CN110290493A - Lead to the non inhabitation islands observation method of No.1 satellite based on day - Google Patents
Lead to the non inhabitation islands observation method of No.1 satellite based on day Download PDFInfo
- Publication number
- CN110290493A CN110290493A CN201910611806.7A CN201910611806A CN110290493A CN 110290493 A CN110290493 A CN 110290493A CN 201910611806 A CN201910611806 A CN 201910611806A CN 110290493 A CN110290493 A CN 110290493A
- Authority
- CN
- China
- Prior art keywords
- target
- island
- satellite
- day
- various kinds
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
-
- 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
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/14—Relay systems
- H04B7/15—Active relay systems
- H04B7/185—Space-based or airborne stations; Stations for satellite systems
- H04B7/1851—Systems using a satellite or space-based relay
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/38—Services specially adapted for particular environments, situations or purposes for collecting sensor information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W84/00—Network topologies
- H04W84/02—Hierarchically pre-organised networks, e.g. paging networks, cellular networks, WLAN [Wireless Local Area Network] or WLL [Wireless Local Loop]
- H04W84/04—Large scale networks; Deep hierarchical networks
- H04W84/06—Airborne or Satellite Networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Astronomy & Astrophysics (AREA)
- Aviation & Aerospace Engineering (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of non inhabitation islands observation methods for leading to No.1 satellite based on day, the method leads to No.1 satellite by day and terrestrial wireless base station in island is established and communicated, using the wireless mesh network of the base stations united various kinds of sensors building non inhabitation islands of terrestrial wireless, and the observation according to island island area design different levels, island are monitored in the way of wireless networking, various kinds of sensors is tracked using dynamic object active tracing mode based on computer vision, comprehensive monitoring can be carried out to island, the wireless mesh network is by mesh client, Grid Router and gateway composition；Mesh client is the actuator and various kinds of sensors composition for being deployed in island, either laptop, mobile phone, Grid Router is by flow from gateway forwards to the gateway of satellite communication node, the information that various kinds of sensors traces into is led into No.1 satellite by day and is sent to command centre to be monitored, is able to achieve non inhabitation islands monitoring.
Description
Technical field
The present invention relates to oceanographic equipment technical field, especially a kind of non inhabitation islands observation for leading to No.1 satellite based on day
Method.
Background technique
Since island communication and the development of intelligent observation technology are insufficient, the utilization of the most of non inhabitation islands in China and guarantor
Shield is chronically at disordered state.And the human activity on island is less, creates the natural enclosed environment in island.Its geographical position
Set, hydrologic condition and the ecosystem it is unique, it is right in all various aspects such as humanity, science, ecology, politics and military affairs with comprehensive value
It carries out the exploitation, management and protection that scientifically and rationally monitoring is beneficial to island, has important research significance, this is also this
The important technological problems and application background that project needs to solve at present.
As country reinforces the attention of sea power consciousness at this stage, island has become the hot spot and coke of countries in the world concern
Point.Reinforce island management, developing Island Economy, oneself becomes the important development direction of the following coastal state.Therefore, island communicates
It builds more important.Currently, Chinese large-sized island are generally with submarine cable for main means of communication.And for small-sized island and
Speech, has the shortcomings that at high cost, system complex, construction period are long.For based on marine environmental monitoring island and nothing
Resident island, communication are its most important parts, and marine environment is changeable, if data can not be transferred to land monitoring in time
Center, marine environmental monitoring will lose meaning.So need to design it is a kind of observe non inhabitation islands scheme, for solve with
Upper problem.
In addition, day leads to No.1 satellite, it is the satellite mobile communication system and China's Space letter of Chinese independent development construction
Cease the important component of infrastructure.System is made of space segment, ground segment and user terminal, and space segment plan is by more ground
Ball geo-stationary orbit moving communication satellite composition.System is to guarantee for mobile platforms such as personal and vehicle, aircraft, ships
In sound, short message, fax, data, video passback etc. based on low speed mobile communication business, it is possible to provide be directly facing in every profession and trade commander
The heart and personal, round-the-clock mobile communication service.The communication system is the important component of Chinese Space information infrastructure,
It is made of more geostationary orbit moving communication satellites, provides high-gain multi-beam using Large deployable antenna on star and covers
Lid advantage, by using space division multiplexing, the advanced technologies such as network management control of class honeycomb, in system user capacity, end
End miniaturization, the suitable dress property of platform etc. have significant advantage, can meet mobile to satellite logical within the scope of military and civilian user's the earth domain
The demand of letter.
Summary of the invention
In order to overcome the problems referred above, the object of the present invention is to provide a kind of non inhabitation islands observations for leading to No.1 satellite based on day
Method is able to achieve the monitoring to non inhabitation islands, and the target information monitored is accurate.
The present invention is realized using following scheme: a kind of non inhabitation islands observation method leading to No.1 satellite based on day, described
Method leads to No.1 satellite by day and terrestrial wireless base station in island is established and communicated, using the base stations united various kinds of sensors of terrestrial wireless
The wireless mesh network of non inhabitation islands, and the observation according to island island area design different levels are constructed, wireless group is utilized
Net mode is monitored island, various kinds of sensors using dynamic object active tracing mode based on computer vision carry out with
Track can carry out comprehensive monitoring to island, and the wireless mesh network is by mesh client, Grid Router and gateway group
At；Mesh client be deployed in island actuator and various kinds of sensors composition or laptop, mobile phone,
The information that various kinds of sensors traces into from gateway forwards to the gateway of satellite communication node, is passed through day by flow by Grid Router
Logical No.1 satellite is sent to command centre to be monitored.
Further, the various kinds of sensors includes thermal camera, sound collection instrument and ecological sensor；It is described
Terrestrial wireless base station includes the satellite communication module being made of RF transceiver chip and baseband chip, GTS radio receiver,
GSC radio reception device and control centre's server composition.
Further, the ecological sensor includes: ultrasonic wave wind transducer, ultrasonic wind speed sensor, illumination biography
Sensor, ultraviolet radiation sensor, barometric pressure sensor, carbon dioxide sensor, global radiation sensor and it is photosynthetic effectively
Radiation sensor.
Further, the various kinds of sensors using dynamic object active tracing mode based on computer vision carry out with
The concrete mode of track includes the following steps:
Step 1: location estimation, i.e., by the way that the correlation filtering tracker of the HOG feature in region of search and Bayes is general
Rate model individually extracts, and will extract obtained subject image as original template, then the response of the two is only by two
Vertical ridge regression problem carries out isomorphism after calculating, and then determines the objective function of characterization object real border, finally using linear
The mode of fusion determines location estimation；
Step 2: size estimation, i.e., using step 1 determine position centered on, the image gray processing that will acquire obtains
The image that one pixel is A, then image is corrected using gamma correction method, then be divided into the nine grids of different scale size
Image block constructs HOG feature, and different image blocks is unified into fixed form size, extracts fhog feature and forms feature gold word
Tower eliminates boundary effect using harm window；
Step 3: the response of correlation filtering tracker and the response of bayesian probability model tracker are subjected to linear fusion,
It is comprehensive to respond at maximum position, it is the location estimation of target；
Two independent ridge regression problem formulations:
Wherein hcfFor filter tracks device, h is utilizedcfMesh is carried out to find the maximum response in target search region
Target
Tracking, θ and β are model parameter, βbayesFor weight vectors, LcfWith LbayesFor loss function, adjusting θ and β makes
Loss function Lcf(θ,Xt) and Lbayes(β,Xt) minimize, XtFor position of the target in t frame, λcfWith λbayesFor
Regularization parameter；
It is as follows to respond amalgamation mode:
F (x)=γ fbayes(x)+(1-γ)fcf(x)
Wherein fbayesIt (x) is the response of bayesian probability model tracker, fcfIt (x) is the response of filter tracker, response
Fusion coefficients γ be 0.2；
Step 4: utilize the selection principle of scale:Wherein, W and H is mesh
It is marked on the width and height of former frame, b is scale factor, and T is scale quantity；Centered on the estimated location of step 3, scale is obtained not
It is unified into fixed form with the image block of size, then by these image block, fhog feature is extracted and forms feature pyramid, utilize
Harm window eliminates boundary effect, corresponding to regard that optimal scale is estimated as；
Step 5: building classifier set Q and target image set G, Q contain M tracking nearest in unobstructed situation
The target image block arrived, set G contain the nearest M in the unobstructed situation target image blocks traced into；
Step 6: the target image block obtained according to step 5 calculates itself and each element in optimal objective image block set G
Between similitude, wherein the smallest similarity measurement distanceReferred to as target similitude；Calculate current target image block with
The similarity distance of 8 image blocks around it, referred to as background similitude, if its minimum range isThe measurement side of similitude
Method include but is not limited to mahalanobis distance, center away from, Euclidean distance, part HOG distance, best partner's similitude BBS；
Step 7: shadowing is updated, ifThen target is blocked, and does not update set and background similitude；IfThen target is not blocked and calculates the background similitude of present frame, while updating classifier set Q and object set
G is closed, the minimum range of obtained image block and the similarity measurement of image block around it isη coefficient is 0.8；
Step 8: prediction shadowing, in n+1 frame, ifThen target is blocked completely, utilizes least energy
Function chooses optimum classifier from classifier set Q, and image block corresponding to optimum classifier is chosen from target collection G
And feature is extracted, using the target of selected classifier and signature tracking n+1 frame, go to step 3；IfThen target is not
It is blocked completely, the classifier and feature updated at this time using n frame carries out the tracking of next frame, goes to step 3；
Used least energy function are as follows:
Wherein,For element in classifier set Q,For each classifier
Energy function,Possibility predication is characterized,It is entropy regularization term；Wherein l={ l1,l2It is neural network
Norm in regularization, it is sparse for parameter, prevent over-fitting, the generalization ability of lift scheme.
The beneficial effects of the present invention are: the present invention to have original advantage in terms of emergency communication, overocean communications, can be with
The communication problem of most non inhabitation islands is solved, provides strong communication support for non inhabitation islands intelligence, informationization.
Meanwhile the island internal information interactive system of wireless mesh ad hoc network is constructed, realize single athermal effect using a set of satellite communication
The two-way communication on island, land can be realized in terminal, significantly reduces non inhabitation islands communications cost.Solution party of the invention
Case can be generalized to wild protection zone, depopulated zone, national boundaries sensitive area with the transmitting that day leads to No. 2 stars, No. 3 stars, give full play to
The advantage and effect of satellite communication.
Detailed description of the invention
Fig. 1 is that the method for the present invention is related to the functional block diagram of hardware.
Fig. 2 is the non inhabitation islands intelligence observation system architecture diagram the present invention is based on wireless mesh ad hoc network.
Fig. 3 is distribution map of the hardware distribution of the present invention on non inhabitation islands.
Fig. 4 is that the CNN based on the fusion of multilayer depth characteristic improves schematic diagram.
Specific embodiment
The present invention will be further described with reference to the accompanying drawing.
Shown in please referring to Fig.1 to Fig.4, the present invention provides the non inhabitation islands observation method for leading to No.1 satellite based on day,
Design it is a kind of based on day lead to No.1 satellite possess full communication link, Study On Intelligent Monitoring Techniques, wireless self-networking " isolated island " letter
Message communication, monitoring and remote data acquisition system, merge low-power consumption Wide Area Network, and land mobile communication system is realized without resident
The data on island acquire, and the high efficiency of transmission in China domestic and sea frontier zone, and data are converged in cloud platform,
Data are handled and are distributed.By integrated transmission and routing, access and control, O&M and administrative mechanism, will be obtained
The environment taken acquires the information such as the status data of information, dynamic data, communication data, various terminals and passes through satellite network and ground
Network transmission realizes the diversified industry such as speech, short message, data (acquisition), video passback, alarm/early warning publication to command centre
Information of being engaged in transmission and integrated service application, realize monitoring, analysis and the control of non inhabitation islands monitoring station.The present invention can be extensive
With the bad environments such as Yu Haiyang, national boundaries, depopulated zone, protection zone, the key area of communication condition difference.
As shown in Figure 1, a kind of non inhabitation islands observation method for leading to No.1 satellite based on day of the invention, the method is logical
It crosses day and leads to No.1 satellite and the foundation communication of island terrestrial wireless base station, nothing is constructed using the base stations united various kinds of sensors of terrestrial wireless
The wireless mesh network on resident island, and the observation according to island island area design different levels (designs different monitoring
Key area forms different grades of level observation), island are monitored in the way of wireless networking, various kinds of sensors is adopted
It is tracked with dynamic object active tracing mode based on computer vision, comprehensive monitoring, institute can be carried out to island
Wireless mesh network is stated to be made of mesh client, Grid Router and gateway；Mesh client is to be deployed in the execution on island
Device and various kinds of sensors composition or laptop, mobile phone, Grid Router is by flow from gateway forwards to satellite
The information that various kinds of sensors traces into is led to No.1 satellite by day and is sent to command centre to carry out by the gateway of communication node
Monitoring.Actuator executes the operation of various kinds of sensors, and the information traced into is carried out control transmission；For the lesser nothing of area
Resident island is not necessarily to Grid Router, directlys adopt single base station joint various kinds of sensors and constitutes wireless mesh network.It is quasi- to utilize
Outdoor WL-Mesh router establishes island wireless mesh network.Gateway can be connected to land server by Satellite Receiving Station.
Wireless access point communication equipment can be used as the central transmitter and receiver of radio wave signal.
Wherein, in the present invention, the various kinds of sensors includes Canon's VB-H410 thermal camera, SongMeterSM4
Sound collection instrument and ecological sensor；The terrestrial wireless base station includes leading to special-purpose radio-frequency transponder chip MSR01B by day
And MSB01A baseband chip forms satellite communication module, GTS radio receiver, GSC radio reception device and control centre
Server composition.GTS radio receiver mainly completes the frequency conversion of radio frequency C frequency band signals, A/D conversion, filtering and baseband modulation
Demodulation process, GSC radio reception device mainly complete the protocol stack processing of signal.The ecology sensor includes: ultrasonic wave wind
To sensor, ultrasonic wind speed sensor, illuminance transducer, ultraviolet radiation sensor, barometric pressure sensor, carbon dioxide
Sensor, global radiation sensor and light together valid radiation sensor.No resident's Eco-Island includes all biological, weathers and oneself
The interaction of right resource.The distributed nature of monitoring station is the intrinsic characteristic of geography information, this makes based on wireless mesh
The non inhabitation islands intelligent monitor system design of ad hoc network needs a global reference time, this will be helpful to the complete of realization event
Office's sequence.The proposed Liru non inhabitation islands intelligent monitor system framework shown in Fig. 2 based on wireless mesh ad hoc network.CPS is mono-
Member is information physical system, is realized by man-machine interactive interface and physics process interaction, using networking space with it is long-range, can
Lean on, in real time, safety, the mode of cooperation manipulate a physical entity.
In addition, what the various kinds of sensors was tracked using dynamic object active tracing mode based on computer vision
Concrete mode includes the following steps:
Step 1: location estimation, i.e., by the way that the correlation filtering tracker of the HOG feature in region of search and Bayes is general
Rate model individually extracts, and will extract obtained subject image as original template, then the response of the two is only by two
Vertical ridge regression problem carries out isomorphism after calculating, and then determines the objective function of characterization object real border, finally using linear
The mode of fusion determines location estimation；
Step 2: size estimation, i.e., using step 1 determine position centered on, the image gray processing that will acquire obtains
The image that one pixel is A, then image is corrected using gamma correction method, then be divided into the nine grids of different scale size
Image block constructs HOG feature, and different image blocks is unified into fixed form size, extracts fhog feature and forms feature gold word
Tower eliminates boundary effect using harm window；
Step 3: the response of correlation filtering tracker and the response of bayesian probability model tracker are subjected to linear fusion,
It is comprehensive to respond at maximum position, it is the location estimation of target；
Two independent ridge regression problem formulations:
Wherein hcfFor correlation filter tracker, h is utilizedcfCome find the maximum response in target search region come into
Row mesh
Target tracking, θ and β are model parameter, βbayesFor weight vectors, LcfWith LbayesFor loss function, θ and β is adjusted
Make loss function Lcf(θ,Xt) and Lbayes(β,Xt) minimize, XtFor position of the target in t frame, λcfWith λbayes
For regularization parameter；
It is as follows to respond amalgamation mode:
F (x)=γ fbayes(x)+(1-γ)fcf(x)
Wherein fbayesIt (x) is the response of bayesian probability model tracker, fcfIt (x) is the response of correlation filtering tracker,
The fusion coefficients γ of response is 0.2；
Step 4: utilize the selection principle of scale:Wherein, W and H is mesh
It is marked on the width and height of former frame, b is scale factor, and T is scale quantity；Centered on the estimated location of step 3, scale is obtained not
It is unified into fixed form with the image block of size, then by these image block, fhog feature is extracted and forms feature pyramid, utilize
Harm window eliminates boundary effect, corresponding to regard that optimal scale is estimated as；
Step 5: building classifier set Q and target image set G, Q contain M tracking nearest in unobstructed situation
The target image block arrived, set G contain the nearest M in the unobstructed situation target image blocks traced into；
Step 6: the target image block obtained according to step 5 calculates itself and each element in optimal objective image block set G
Between similitude, wherein the smallest similarity measurement distance Xn minReferred to as target similitude；Calculate current target image block with
The similarity distance of 8 image blocks around it, referred to as background similitude, if its minimum range isThe measurement side of similitude
Method include but is not limited to mahalanobis distance, center away from, Euclidean distance, part HOG distance, best partner's similitude BBS；
Step 7: shadowing is updated, ifThen target is blocked, and does not update set and background similitude；IfThen target is not blocked and calculates the background similitude of present frame, while updating classifier set Q and object set
G is closed, the minimum range of obtained image block and the similarity measurement of image block around it isη coefficient is 0.8；
Step 8: prediction shadowing, in n+1 frame, ifThen target is blocked completely, utilizes least energy
Function chooses optimum classifier from classifier set Q, and image block corresponding to optimum classifier is chosen from target collection G
And feature is extracted, using the target of selected classifier and signature tracking n+1 frame, go to step 3；IfThen target is not
It is blocked completely, the classifier and feature updated at this time using n frame carries out the tracking of next frame, goes to step 3；
Used least energy function are as follows:
Wherein,For element in classifier set Q,For each classifier
Energy function,Possibility predication is characterized,It is entropy regularization term.Wherein l={ l1,l2It is neural network
Norm in regularization, it is sparse for parameter, prevent over-fitting, the generalization ability of lift scheme.W and H is target in former frame
Width and height, T be scale quantity, mention in steps of 5；
Here it is worth noting that: the design of the target active tracking technology experiment porch of view-based access control model with build, use
Canon Pan/Tilt/Zoom camera VB-H410 grinds magnificent embedded main board MIO-5271.
The on-line real-time test of computer vision, based on improve CNN wildlife species property insurance recognizer, using it is sparse from
Dynamic coding network, convolutional neural networks technology, by improving layer inner connecting way, by the convolution process of convolutional layer with accordingly
Shallow-layer perceptron network replaces, and improves accuracy of identification, shortens inference time.
In order to which the feature and model that learn Cloud Server are more representative, need to improve the non-linear of network.And it mentions
The nonlinear method of high network has two: first is that increasing network depth, second is that changing layer inner connecting way.This patent will select
Two kinds of modes improve network, i.e., replace the convolution process of convolutional layer with corresponding shallow-layer perceptron network.In order to reduce this behaviour
Make the increase of bring parameter and network over-fitting, enters deep learning dropout mechanism in part of neural network, sample is defeated each time
When entering trained, part hidden layer node is thrown away at random, prevents over-fitting.
Convolution operation is first carried out to complete a series of linear activation in convolutional layer, is then each linearly input to non-linear
In activation primitive, such as ReLU activation primitive, tanh activation primitive.In convolutional layer, input is carried out with a series of convolution kernel
Convolution study, convolution is a kind of linear operation on two kinds of signals, such as: there are two function x (t) and ω (t), convolution behaviour
Making can be with is defined as:
Wherein, x indicates signal input, and ω indicates convolution kernel (filter), and output h indicates Feature Mapping, and t indicates time, a
Indicate mobile step-length.The each output of traditional neural network connects each input, and convolutional neural networks possess local sense
By open country, it means that each output unit is connected solely to a subset of input, utilizes the phase between the adjacent cells of space part
Closing property carries out convolution algorithm.Another significant properties of CNN is parameter sharing, and parameter sharing used in convolutional layer means often
Identical parameter (weight and deviation) is shared in a position, reduces the number of parameters of whole network, improves the efficiency of calculating.
After the CNN based on structure fusion is improved, it is contemplated that the complexity of target identification, in order to obtain more fully feature
The feature output that every layer of CNN learns is extracted, carries out Fusion Features, realize higher accuracy of identification, schematic diagram by expression
As shown in Figure 4.
Finally, we pass through cloud server, satellite communication based on the target online recognition algorithm for improving CNN for final
Network is updated into computer vision hardware, so that image capture device is online with the target of composition view-based access control model with recognizer
Identification technology.
The foregoing is merely presently preferred embodiments of the present invention, all equivalent changes done according to scope of the present invention patent with
Modification, is all covered by the present invention.
Claims (4)
1. a kind of non inhabitation islands observation method for leading to No.1 satellite based on day, it is characterised in that: the method passes through day logical one
Number satellite and island terrestrial wireless base station, which are established, to be communicated, and constructs non inhabitation islands using the base stations united various kinds of sensors of terrestrial wireless
Wireless mesh network, and according to island island area design different levels observation, in the way of wireless networking to island into
Row monitoring, various kinds of sensors are tracked using dynamic object active tracing mode based on computer vision, can be to island
Comprehensive monitoring is carried out, the wireless mesh network is made of mesh client, Grid Router and gateway；Mesh client
Be be deployed in island actuator and various kinds of sensors composition or laptop, mobile phone, Grid Router will flow
Amount from gateway forwards to the gateway of satellite communication node, lead to No.1 satellite by day and send by the information that various kinds of sensors is traced into
To command centre to be monitored.
2. the non inhabitation islands observation method according to claim 1 for leading to No.1 satellite based on day, it is characterised in that: described
Various kinds of sensors includes thermal camera, sound collection instrument and ecological sensor；The terrestrial wireless base station includes by radio frequency
The satellite communication module of transponder chip and baseband chip composition, GTS radio receiver, GSC radio reception device and control
Central server composition processed.
3. the non inhabitation islands observation method according to claim 2 for leading to No.1 satellite based on day, it is characterised in that: described
Ecological sensor include: ultrasonic wave wind transducer, ultrasonic wind speed sensor, illuminance transducer, ultraviolet radiation sensor,
Barometric pressure sensor, carbon dioxide sensor, global radiation sensor and light together valid radiation sensor.
4. the non inhabitation islands observation method according to claim 1 for leading to No.1 satellite based on day, it is characterised in that: described
Various kinds of sensors includes as follows using the concrete mode that dynamic object active tracing mode based on computer vision is tracked
Step:
Step 1: location estimation, i.e., by by region of search HOG feature filter tracker and bayesian probability model list
It solely extracts, obtained subject image will be extracted and returned as original template, then by the response of the two by two independent ridges
Isomorphism is carried out after returning problem to calculate, then determines the objective function of characterization object real border, finally utilizes the side of linear fusion
Formula determines location estimation；
Step 2: size estimation, i.e., using step 1 determine position centered on, the image gray processing that will acquire obtains one
Pixel is the image of A, then is corrected to image using gamma correction method, then be divided into the nine grids figure picture of different scale size
Block constructs HOG feature, and different image blocks is unified into fixed form size, extracts fhog feature and forms feature pyramid, benefit
Boundary effect is eliminated with harm window；
Step 3: the response of filter tracker and the response of bayesian probability model tracker are subjected to linear fusion, comprehensive response
It is the location estimation of target at maximum position；
Two independent ridge regression problem formulations:
Wherein hcfFor filter tracks device, h is utilizedcfCome find the maximum response in target search region carry out target with
Track, θ and β are model parameter, βbayesFor weight vectors, LcfWith LbayesFor loss function, adjusting θ and β makes loss function Lcf(θ,
Xt) and Lbayes(β,Xt) minimize, XtFor position of the target in t frame, λcfWith λbayesFor regularization parameter；
It is as follows to respond amalgamation mode:
F (x)=γ fbayes(x)+(1-γ)fcf(x)
Wherein fbayesIt (x) is the response of bayesian probability model tracker, fcfIt (x) is the response of filter tracker, response is melted
Closing coefficient gamma is 0.2；
Step 4: utilize the selection principle of scale:Wherein, W and H is that target exists
The width and height of former frame, b are scale factor, and T is scale quantity；Centered on the estimated location of step 3, it is different big to obtain scale
Small image block, then these image block is unified into fixed form, it extracts fhog feature and forms feature pyramid, utilize harm
Window eliminates boundary effect, corresponding to regard that optimal scale is estimated as；
Step 5: building classifier set Q and target image set G, Q contain M nearest in unobstructed situation and trace into
Target image block, set G contain the nearest M in the unobstructed situation target image blocks traced into；
Step 6: the target image block obtained according to step 5 calculates it between each element in optimal objective image block set G
Similitude, wherein the smallest similarity measurement distanceReferred to as target similitude；Calculate current target image block and its week
The similarity distance of 8 image blocks, referred to as background similitude are enclosed, if its minimum range isThe measure packet of similitude
Include but be not limited to mahalanobis distance, center away from, Euclidean distance, part HOG distance, best partner's similitude BBS；
Step 7: shadowing is updated, ifThen target is blocked, and does not update set and background similitude；IfThen target is not blocked and calculates the background similitude of present frame, while updating classifier set Q and object set
G is closed, the minimum range of obtained image block and the similarity measurement of image block around it isη coefficient is 0.8；
Step 8: prediction shadowing, in n+1 frame, ifThen target is blocked completely, utilizes least energy function
Optimum classifier is chosen from classifier set Q, and is chosen image block corresponding to optimum classifier from target collection G and mentioned
Feature is taken, using the target of selected classifier and signature tracking n+1 frame, goes to step 3；IfThen target not by
It blocks completely, the classifier and feature updated at this time using n frame carries out the tracking of next frame, goes to step 3；
Used least energy function are as follows:
Wherein,For element in classifier set Q,For each classifier energy
Function,Possibility predication is characterized,It is entropy regularization term；Wherein l={ l1,l2It is neural network canonical
Norm in change, it is sparse for parameter, prevent over-fitting, the generalization ability of lift scheme.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910611806.7A CN110290493A (en) | 2019-07-08 | 2019-07-08 | Lead to the non inhabitation islands observation method of No.1 satellite based on day |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910611806.7A CN110290493A (en) | 2019-07-08 | 2019-07-08 | Lead to the non inhabitation islands observation method of No.1 satellite based on day |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110290493A true CN110290493A (en) | 2019-09-27 |
Family
ID=68020916
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910611806.7A Pending CN110290493A (en) | 2019-07-08 | 2019-07-08 | Lead to the non inhabitation islands observation method of No.1 satellite based on day |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110290493A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110958591A (en) * | 2019-12-20 | 2020-04-03 | 中国船舶工业系统工程研究院 | Marine cross-domain communication management and control system of wide area ocean thing networking |
| CN115002709A (en) * | 2022-05-19 | 2022-09-02 | 四创科技有限公司 | Water conservancy observation method based on Tiantong number one |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106982243A (en) * | 2016-12-01 | 2017-07-25 | 天津春荣合陞科技发展有限公司 | Internet of Things greenhouse based on Wireless MESH sensor network technique |
| CN107480704A (en) * | 2017-07-24 | 2017-12-15 | 南开大学 | It is a kind of that there is the real-time vision method for tracking target for blocking perception mechanism |
| CN107846243A (en) * | 2017-09-28 | 2018-03-27 | 成都傅立叶电子科技有限公司 | A kind of remote data acquisition system for leading to No.1 moving communication satellite based on day |
-
2019
- 2019-07-08 CN CN201910611806.7A patent/CN110290493A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106982243A (en) * | 2016-12-01 | 2017-07-25 | 天津春荣合陞科技发展有限公司 | Internet of Things greenhouse based on Wireless MESH sensor network technique |
| CN107480704A (en) * | 2017-07-24 | 2017-12-15 | 南开大学 | It is a kind of that there is the real-time vision method for tracking target for blocking perception mechanism |
| CN107846243A (en) * | 2017-09-28 | 2018-03-27 | 成都傅立叶电子科技有限公司 | A kind of remote data acquisition system for leading to No.1 moving communication satellite based on day |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110958591A (en) * | 2019-12-20 | 2020-04-03 | 中国船舶工业系统工程研究院 | Marine cross-domain communication management and control system of wide area ocean thing networking |
| CN115002709A (en) * | 2022-05-19 | 2022-09-02 | 四创科技有限公司 | Water conservancy observation method based on Tiantong number one |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3699825A2 (en) | Systems and methods for deploying and updating neural networks at the edge of a network | |
| Lu et al. | CONet: A cognitive ocean network | |
| US11232330B2 (en) | Adaptive neural network selection to extract particular results | |
| US20140097979A1 (en) | Device & method for cognitive radar information network | |
| CN104049639B (en) | A kind of unmanned boat antisurge based on support vector regression controls apparatus and method | |
| CN206087337U (en) | Anti - unmanned aerial vehicle system of interference formula | |
| CN110290493A (en) | Lead to the non inhabitation islands observation method of No.1 satellite based on day | |
| CN107450076A (en) | The collaboration networking measuring method and device of cloud detection radar | |
| Ajith et al. | Deep learning algorithms for very short term solar irradiance forecasting: A survey | |
| CN117114207B (en) | Marine personnel drift track prediction method | |
| CN107423670A (en) | MODIS mist monitoring methods based on depth confidence network | |
| CN116720156A (en) | Weather element forecasting method based on graph neural network multi-mode weather data fusion | |
| Sai et al. | A comprehensive survey on artificial intelligence for unmanned aerial vehicles | |
| Lou et al. | Wave height prediction suitable for maritime transportation based on green ocean of things | |
| Zhang et al. | Generative AI for space-air-ground integrated networks (SAGIN) | |
| Hassan et al. | Internet of Things-Enabled Vessel Monitoring System for Enhanced Maritime Safety and Tracking at Sea | |
| CN116562466B (en) | Planning method for high orbit satellite observation task | |
| Gricius et al. | Advanced approach of multiagent based buoy communication | |
| Xu et al. | Collaborative Human Recognition with Lightweight Models in Drone-based Search and Rescue Operations | |
| Otero et al. | Intelligent system for predicting wireless sensor network performance in on-demand deployments | |
| Ahmad | Smart remote sensing network for disaster management: an overview | |
| Cao et al. | Research on image recognition of Wushu action based on remote sensing image and embedded system | |
| Tarekegn et al. | Channel Quality Estimation in 3D Drone Base Station for Future Wireless Network | |
| KR20220066517A (en) | System and method for providing sailing information of vessel, computer readable medium storing the same | |
| Zhang et al. | Small-scale location identification in natural environments with deep learning based on biomimetic sonar echoes |
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: 20190927 |