CN104809205B - A kind of online network of waterways space-time accident detection method - Google Patents
A kind of online network of waterways space-time accident detection method Download PDFInfo
- Publication number
- CN104809205B CN104809205B CN201510206245.4A CN201510206245A CN104809205B CN 104809205 B CN104809205 B CN 104809205B CN 201510206245 A CN201510206245 A CN 201510206245A CN 104809205 B CN104809205 B CN 104809205B
- Authority
- CN
- China
- Prior art keywords
- node
- time
- sensor
- network
- space
- 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
Abstract
The invention discloses a kind of online network of waterways space-time accident detection method, comprise the following steps：First, backbone node is selected using connected dominating set algorithm, only the state of detection backbone node in each timestamp Sensor Network；It is then detected that whether backbone node occurs space-time anomalous event；If the child nodes of backbone node are added in the Candidate Set of detection by backbone node there occurs time anomaly；The abnormal backbone node of time of origin is compared by the expecting state that Bayesian Network Learning obtains and observer state, sees whether deviate from the spatial relationship that study obtains, in the event of free air anomaly, space-time anomalous counts device adds 1.In next timestamp, expand the node in detection range, including backbone node and Candidate Set, continue to detect whether that free air anomaly or time anomaly occurs.Finally, when the number of backbone node and its Candidate Set node space-time exception reaches threshold value Q, just think space-time anomalous event occurs.
Description
Technical field
The present invention relates to wireless sense network detection technique field, and in particular to a kind of wireless to the network of waterways using scattered method
The method that sensing network carries out space-time accident detection.
Background technology
With sensor technology fast development and wireless communication technology it is increasingly mature, wireless sensor network is extensive
Be applied to ecological science observation, environmental monitoring, the field such as disaster early warning and national defense and military.In these practical applications, from
The accident detection function of dynamicization is essential.As a rule, in order to making phase for Network Abnormal as soon as possible
The remedial measure answered is, it is necessary to all events impacted to system application of detection burst in real time, including physical phenomenon, pollution
Thing monitoring etc., then makes early warning.Sensor deployment monitors material different in water such as chlorine in real time in the network of waterways
Concentration, the too high levels of free chlorine are the most important reasons of water pollution in water, so selection monitoring free chloro concentration judges
The generation of anomalous event.Sensor can produce substantial amounts of data flow, it is necessary in real time analyte sensors monitor value, to detect exception
Event, the anomalous event of network of waterways water environment pollution can be caused by being particularly those.
Recent years, the research to accident detection have much been paid close attention in radio sensing network field.Tradition
Radio sensing network accident detection method can be roughly divided into three classes：Detection method based on threshold value, based on pattern
Detection method and the reasoning detection method based on study.
Above-mentioned a few class methods are all solely limited to the reasoning and study of time or spatial domain, and time and spaces union reasoning
Referred to seldom with study.And in real world, sensor is all by certain space relational layout, and sensing data is then by a timing
Between the cycle gather, therefore, sensing data has shown strong temporal correlation, and such time-space relationship is to event detection
Accuracy is most important.A kind of effective accident detection method, not only detection time is abnormal for this method, and detects space
It is abnormal, the propagation path of trace exception event, it is that accident detection will solve the problems, such as, and to be solved by this invention ask
Topic.
The content of the invention
Goal of the invention：A kind of method of online network of waterways space-time accident detection, solution can not integrate in the prior art examines
The problem of considering the temporal correlation of network exception event.In order to consider time and space factor, select probability graph model
Simulated time and spatial data respectively, using the Markov Chain discovery time event based on learning method, use Bayesian network
Speculate the probability of occurrence of spatial event, consider time and free air anomaly event, determine whether there is the generation of space-time anomalous event.
And the node important by the use of connected dominating set algorithms selection replaces all nodes to monitor on-line as key contact.
Technical scheme：A kind of online network of waterways space-time accident detection method, including in terms of following four：
(1) connected dominating set algorithms selection backbone node is utilized, monitors time and the spatiality of backbone node, generation on-line
For the whole nodes of monitoring；
(2) Markov Chain simulated time process is used, the backbone node selected according to step (1) individually passes through Ma Erke
Husband's chain, utilize transition probability predicted events development tendency between state.When by Markov Chain, if the bone of prediction
Dry junction sensor stateThe sequence τ and sensor states X (s of actual measurement for some timei, t+ τ) it is inconsistent when, recognize
For there occurs time anomaly；
(3) spatial distribution of network of waterways sensor is represented with Bayesian network, and each sensor is regarded as Bayesian network
One node, the flow direction in river regard the direction on Bayesian network side as.Sensor network can be regarded as a directed acyclic
Bayesian network, with Bayesian network speculate spatial event probability of occurrence judge whether occur free air anomaly；
(4) see whether having time anomalous event occurs backbone node in the network of waterways according to Markov state conversion, Ran Hougen
Whether there is space anomalous event according to Bayesian network forecasting backbone node, finally determine whether there is space-time anomalous event hair
Raw, time and spatial coherence model are coordinated mutual to detect space-time anomalous event；
A kind of method of online network of waterways space-time accident detection of present disclosure, the content (1) utilize connected component
Backbone node is selected with set algorithm, monitors time and the spatiality of backbone node on-line, instead of monitoring the specific of whole nodes
Step is as follows：
(1.1) labeling process：During initialization, all nodes are all non-dominant nodes；Each node u with its communication radius
Interior RuNeighbours exchange the information of respective neighbours, now node u is according to neighbor information, judge whether two can not direct phase
The neighbor node v and w of mutual communication, if in the presence of node u dominates node as candidate and participates in structure CDS, otherwise shows node u
Can be by other node controls, not as CDS nodes.
(1.2) prune rule：The candidate's domination number of nodes obtained by labeling process is excessive, and candidate, which dominates node, to be needed
According to prune rule, the quantity for dominating set of node is reduced.If the K neighbor node that candidate dominates node u can pass through other
Dominate node to be communicated, then the set of node that deletion of node u is obtained from CDS is still CDS, therefore dominates set of node from candidate
Middle deletion of node u, node u turn into non-CDS nodes.
(1.3) node in connected dominating set is as key contact, it is only necessary to which the data flow for monitoring backbone node replaces prison
Survey all nodes.This strategy can guarantee that non-selected node can find at least one key section from its neighbor node
Point, when anomalous event occurs in non-selected non-key near nodal, key contact can be traveled to as soon as possible, can be examined rapidly
Abnormal generation is measured, reduces loss.
A kind of method of online network of waterways space-time accident detection of present disclosure, the content (2) use Markov
Chain simulated time process, the traffic diagram of wireless network is covered, all sensor nodes individually pass through Markov Chain, utilize
Transition probability predicted events development tendency between state.When by Markov Chain, if the backbone node of prediction passes
Sensor stateThe sequence τ and sensor states X (s of actual measurement for some timei, t+ τ) it is inconsistent when, it is believed that there occurs
Time anomaly.Time anomaly detection comprises the following steps that：
(2.1) Markov Chain is a limited or denumerable sequence of events (E on discrete time point1,E2...)
Set, at any time, following behavior of process are dependent only on current state, unrelated with state before.Use Ma Erke
Husband's chain simulated time process, the backbone node selected according to step (1) individually pass through Markov Chain, turn using between state
Move the state that probabilistic forecasting event occurs.
(2.2) free chloro concentration is quantized into three state in the Sensor monitoring network of waterways, and Low (low concentration), Medium is (dense
Spend moderate), the sensor stream of High (high concentration) monitorings individually passes through Markov Chain, defines current state first, then
Predict the state of next step.Assuming that sensor is in moment StState be Ni, moment St+1State be Nj, according to Markov Chain
State transition probability Pij=P (St+1=Nj|St=Ni)=P (Nj|Ni) predicted events occur state and its development and change become
Gesture.
(2.3) sensor deviates normal time behavior and can be regarded as noise or be used as a time anomaly event.By
Frequently occur in deviateing, so rejecting noise this possibility, therefore, the deviation continuously occurred can regard time anomaly as
Event, when by Markov Chain, if the backbone node sensor states of predictionSequence τ for some time
With the sensor states X (s of actual measurementi, t+ τ) it is inconsistent when, it is believed that there occurs time anomaly.
A kind of method of online network of waterways space-time accident detection of present disclosure, content (3) network of waterways sensor
Spatial distribution represent that each sensor is regarded as a node of Bayesian network, the flow direction in river is seen with Bayesian network
Make the direction on Bayesian network side.Sensor network can be regarded as the Bayesian network of a directed acyclic, use Bayes
The probability of occurrence of network guessing spatial event judges whether that free air anomaly occurs.Free air anomaly detection comprises the following steps that：
(3.1) spatial relationship of analog sensor node and its neighbor node is carried out for network of waterways structure Bayesian network G, every
Individual sensor regards a node of Bayesian network as, and the flow direction in river regards the direction on Bayesian network side as, it is possible to
Sensor network regards the Bayesian network of a directed acyclic as.
(3.2) training data of Bayesian network is by S tuplesComposition,Represent sensing
DeviceState.It can be concentrated from historical data and produce many training datas, then can learned with maximum likelihood estimation algorithm
Parameter of the acquistion to Bayesian network.
(3.3) after the training stage, to each sensor, a conditional probability table can be obtained, it is pre- according to conditional probability table
Survey the state of sensor.Predicted state and true observer state are contrasted, if sensor SiObservation be not belonging to that state,
Think that actual value deviate from the spatial relationship that study obtains, there is free air anomaly.
A kind of method of online network of waterways space-time accident detection of present disclosure, the content (4) see bone in the network of waterways
Whether having time anomalous event occurs dry node, then sees whether backbone node has free air anomaly in the Bayesian network of structure again
Event occurs, and is finally determined with the generation of space-time anomalous event, time and spatial coherence model and coordinates mutual to detect space-time
Anomalous event, comprise the following steps that：
(4.1) backbone node selected according to step (1), see backbone node whether see extremely by having time according to step (2)
Whether having time anomalous event occurs backbone node.If backbone node is there occurs time anomaly, by the bone that time of origin is abnormal
The child nodes of dry node are as Candidate Set.If abnormal without time of origin, continue to detect the next timestamp of backbone node
State.
(4.2) node of the abnormal backbone node of time of origin and its Candidate Set is passed through by Bayesian network according to step (3)
Network study obtain expecting state compared with observer state, see whether sensor SiIt deviate from the spatial relationship that study obtains.
(4.3) if backbone node SE and SI (SE is sensor node predicted value, and SI is observation) is different, space occurs
Abnormal, due to time anomaly, the value of space-time anomalous counts device adds 1.
(4.4) if the node in Candidate Set has been detected free air anomaly, then the child nodes of the both candidate nodes
It is added to Candidate Set.Detect simultaneously and candidate's centralized node of free air anomaly occurs whether having time is abnormal, if so, space-time is abnormal
Counter adds 1, similarly, continues to detect whether candidate's centralized node has free air anomaly, if so, detection time is abnormal, during generation
Between it is abnormal, space-time anomalous counts device adds 1.When the number of backbone node and its Candidate Set space-time exception reaches threshold value Q, just recognize
For space-time anomalous event occurs.
The present invention uses above-mentioned technical proposal, has the advantages that：
1. the monitoring of selection backbone node rather than all nodes of monitoring Sensor Network improve detection efficiency and shorten the response time.
2. existing method but generally considers time and spatial data property point, the characteristic pair of anomalous event can not be combined
Global network makes unified space-time detection.The space-time accident detection algorithm energy binding time and the aspect of space two of the present invention
Factor considers, improves the accuracy of accident detection, reduces rate of false alarm.
3. can not only detect anomalous event, pollution spread can be terminated as early as possible with the propagation path of trace exception event,
Reduce loss.
Brief description of the drawings
Fig. 1 is the overall framework figure of the inventive method embodiment；
Fig. 2 is the flow chart of space-time accident detection method of the embodiment of the present invention；
Fig. 3 is actual network of waterways basin figure；
Fig. 4 is the sensor distribution map for being deployed in the network of waterways；
Fig. 5 represents the Markov transfer process of free chloro concentration；
Fig. 6 is the node state in certain moment Sensor Network；
The State Transferring of Fig. 7 free chloro concentrations at different moments.
Embodiment
With reference to specific embodiment, the present invention is furture elucidated, it should be understood that these embodiments are merely to illustrate the present invention
Rather than limitation the scope of the present invention, after the present invention has been read, various equivalences of the those skilled in the art to the present invention
The modification of form falls within the application appended claims limited range.
The present invention provides a kind of embodiment of online network of waterways space-time accident detection algorithm, and Fig. 1 is online network of waterways space-time
The overall framework figure of accident detection.As can be seen that the key step of the embodiment of the present invention is as follows：
(1) connected dominating set algorithms selection backbone node is utilized, monitors time and the spatiality of backbone node, generation on-line
For the whole nodes of monitoring；
(2) Markov Chain simulated time process is used, the backbone node selected according to step (1) individually passes through Ma Erke
Husband's chain, utilize transition probability predicted events development tendency between state.When by Markov Chain, if the bone of prediction
Dry junction sensor stateThe sequence τ and sensor states X (s of actual measurement for some timei, t+ τ) it is inconsistent when, recognize
For there occurs time anomaly；
(3) spatial distribution of network of waterways sensor is represented with Bayesian network, and each sensor is regarded as Bayesian network
One node, the flow direction in river regard the direction on Bayesian network side as.Sensor network can be regarded as a directed acyclic
Bayesian network, with Bayesian network speculate spatial event probability of occurrence judge whether occur free air anomaly；
(4) see whether having time anomalous event occurs backbone node in the network of waterways according to Markov state conversion, Ran Hougen
Whether there is space anomalous event according to Bayesian network forecasting backbone node, finally determine whether there is space-time anomalous event hair
Raw, time and spatial coherence model are coordinated mutual to detect space-time anomalous event；
Wherein step (1) concretely comprises the following steps：
(1.1) labeling process：During initialization, all nodes are all non-dominant nodes；Each node u with its communication radius
Interior RuNeighbours exchange the information of respective neighbours, now node u is according to neighbor information, judge whether two can not direct phase
The neighbor node v and w of mutual communication, if in the presence of node u dominates node as candidate and participates in structure CDS, otherwise shows node u
Can be by other node controls, not as CDS nodes.
(1.2) prune rule：The candidate's domination number of nodes obtained by labeling process is excessive, and candidate, which dominates node, to be needed
According to prune rule, the quantity for dominating set of node is reduced.If the K neighbor node that candidate dominates node u can pass through other
Dominate node to be communicated, then the set of node that deletion of node u is obtained from CDS is still CDS, therefore dominates set of node from candidate
Middle deletion of node u, node u turn into non-CDS nodes.
(1.3) node in connected dominating set is as key contact, it is only necessary to which the data flow for monitoring backbone node replaces prison
Survey all nodes.This strategy can guarantee that non-selected node can find at least one key section from its neighbor node
Point, when anomalous event occurs in non-selected non-key near nodal, key contact can be traveled to as soon as possible, can be examined rapidly
Abnormal generation is measured, reduces loss.
Wherein step (2) concretely comprise the following steps：
(2.1) Markov Chain is a limited or denumerable sequence of events (E on discrete time point1,E2...)
Set, at any time, following behavior of process are dependent only on current state, unrelated with state before.Use Ma Erke
Husband's chain simulated time process, the traffic diagram of wireless network is covered, all sensor nodes individually pass through Markov Chain, profit
The state occurred with transition probability predicted events between state.
(2.2) as shown in figure 5, being quantized into three state by free chloro concentration in the network of waterways of Sensor monitoring, Low is (low dense
Degree), each sensor stream of Medium (moderate concentration), High (high concentration) individually passes through markoff process, and definition first is worked as
Preceding state, then predict the state of next step.Assuming that sensor is in moment StState be Ni, in moment St+1State be
Nj, according to markovian state transition probability Pij=P (St+1=Nj|St=Ni)=P (Nj|Ni) predicted events occur shape
State and its development tendency.
(2.3) sensor deviates normal time behavior and can be regarded as noise or be used as a time anomaly event.By
Frequently occur in deviateing, so rejecting noise this possibility, therefore, the deviation continuously occurred can regard time anomaly as
Event, when by Markov Chain, if the backbone node sensor states of predictionSequence τ for some time
With the sensor states X (s of actual measurementi, t+ τ) it is inconsistent when, it is believed that there occurs time anomaly.
Wherein step (3) concretely comprise the following steps：
(3.1) as shown in figure 4, carrying out the sky of analog sensor node and its neighbor node for network of waterways structure Bayesian network G
Between relation, each sensor is regarded as a node of Bayesian network, the flow direction in river regards the direction on Bayesian network side as,
So sensor network can be regarded as the Bayesian network of a directed acyclic.
(3.2) training data of Bayesian network is by S tuplesComposition,Represent sensing
DeviceState.It can be concentrated from historical data and produce many training datas, then can learned with maximum likelihood estimation algorithm
Parameter of the acquistion to Bayesian network.
(3.3) after the training stage, to each sensor, a conditional probability table can be obtained, it is pre- according to conditional probability table
Survey the state of sensor.Predicted state and true observer state are contrasted, if sensor SiObservation be not belonging to that state,
Think that actual value deviate from the spatial relationship that study obtains, there is free air anomaly.
Fig. 2 is space-time accident detection algorithm flow chart, it can be seen that the idiographic flow of step (4) is：
(4.1) backbone node selected according to step (1), see backbone node whether see extremely by having time according to step (2)
Whether having time anomalous event occurs backbone node.If backbone node is there occurs time anomaly, by the bone that time of origin is abnormal
The child nodes of dry node are as Candidate Set.If abnormal without time of origin, continue to detect the next timestamp of backbone node
State.
(4.2) node of the abnormal backbone node of time of origin and its Candidate Set is passed through by Bayesian network according to step (3)
The expecting state that network study obtains compares with observer state, sees sensor SiWhether spatial relationship that study obtain deviate from.
(4.3) if backbone node SE and SI (SE is sensor node predicted value, and SI is observation) is different, space occurs
Abnormal, due to time anomaly, the value of space-time anomalous counts device adds 1.
(4.4) if the node in Candidate Set has been detected free air anomaly, then the child nodes of the both candidate nodes
It is added to Candidate Set.Detect simultaneously and candidate's centralized node of free air anomaly occurs whether having time is abnormal, if so, space-time is abnormal
Counter adds 1, similarly, continues to detect whether candidate's centralized node has free air anomaly, if so, detection time is abnormal, during generation
Between it is abnormal, space-time anomalous counts device adds 1.When the number of backbone node and its Candidate Set space-time exception reaches threshold value Q, just recognize
For space-time anomalous event occurs.
Fig. 3 represents network of waterways sensor network distribution map, figure interior joint S for basin figure, Fig. 4 of the network of waterwaysiTo be deployed in river course
In sensor, (Si,Sj) represent to connect the side of sensor, the direction on side is the flow direction in river.The spatial distribution of network of waterways sensor
Represented with Bayesian network, each sensor is regarded as a node of Bayesian network, the flow direction in river regards Bayesian network as
The direction on network side.Sensor network can be regarded as the Bayesian network of a directed acyclic.Bayesian Network Learning parameter
Training data by S tuplesComposition.RepresentState.It can be concentrated from historical data
Produce from timestamp 1 to many training datas of sequence tThen learnt with maximum likelihood estimation algorithm
To the parameter of Bayesian network.
Fig. 5 represents the Markov transfer process of free chloro concentration, is quantized into 3 by the free chloro concentration of Sensor monitoring
Kind state, Low (low concentration), Medium (moderate concentration), High (high concentration), Fig. 5 represent that free chlorine is low, in, high three state
Between transfer process.
Fig. 6 is the state of certain moment Sensor Network interior joint free chloro concentration, there is low, tri- kinds of different shapes of medium, high
State, Fig. 7 represent the State Transferring of free chloro concentration at different moments, represent in interval of time N1~N5, free chloro concentration
State Transferring.Each sensor stream individually passes through markoff process, defines current state first, then predicts next step
State.Assuming that sensor is in moment StState be Ni, in moment St+1State be Nj, according to markovian State Transferring
Probability Pij=P (St+1=Nj|St=Ni)=P (Nj|Ni) predicted events occur state and its development tendency.Sensor is inclined
Noise (due to sensor failure) is can be regarded as from normal time behavior or is used as a time anomaly event.Due to
Deviation frequently occurs, so rejecting noise this possibility, therefore, the deviation continuously occurred can regard time anomaly thing as
Part.When by Markov Chain, if the backbone node sensor states of predictionSequence τ and reality for some time
Sensor states X (the s of surveyi, t+ τ) it is inconsistent when, it is believed that there occurs time anomaly.
Claims (5)
- A kind of 1. online network of waterways space-time accident detection method, it is characterised in that comprise the following steps：(1) connected dominating set algorithms selection backbone node is utilized, monitors time and the spatiality of backbone node on-line, instead of prison Survey whole nodes；(2) Markov Chain simulated time process is used, the backbone node selected according to step (1) individually passes through Markov Chain, utilize transition probability predicted events development tendency between state；When by Markov Chain, if the backbone of prediction Junction sensor stateThe sequence τ and sensor states X (s of actual measurement for some timei, t+ τ) it is inconsistent when, it is believed that There occurs time anomaly；(3) spatial distribution of network of waterways sensor is represented with Bayesian network, and each sensor is regarded as one of Bayesian network Node, the flow direction in river regard the direction on Bayesian network side as；Sensor network can be regarded as the shellfish of a directed acyclic This network of leaf, speculate that the probability of occurrence of spatial event judges whether that free air anomaly occurs with Bayesian network；(4) see whether having time anomalous event occurs backbone node in the network of waterways according to Markov state conversion, then according to shellfish Whether this neural network forecast backbone node of leaf has space anomalous event, finally determines whether there is the generation of space-time anomalous event, when Between and spatial coherence model will coordinate mutually to detect space-time anomalous event.
- 2. online network of waterways space-time accident detection method according to claim 1, it is characterised in that the step (1) Concretely comprise the following steps：(1.1) labeling process：During initialization, all nodes are all non-dominant nodes；The each node u and R in its communication radiusu Neighbours exchange the information of respective neighbours, now node u is according to neighbor information, judge whether two can not directly phase intercommunication The neighbor node v and w of letter, if dominating node as candidate in the presence of, node u participates in structure CDS, otherwise show that node u can be with By other node controls, not as CDS nodes；(1.2) prune rule：The candidate's domination number of nodes obtained by labeling process is excessive, and candidate, which dominates node, needs basis Prune rule, reduce the quantity for dominating set of node；If the K neighbor node that candidate dominates node u can be by other dominations Node is communicated, then the set of node that deletion of node u is obtained from CDS is still CDS, therefore dominates in set of node and delete from candidate Except node u, node u turns into non-CDS nodes；(1.3) node in connected dominating set is as key contact, it is only necessary to which the data flow for monitoring backbone node replaces monitoring institute There is node；This strategy can guarantee that non-selected node can find at least one backbone node from its neighbor node, when Anomalous event occurs in non-selected non-key near nodal, can travel to key contact as soon as possible, can quickly detect Abnormal generation, reduce loss.
- 3. online network of waterways space-time accident detection method according to claim 1, it is characterised in that the step (2) With Markov Chain simulated time process, by comparison prediction value and measured value, judge whether time of origin exception；Detailed process It is as follows：(2.1) Markov Chain is a limited or denumerable sequence of events (E on discrete time point1,E2...) set, At any time, following behavior of process is dependent only on current state, unrelated with state before；With Markov Chain mould Pseudotime process, the backbone node selected according to step (1) individually pass through Markov Chain, utilize transition probability between state The state that predicted events occur；(2.2) free chloro concentration is quantized into the sensor stream of three state and individually passes through Markov in the Sensor monitoring network of waterways Chain, current state is defined first, then predict the state of next step；(2.3) sensor deviates normal time behavior and can be regarded as noise or be used as a time anomaly event；Due to inclined From frequently occurring, so rejecting noise this possibility, therefore, the deviation continuously occurred can regard time anomaly thing as Part, when by Markov Chain, if the backbone node sensor states of predictionFor some time sequence τ with Sensor states X (the s of actual measurementi, t+ τ) it is inconsistent when, it is believed that there occurs time anomaly.
- 4. online network of waterways space-time accident detection method according to claim 1, it is characterised in that the step (3) It is as follows with the spatial coherence between Bayesian network analog sensor, detailed process：(3.1) carry out the spatial relationship of analog sensor node and its neighbor node for network of waterways structure Bayesian network G, passed each Sensor regards a node of Bayesian network as, and the flow direction in river regards the direction on Bayesian network side as, it is possible to sensing Device network regards the Bayesian network of a directed acyclic as；(3.2) training data of Bayesian network is by S tuplesComposition,Representative sensor State；It can be concentrated from historical data and produce training data, then can learn to obtain shellfish with maximum likelihood estimation algorithm The parameter of this network of leaf；(3.3) after the training stage, to each sensor, a conditional probability table can be obtained, is predicted and passed according to conditional probability table The state of sensor；Predicted state and true observer state are contrasted, if sensor SiObservation be not belonging to that state, it is believed that Actual value deviate from the spatial relationship that study obtains, and have free air anomaly.
- 5. online network of waterways space-time accident detection method according to claim 1, it is characterised in that the step (4) Time and spatial coherence model will be coordinated mutually to detect space-time anomalous event, and detailed process is as follows：(4.1) backbone node selected according to step (1), see backbone node whether send out by having time anomalous event according to step (2) It is raw；If backbone node is there occurs time anomaly, using the child nodes of the abnormal backbone node of time of origin as Candidate Set；Such as Fruit does not have time of origin abnormal, continues to detect the next timestamp state of backbone node；(4.2) node of the abnormal backbone node of time of origin and its Candidate Set is passed through by Bayesian network according to step (3) Practise obtain expecting state compared with observer state, see whether sensor SiIt deviate from the spatial relationship that study obtains；(4.3) if backbone node SE is different from SI, free air anomaly occurs, due to time anomaly, the value of space-time anomalous counts device Add 1, wherein SE is sensor node predicted value, and SI is observation；(4.4) if the node in Candidate Set has been detected free air anomaly, then the child nodes of both candidate nodes are also added to Candidate Set；Detect simultaneously and candidate's centralized node of free air anomaly occurs whether having time is abnormal, if so, space-time anomalous counts device Add 1, similarly, continue to detect whether candidate's centralized node has free air anomaly, if so, detection time is abnormal, time of origin is abnormal, Space-time anomalous counts device adds 1；When the number of backbone node and its Candidate Set space-time exception reaches threshold value Q, just think to occur Space-time anomalous event.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510206245.4A CN104809205B (en) | 2015-04-27 | 2015-04-27 | A kind of online network of waterways space-time accident detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510206245.4A CN104809205B (en) | 2015-04-27 | 2015-04-27 | A kind of online network of waterways space-time accident detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104809205A CN104809205A (en) | 2015-07-29 |
| CN104809205B true CN104809205B (en) | 2018-03-20 |
Family
ID=53694027
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510206245.4A Active CN104809205B (en) | 2015-04-27 | 2015-04-27 | A kind of online network of waterways space-time accident detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104809205B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11354184B2 (en) * | 2019-06-21 | 2022-06-07 | Palo Alto Research Center Incorporated | Method and system for performing automated root cause analysis of anomaly events in high-dimensional sensor data |
| CN110969556B (en) * | 2019-09-30 | 2023-11-21 | 上海仪电（集团）有限公司中央研究院 | Machine learning multidimensional multi-model fusion river channel water quality anomaly detection method and device |
| CN111858712A (en) * | 2020-07-20 | 2020-10-30 | 上海仪电（集团）有限公司中央研究院 | In-situ water quality inspection data time-space analysis and anomaly detection method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102201065A (en) * | 2011-05-16 | 2011-09-28 | 天津大学 | Method for detecting monitored video abnormal event based on trace analysis |
| CN102752784A (en) * | 2012-06-19 | 2012-10-24 | 电子科技大学 | Detection method of distribution type event domain based on graph theory in wireless sensor network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8478048B2 (en) * | 2010-07-08 | 2013-07-02 | International Business Machines Corporation | Optimization of human activity determination from video |
-
2015
- 2015-04-27 CN CN201510206245.4A patent/CN104809205B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102201065A (en) * | 2011-05-16 | 2011-09-28 | 天津大学 | Method for detecting monitored video abnormal event based on trace analysis |
| CN102752784A (en) * | 2012-06-19 | 2012-10-24 | 电子科技大学 | Detection method of distribution type event domain based on graph theory in wireless sensor network |
Non-Patent Citations (3)
| Title |
|---|
| "Bayesian Analysis of Elapsed Times in Continuous-Time Markov Chains";Marco A. R. Ferreira .etc;《Canadian Journal of Statistics》;20101231;全文 * |
| "Empirical Bayes Estimation for Markov Chain Models of Drought Events in Peninsular Malaysia";Wahidah Sanusi .etc;《AIP Conference Proceedings》;20141231;全文 * |
| "基于贝叶斯推理的水环境系统参数识别";朱嵩，等;《江苏大学学报（自然科学版）》;20070531;第28卷(第3期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104809205A (en) | 2015-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | A survey on fault diagnosis in wireless sensor networks | |
| CN106872657B (en) | A kind of multivariable water quality parameter time series data accident detection method | |
| CN105764162A (en) | Wireless sensor network abnormal event detecting method based on multi-attribute correlation | |
| CN103533571B (en) | Fault-tolerant event detecting method based on temporal voting strategy | |
| CN105636094A (en) | Wireless sensor network early warning method and system based on clustering compressed sensing | |
| CN102752784B (en) | Detection method of distribution type event domain based on graph theory in wireless sensor network | |
| CN103813355A (en) | Identification method for anomalous points of cooperative synchronization in distributed network | |
| Ayadi et al. | Spatio-temporal correlations for damages identification and localization in water pipeline systems based on WSNs | |
| CN104809205B (en) | A kind of online network of waterways space-time accident detection method | |
| Zhang et al. | Urban traffic flow forecast based on FastGCRNN | |
| CN114265913A (en) | Space-time prediction algorithm based on federal learning on industrial Internet of things edge equipment | |
| Krishnan et al. | An improved clustering with particle swarm optimization-based mobile sink for wireless sensor networks | |
| Dai et al. | Temporal multi-view graph convolutional networks for citywide traffic volume inference | |
| Wang et al. | Correlated time series self-supervised representation learning via spatiotemporal bootstrapping | |
| Qiu et al. | Evolution of node behavior in link prediction | |
| Zhang et al. | A spatiotemporal graph wavelet neural network for traffic flow prediction | |
| Syed et al. | A novel adaptive neuro-fuzzy inference system-differential evolution (Anfis-DE) assisted software fault-tolerance methodology in wireless sensor network (WSN) | |
| Abraham et al. | Computational Intelligence in Wireless Sensor Networks: Recent Advances and Future Challenges | |
| Zhang et al. | Fault diagnosis in clustering WSN based on neighbor cooperation | |
| Mao et al. | Real-time event detection with water sensor networks using a spatio-temporal model | |
| Naït-Abdesselam et al. | Data Quality Improvements for Internet of Things Using Artificial Neural Networks | |
| Mao et al. | Online event detection based on the spatio-temporal analysis in the river sensor networks | |
| Zhou et al. | Towards learning in grey spatiotemporal systems: A prophet to non-consecutive spatiotemporal dynamics | |
| Liu et al. | Research on ship abnormal behavior detection method based on graph neural network | |
| Chavan et al. | Efficient bio-inspired method for disaster monitoring in wireless sensor networks using improved PSO |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Wang Longbao Inventor after: Mao Yingchi Inventor after: Jie Qing Inventor after: Jia Bicong Inventor after: Ping Ping Inventor after: Xu Feng Inventor after: Zhou Xiaofeng Inventor before: Mao Yingchi Inventor before: Jie Qing Inventor before: Jia Bicong Inventor before: Ping Ping Inventor before: Wang Longbao Inventor before: Xu Feng Inventor before: Zhou Xiaofeng |