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
- sensors
- satellite
- image
- 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
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000004891 communication Methods 0.000 claims abstract description 27
- 230000006854 communication Effects 0.000 claims abstract description 27
- 238000012544 monitoring process Methods 0.000 claims abstract description 17
- 230000006855 networking Effects 0.000 claims abstract description 4
- 230000004044 response Effects 0.000 claims description 29
- 230000006870 function Effects 0.000 claims description 23
- 230000004927 fusion Effects 0.000 claims description 15
- 230000005855 radiation Effects 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 8
- 230000000694 effects Effects 0.000 claims description 7
- 238000011524 similarity measure Methods 0.000 claims description 7
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 229910002092 carbon dioxide Inorganic materials 0.000 claims description 3
- 239000001569 carbon dioxide Substances 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 2
- 238000013461 design Methods 0.000 abstract description 4
- 238000013527 convolutional neural network Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 238000010295 mobile communication Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 4
- 238000001994 activation Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000007175 bidirectional communication Effects 0.000 description 1
- 210000003850 cellular structure Anatomy 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
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 invention relates to the technical field of marine equipment, in particular to a resident-free sea island observation method based on an Tiantong one-number satellite.
Background
Due to the development shortage of island communication and intelligent observation technology, most of the inhabitant-free islands in China are in a disordered state for a long time in utilization and protection. And the artificial activities of the island are less, so that the natural closed environment of the island is created. The geographical position, the hydrological condition and the ecological system are unique, the comprehensive value is realized in the aspects of humanity, science, ecology, politics and military and the like, the scientific and reasonable monitoring of the comprehensive value is beneficial to the development, management and protection of islands, and the comprehensive value has important research significance, and the comprehensive value is also an important technical problem and an application background which need to be solved by the project at present.
With the increasing emphasis on the consciousness of sea rights in the current countries, the sea island has become a hot spot and focus of attention of all countries in the world. Strengthening the island management and developing the island economy become important development directions in coastal countries in the future. Therefore, the construction of island communication is becoming more important. At present, submarine cables are generally used as main communication means in large islands in China. For small islands, the cost is high, the system is complex, and the construction period is long. For islands mainly monitoring marine environments and non-residential islands, communication is the most important part, the marine environment is measured in a magic way, and if data cannot be transmitted to a land monitoring center in time, the marine environment monitoring is meaningless. Therefore, it is necessary to design a solution for observing the residential island to solve the above problems.
In addition, the satellite of heaven-earth-through one is a satellite mobile communication system which is independently developed and constructed in China and is also an important component of the infrastructure of the Chinese space information. The system consists of a space section, a ground section and a user terminal, wherein the space section plan consists of a plurality of geosynchronous orbit mobile communication satellites. The system mainly ensures the medium and low speed mobile communication services such as voice, short message, fax, data, video return and the like aiming at the mobile platforms such as individuals, vehicles, airplanes, ships and the like, and can provide all-weather mobile communication service directly facing command centers and individuals of various industries. The communication system is an important component of the national spatial information infrastructure, consists of a plurality of geosynchronous orbit mobile communication satellites, provides high-gain multi-beam coverage advantage by utilizing a large deployable antenna on the satellite, has obvious advantages in the aspects of system user capacity, terminal miniaturization, platform adaptability and the like by adopting advanced technologies such as space division multiplexing, network management control of a similar cellular structure and the like, and can meet the requirements of military and civil users on satellite mobile communication in a large geographical range.
Disclosure of Invention
In order to overcome the problems, the invention aims to provide a resident-free island observation method based on the Tiantong satellite I, which can realize monitoring on a resident-free island and has accurate monitored target information.
The invention is realized by adopting the following scheme: a resident-free sea island observation method based on a satellite Tiantong I is characterized in that communication is established between the satellite Tiantong I and an island ground wireless base station, the ground wireless base station is combined with various sensors to construct a resident-free sea island wireless mesh network, different levels of observation are designed according to the area of the sea island, the island is monitored in a wireless networking mode, various sensors are tracked in a dynamic target active tracking mode based on computer vision, the sea island can be monitored in an all-around mode, and the wireless mesh network consists of a mesh client, a grid router and a gateway; the mesh client is composed of an actuator deployed on the island and various sensors, or is composed of a notebook computer and a mobile phone, the grid router forwards the flow from the gateway to the gateway of the satellite communication node, and the information tracked by the various sensors is sent to the command center through the Tiantong one satellite for monitoring.
Furthermore, the various sensors comprise an infrared camera, a sound collector and an ecological sensor; the ground wireless base station comprises a satellite communication module consisting of a radio frequency transceiver chip and a baseband chip, GTS wireless receiving equipment, GSC wireless access equipment and a control center server.
Further, the ecological sensor includes: ultrasonic wind direction sensors, ultrasonic wind speed sensors, illuminance sensors, ultraviolet radiation sensors, barometric pressure sensors, carbon dioxide sensors, total radiation sensors, and photosynthetically active radiation sensors.
Further, the specific method for tracking various sensors by adopting a dynamic target active tracking method based on computer vision comprises the following steps:
the method comprises the following steps of 1, position estimation, namely, extracting HOG characteristics in a search area by using a correlation filtering tracker and a Bayes probability model independently, taking an extracted object image as an initial template, calculating the response of the object image and the response of the object image by using two independent ridge regression problems, isomorphism, determining a target function representing the real boundary of an object, and determining the position estimation by using a linear fusion mode;
step 2, scale estimation, namely graying the obtained image by using the position determined in the step 1 as a center to obtain an image with a pixel of A, correcting the image by using a gamma correction method, dividing the image into Sudoku image blocks with different scales, constructing HOG characteristics, unifying the different image blocks into a fixed template size, extracting fhog characteristics to form a characteristic pyramid, and eliminating a boundary effect by using a Harm window;
step 3, carrying out linear fusion on the response of the correlation filtering tracker and the response of the Bayesian probability model tracker, and comprehensively estimating the position of the maximum response as the position of the target;
two independent ridge regression problem equations:
wherein h iscfFor filter trackers, using hcfTo find the maximum response value in the target search area for targeting
Tracking, θ and β are model parameters, βbayesIs a weight vector, LcfAnd LbayesAdjust θ and β as a loss function
Loss function Lcf(θ,Xt) And Lbayes(β,Xt) Minimization, XtIs the position of the target in the t-th frame, λcfAnd λbayesIs composed of
A regularization parameter;
the response fusion mode is as follows:
f(x)＝γfbayes(x)+(1-γ)fcf(x)
wherein f isbayes(x) For the response of the Bayesian probabilistic model tracker, fcf(x) For the response of the filter tracker, the fusion coefficient γ of the response is 0.2;
and 4, utilizing a selection principle of the scale:wherein, W and H are the width and height of the target in the previous frame, b is a scale factor, and T is the number of scales; taking the estimated position of the step 3 as a center to obtain a scaleUnifying image blocks with different sizes into a fixed template, extracting fhog characteristics to form a characteristic pyramid, eliminating the boundary effect by using a harm window, and correspondingly considering the image blocks as optimal scale estimation;
step 5, constructing a classifier set Q and a target image set G, wherein the Q comprises the latest M tracked target image blocks under the condition of no shielding, and the set G comprises the latest M tracked target image blocks under the condition of no shielding;
step 6, according to the target image block obtained in the step 5, calculating the similarity between the target image block and each element in the optimal target image block set G, wherein the minimum similarity measures the distanceReferred to as target similarity; calculating the similarity distance between the current target image block and the 8 surrounding image blocks, called background similarity, and setting the minimum distance asSimilarity measures include, but are not limited to, mahalanobis distance, center-to-center distance, euclidean distance, local HOG distance, best partner similarity BBS;
step 7, updating the occlusion judgment, if soThe target is occluded and the set and background similarity are not updated; if it isThen the target is not blocked and the background similarity of the current frame is calculated, and the classifier set Q and the target set G are updated at the same time, and the minimum distance of the similarity measurement between the obtained image block and the surrounding image blocks isη coefficient is 0.8;
step 8, predictionOcclusion is judged, in n +1 frame, ifThe target is completely shielded, an optimal classifier is selected from the classifier set Q by using a minimum energy function, an image block corresponding to the optimal classifier is selected from the target set G and features are extracted, the target of the n +1 frame is tracked by using the selected classifier and the features, and the step 3 is carried out; if it isIf the target is not completely shielded, tracking the next frame by using the classifiers and the characteristics updated by the n frames, and turning to the step 3;
the minimum energy function used is:
wherein,for the elements in the set Q of classifiers,for each of the energy functions of the classifiers,the likelihood estimates are characterized in that,is an entropy regularization term; wherein l ═ { l ═ l1,l2And the norm is the norm in the regularization of the neural network, and is used for parameter sparseness, so that overfitting is prevented, and the generalization capability of the model is improved.
The invention has the beneficial effects that: the invention has unique advantages in the aspects of emergency communication and maritime communication, can solve the communication problem of most of non-residential islands, and provides powerful communication guarantee for intellectualization and informatization of the non-residential islands. Meanwhile, an island internal information interaction system of the wireless mesh ad hoc network is constructed, bidirectional communication between an island and a land can be realized by adopting one set of satellite communication terminal for a single unmanned island, and the communication cost of the resident-free island is greatly reduced. The solution of the invention can be popularized to wild protection areas, unmanned areas and sensitive areas of national boundaries along with the emission of Tiantong No. 2 star and No. 3 star, and fully exerts the advantages and functions of satellite communication.
Drawings
FIG. 1 is a functional block diagram of the hardware involved in the method of the present invention.
Fig. 2 is a wireless mesh ad hoc network-based resident-free intelligent sea island observation system architecture diagram.
Fig. 3 is a distribution diagram of the distribution of the hardware of the invention on the inhabited island.
Fig. 4 is a CNN improvement scheme based on multi-layer depth feature fusion.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
Referring to fig. 1 to 4, the invention provides a residential island observation method based on an skyness one satellite, namely, designing an island information communication, monitoring and remote data acquisition system based on the skyness one satellite, which has a complete communication link, an intelligent monitoring technology and a wireless ad hoc network, integrating a low-power wide area network and a ground mobile communication system, realizing data acquisition of the residential island and efficient transmission in local and territorial areas of China, converging data on a cloud platform, and processing and distributing the data. Through integrated transmission and routing, access and control, operation, maintenance and management mechanisms, the acquired environment acquisition information, dynamic data, communication data, state data of various terminals and other information are transmitted to a command center through a satellite network and a ground network, so that diversified service information transmission and comprehensive service application such as voice, short messages, data (acquisition), video return, alarm/early warning issue and the like are realized, and monitoring, analysis and control of the monitoring station of the non-residential island are realized. The invention can be widely applied to key areas with severe environment and poor communication conditions, such as oceans, national boundaries, unmanned areas, protection areas and the like.
As shown in fig. 1, in the method for observing the non-residential island based on the satellite skyngie one, the satellite skyngie one is communicated with an island ground wireless base station, the ground wireless base station is combined with various sensors to construct a wireless mesh network of the non-residential island, different levels of observation are designed according to the island area of the island (namely different monitoring key areas are designed to form different levels of layered observation), the island is monitored in a wireless networking mode, the sensors are tracked in a dynamic target active tracking mode based on computer vision, the island can be monitored in all directions, and the wireless mesh network consists of a mesh client, a mesh router and a gateway; the mesh client is composed of an actuator deployed on the island and various sensors, or is composed of a notebook computer and a mobile phone, the grid router forwards the flow from the gateway to the gateway of the satellite communication node, and the information tracked by the various sensors is sent to the command center through the Tiantong one satellite for monitoring. The actuator executes the operation of various sensors and controls and sends the tracked information; aiming at the inhabitant-free islands with small area, a grid router is not needed, and a single base station is directly combined with various sensors to form a wireless mesh network. An island wireless Mesh network is to be established using an outdoor WL-Mesh router. The gateway may be connected to a terrestrial server through a satellite receiving station. The wireless access point communication device may act as a central transmitter and receiver of radio wave signals.
In the invention, the various sensors comprise a Canon VB-H410 infrared camera, a SongMeterSM4 sound acquisition instrument and an ecological sensor; the ground wireless base station comprises a satellite communication module consisting of an OTTO special radio frequency transceiver chip MSR01B and an MSB01A baseband chip, GTS wireless receiving equipment, GSC wireless access equipment and a control center server. The GTS wireless receiving equipment mainly completes frequency conversion, A/D conversion, filtering and baseband modulation and demodulation processing of radio frequency C frequency band signals, and the GSC wireless access equipment mainly completes protocol stack processing of the signals. The ecological sensor includes: ultrasonic wind direction sensors, ultrasonic wind speed sensors, illuminance sensors, ultraviolet radiation sensors, barometric pressure sensors, carbon dioxide sensors, total radiation sensors, and photosynthetically active radiation sensors. The inhabitant-free ecological island includes the interaction of all biological, climatic and natural resources. The distributed characteristic of the monitoring station is inherent to the geographic information, so that the design of the resident-free island intelligent monitoring system based on the wireless mesh ad hoc network needs a global reference time, which is helpful for realizing the global sequence of events. An intelligent monitoring system architecture without residents on the sea island based on the wireless mesh ad hoc network is proposed to be established, as shown in fig. 2. The CPS unit is an information physical system, realizes interaction with a physical process through a human-computer interaction interface, and controls a physical entity in a remote, reliable, real-time, safe and cooperative mode by using a networked space.
In addition, the specific mode of tracking various sensors by adopting a dynamic target active tracking mode based on computer vision comprises the following steps:
the method comprises the following steps of 1, position estimation, namely, extracting HOG characteristics in a search area by using a correlation filtering tracker and a Bayes probability model independently, taking an extracted object image as an initial template, calculating the response of the object image and the response of the object image by using two independent ridge regression problems, isomorphism, determining a target function representing the real boundary of an object, and determining the position estimation by using a linear fusion mode;
step 2, scale estimation, namely graying the obtained image by using the position determined in the step 1 as a center to obtain an image with a pixel of A, correcting the image by using a gamma correction method, dividing the image into Sudoku image blocks with different scales, constructing HOG characteristics, unifying the different image blocks into a fixed template size, extracting fhog characteristics to form a characteristic pyramid, and eliminating a boundary effect by using a Harm window;
step 3, carrying out linear fusion on the response of the correlation filtering tracker and the response of the Bayesian probability model tracker, and comprehensively estimating the position of the maximum response as the position of the target;
two independent ridge regression problem equations:
wherein h iscfFor correlation filter trackers, use is made of hcfTo find the maximum response value in the target search area for the purpose
Target tracking, θ and β model parameters, βbayesIs a weight vector, LcfAnd LbayesAdjust θ and β as a loss function
Make the loss function Lcf(θ,Xt) And Lbayes(β,Xt) Minimization, XtIs the position of the target in the t-th frame, λcfAnd λbayes
Is a regularization parameter;
the response fusion mode is as follows:
f(x)＝γfbayes(x)+(1-γ)fcf(x)
wherein f isbayes(x) For the response of the Bayesian probabilistic model tracker, fcf(x) For the response of the correlation filtering tracker, responseThe fusion coefficient gamma of (a) is 0.2;
and 4, utilizing a selection principle of the scale:wherein, W and H are the width and height of the target in the previous frame, b is a scale factor, and T is the number of scales; taking the estimated position in the step 3 as a center, obtaining image blocks with different scales, unifying the image blocks into a fixed template, extracting fhog characteristics to form a characteristic pyramid, eliminating the boundary effect by using a harm window, and correspondingly considering the image blocks as optimal scale estimation;
step 5, constructing a classifier set Q and a target image set G, wherein the Q comprises the latest M tracked target image blocks under the condition of no shielding, and the set G comprises the latest M tracked target image blocks under the condition of no shielding;
step 6, according to the target image block obtained in the step 5, calculating the similarity between the target image block and each element in the optimal target image block set G, wherein the minimum similarity measures the distance Xn minReferred to as target similarity; calculating the similarity distance between the current target image block and the 8 surrounding image blocks, called background similarity, and setting the minimum distance asSimilarity measures include, but are not limited to, mahalanobis distance, center-to-center distance, euclidean distance, local HOG distance, best partner similarity BBS;
step 7, updating the occlusion judgment, if soThe target is occluded and the set and background similarity are not updated; if it isThen the target is not blocked, the background similarity of the current frame is calculated, and the classifier set Q and the target set G are updated simultaneously to obtain an imageThe minimum distance of the similarity measure of a block to its surrounding image blocks isη coefficient is 0.8;
step 8, predicting and judging the occlusion, if the occlusion is judged in the n +1 frameThe target is completely shielded, an optimal classifier is selected from the classifier set Q by using a minimum energy function, an image block corresponding to the optimal classifier is selected from the target set G and features are extracted, the target of the n +1 frame is tracked by using the selected classifier and the features, and the step 3 is carried out; if it isIf the target is not completely shielded, tracking the next frame by using the classifiers and the characteristics updated by the n frames, and turning to the step 3;
the minimum energy function used is:
wherein,for the elements in the set Q of classifiers,for each of the energy functions of the classifiers,the likelihood estimates are characterized in that,is an entropy regularization term. Wherein l ═ { l ═ l1,l2Is a neural netNorm in the regularization is used for parameter sparsity, overfitting is prevented, and generalization capability of the model is improved. W and H are the width and height of the target in the previous frame, T is the number of scales, mentioned in step 5;
here, it is worth mentioning: the design and the construction of an experimental platform of the target active tracking technology based on vision adopt a Canon PTZ camera VB-H410 and an embedded mainboard MIO-5271 for porphyrization.
The computer vision on-line identification technology is based on the wild species financial risk identification algorithm of the improved CNN, adopts a sparse automatic coding network and a convolutional neural network technology, replaces the convolution process of a convolutional layer with a corresponding shallow layer perceptron network by improving the internal connection mode of the layer, improves the identification precision and shortens the inference time.
In order to make the features and models learned by the cloud server more representative, the nonlinearity of the network needs to be improved. The nonlinear method for improving the network is two: firstly, the network depth is increased, and secondly, the layer internal connection mode is changed. The patent will choose the second way to improve the network, i.e. replace the convolution process of the convolutional layer with the corresponding shallow perceptron network. In order to reduce parameter increase and network overfitting caused by the operation, a deep learning dropout mechanism is arranged on a neural network part, and part of hidden layer nodes are thrown away randomly during each sample input training to prevent overfitting.
The convolutional layer first performs a convolution operation to complete a series of linear activations, and then each linear input is input into a non-linear activation function, such as a ReLU activation function, tanh activation function. In convolutional layers, where the input is convolved with a series of convolution kernels, convolution is a linear operation on two signals, e.g., two functions x (t) and ω (t), and the convolution operation can be defined as:
where x represents the signal input, ω represents the convolution kernel (filter), the output h represents the feature map, t represents time, and a represents the step size of the move. Conventional neural networks have each output connected to each input, whereas convolutional neural networks have a local receptive field, which means that each output cell is connected to only a subset of the inputs, and the convolution operation is performed using the correlation between spatially local neighboring cells. Another significant characteristic of CNN is parameter sharing, which used in convolutional layers means that each location shares the same parameters (weights and offsets), reducing the number of parameters in the whole network and improving the efficiency of the computation.
After the CNN based on the structure fusion is improved, in consideration of the complexity of target recognition, in order to obtain more sufficient feature expression, feature outputs learned by each layer of the CNN are extracted, feature fusion is performed, and higher recognition accuracy is achieved, and a schematic diagram is shown in fig. 4.
Finally, the final target online identification algorithm based on the improved CNN is updated into computer vision hardware through a cloud server and a satellite communication network, so that the image acquisition equipment and the identification algorithm form a vision-based target online identification technology.
The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.
Claims (4)
1. A resident-free sea island observation method based on an Tiantong satellite I is characterized by comprising the following steps: the method comprises the steps that communication is established between an Tiantong satellite I and an island ground wireless base station, the ground wireless base station is combined with various sensors to construct a wireless mesh network without residential islands, different levels of observation are designed according to the area of the islands, the islands are monitored in a wireless networking mode, various sensors are tracked in a dynamic target active tracking mode based on computer vision, the islands can be monitored in an all-around mode, and the wireless mesh network consists of a mesh client, a mesh router and a gateway; the mesh client is composed of an actuator deployed on the island and various sensors, or is composed of a notebook computer and a mobile phone, the grid router forwards the flow from the gateway to the gateway of the satellite communication node, and the information tracked by the various sensors is sent to the command center through the Tiantong one satellite for monitoring.
2. The observation method of the resident-free island based on the Tiantong-I satellite according to claim 1, characterized in that: the various sensors comprise an infrared camera, a sound acquisition instrument and an ecological sensor; the ground wireless base station comprises a satellite communication module consisting of a radio frequency transceiver chip and a baseband chip, GTS wireless receiving equipment, GSC wireless access equipment and a control center server.
3. The observation method of the resident-free island based on the Tiantong-I satellite according to claim 2, characterized in that: the ecological sensor includes: ultrasonic wind direction sensors, ultrasonic wind speed sensors, illuminance sensors, ultraviolet radiation sensors, barometric pressure sensors, carbon dioxide sensors, total radiation sensors, and photosynthetically active radiation sensors.
4. The observation method of the resident-free island based on the Tiantong-I satellite according to claim 1, characterized in that: the specific method for tracking various sensors by adopting a dynamic target active tracking mode based on computer vision comprises the following steps:
the method comprises the following steps of 1, position estimation, namely, extracting HOG characteristics in a search area by using a filtering tracker and a Bayesian probability model independently, taking an extracted object image as an initial template, calculating the response of the object image and the response of the object image by using two independent ridge regression problems, isomorphism, determining a target function representing the real boundary of an object, and determining the position estimation by using a linear fusion mode;
step 2, scale estimation, namely graying the obtained image by using the position determined in the step 1 as a center to obtain an image with a pixel of A, correcting the image by using a gamma correction method, dividing the image into Sudoku image blocks with different scales, constructing HOG characteristics, unifying the different image blocks into a fixed template size, extracting fhog characteristics to form a characteristic pyramid, and eliminating a boundary effect by using a Harm window;
step 3, carrying out linear fusion on the response of the filtering tracker and the response of the Bayesian probability model tracker, and comprehensively estimating the position of the maximum response as the position of the target;
two independent ridge regression problem equations:
wherein h iscfFor filter trackers, using hcfTo find the maximum response value in the target search area for tracking the target, θ and β are model parameters, βbayesIs a weight vector, LcfAnd LbayesFor the loss function, θ and β are adjusted to make the loss function Lcf(θ,Xt) And Lbayes(β,Xt) Minimization, XtIs the position of the target in the t-th frame, λcfAnd λbayesIs a regularization parameter;
the response fusion mode is as follows:
f(x)＝γfbayes(x)+(1-γ)fcf(x)
wherein f isbayes(x) For the response of the Bayesian probabilistic model tracker, fcf(x) For the response of the filter tracker, the fusion coefficient γ of the response is 0.2;
and 4, utilizing a selection principle of the scale:wherein, W and H are the width and height of the target in the previous frame, b is a scale factor, and T is the number of scales; taking the estimated position in the step 3 as a center, obtaining image blocks with different scales, unifying the image blocks into a fixed template, extracting fhog characteristics to form a characteristic pyramid, eliminating the boundary effect by using a harm window, and correspondingly considering the image blocks as optimal scale estimation;
step 5, constructing a classifier set Q and a target image set G, wherein the Q comprises the latest M tracked target image blocks under the condition of no shielding, and the set G comprises the latest M tracked target image blocks under the condition of no shielding;
step 6, according to the target image block obtained in the step 5, calculating the similarity between the target image block and each element in the optimal target image block set G, wherein the minimum similarity measures the distanceReferred to as target similarity; calculating the similarity distance between the current target image block and the 8 surrounding image blocks, called background similarity, and setting the minimum distance asSimilarity measures include, but are not limited to, mahalanobis distance, center-to-center distance, euclidean distance, local HOG distance, best partner similarity BBS;
step 7, updating the occlusion judgment, if soThe target is occluded and the set and background similarity are not updated; if it isThen the target is not blocked and the background similarity of the current frame is calculated, and the classifier set Q and the target set G are updated at the same time, and the minimum distance of the similarity measurement between the obtained image block and the surrounding image blocks isη coefficient is 0.8;
step 8, predicting and judging the occlusion, if the occlusion is judged in the n +1 frameThe target is completely shielded, an optimal classifier is selected from the classifier set Q by using a minimum energy function, an image block corresponding to the optimal classifier is selected from the target set G and features are extracted, the target of the n +1 frame is tracked by using the selected classifier and the features, and the step 3 is carried out; if it isIf the target is not completely shielded, tracking the next frame by using the classifiers and the characteristics updated by the n frames, and turning to the step 3;
the minimum energy function used is:
wherein,for the elements in the set Q of classifiers,for each of the energy functions of the classifiers,the likelihood estimates are characterized in that,is an entropy regularization term; wherein l ═ { l ═ l1,l2And the norm is the norm in the regularization of the neural network, and is used for parameter sparseness, so that overfitting is prevented, and the generalization capability of the model is improved.
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 |
|---|---|---|
| Fourati et al. | Artificial intelligence for satellite communication: A review | |
| Ullah et al. | Applications of artificial intelligence and machine learning in smart cities | |
| US20200272899A1 (en) | Systems and Methods for Deploying and Updating Neural Networks at the Edge of a Network | |
| Lv | The security of Internet of drones | |
| US9060341B2 (en) | System and method for hybriding wireless location techniques | |
| US20210027112A1 (en) | Semantic State Based Sensor Tracking and Updating | |
| US20030146871A1 (en) | Wireless location using signal direction and time difference of arrival | |
| Xiao et al. | Abnormal behavior detection scheme of UAV using recurrent neural networks | |
| CN116720156A (en) | Weather element forecasting method based on graph neural network multi-mode weather data fusion | |
| Almalki et al. | Coupling multifunction drones with AI in the fight against the coronavirus pandemic | |
| Huang et al. | Space-based electromagnetic spectrum sensing and situation awareness | |
| CN111356108A (en) | Neural network-based underwater wireless sensor network anomaly diagnosis method | |
| CN110290493A (en) | Lead to the non inhabitation islands observation method of No.1 satellite based on day | |
| Wang et al. | Intelligent drone-assisted fault diagnosis for B5G-enabled space-air-ground-space networks | |
| Elawaad et al. | DeepCReg: Improving cellular-based outdoor localization using CNN-based regressors | |
| Zhang et al. | Generative AI for space-air-ground integrated networks (SAGIN) | |
| Yu et al. | Deep precipitation downscaling | |
| Kim et al. | Twenty-four-hour cloud cover calculation using a ground-based imager with machine learning | |
| CN108934028A (en) | A kind of more Bernoulli Jacob's filter distribution fusion methods | |
| CN115278698B (en) | Unmanned aerial vehicle base station dynamic deployment method and device based on dynamic user distribution prediction | |
| CN113726401B (en) | Satellite constellation reliability assessment method based on satellite survivability and link survivability | |
| Farmanifard et al. | A context-aware hybrid deep learning model for the prediction of tropical cyclone trajectories | |
| Zeng et al. | The study of DDPG based spatiotemporal dynamic deployment optimization of Air-Ground ad hoc network for disaster emergency response | |
| Penteliuc et al. | Prediction of cloud movement from satellite images using neural networks | |
| Otero et al. | Intelligent system for predicting wireless sensor network performance in on-demand deployments |
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 |