106682592A (en) | Automatic image recognition system and method based on neural network method | |
| Kim et al. | Edge-network-assisted real-time object detection framework for autonomous driving | |
| CN105049790A (en) | Video monitoring system image acquisition method and apparatus | |
| CN112215074A (en) | Real-time target identification and detection tracking system and method based on unmanned aerial vehicle vision | |
| CN104486585B (en) | A kind of city magnanimity monitor video management method and system based on GIS | |
| CN108132054A (en) | For generating the method and apparatus of information | |
| CN105120232A (en) | Image monitoring and transmitting method for unmanned plane | |
| CN108764051A (en) | Image processing method, device and mobile terminal | |
| CN112419150A (en) | Random multiple image super-resolution reconstruction method based on bilateral up-sampling network | |
| CN115393635A (en) | Infrared small target detection method based on super-pixel segmentation and data enhancement | |
| CN112668675B (en) | Image processing method and device, computer equipment and storage medium | |
| CN112926415A (en) | Pedestrian avoiding system and pedestrian monitoring method | |
| Xing et al. | Building extraction from *** earth images | |
| CN109215122B (en) | Streetscape three-dimensional reconstruction system and method and intelligent trolley | |
| CN115880538A (en) | Method and equipment for domain generalization of image processing model and image processing | |
| CN115115713A (en) | Unified space-time fusion all-around aerial view perception method | |
| CN115272450A (en) | Target positioning method based on panoramic segmentation | |
| CN114639084A (en) | Road side end vehicle sensing method based on SSD (solid State disk) improved algorithm | |
| Alamouri et al. | The joint research project ANKOMMEN–Exploration using automated UAV and UGV | |
| Xie et al. | Application of intelligence binocular vision sensor: Mobility solutions for automotive perception system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |