CN101868045B - Moving target classification identification method based on compound sensor Ad Hoc network - Google Patents
Moving target classification identification method based on compound sensor Ad Hoc network Download PDFInfo
- Publication number
- CN101868045B CN101868045B CN2009101850784A CN200910185078A CN101868045B CN 101868045 B CN101868045 B CN 101868045B CN 2009101850784 A CN2009101850784 A CN 2009101850784A CN 200910185078 A CN200910185078 A CN 200910185078A CN 101868045 B CN101868045 B CN 101868045B
- Authority
- CN
- China
- Prior art keywords
- target
- signal
- network node
- personnel
- signals
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a moving target classification identification method based on a compound sensor Ad Hoc network. The target classification identification function is completed through the collection on the magnetic resistance and slight shock signals actually generated by ground moving targets. The common targets are classified into six types such as bare-handed personnel and the like according to the compound features of the magnetic resistance and the slight shock signals of the targets. On a deployed sensor network system, firstly, target sensing voltage threshold values of two sensor nodes are determined, then, feature signals generated by the moving targets are converted into three-value signal modes, i.e. {+1, 0, -1}, and a sample base of the target classification identification is established and constructed in advance in the test. When a collection system collects two kinds of actual signals of the moving targets, the actual signals are compared to a standard feature signal in the sample base, the distance among the actual signals and the standard feature signal is calculated, and a target type corresponding to the signal with the shortest distance is used as a classification identification result. The method is characterized in that the calculation amount is small, the real-time performance is strong, and the classification correctness is high, and the invention is used for the identification problem of various kinds of moving targets.
Description
Technical field
The invention belongs to the applied business aspect of self-organizing sensor network network system, relate to the application technology of target identification and MANET, relate in particular to a kind of moving target classification identification method that adopts the compound sensor Ad Hoc network system.
Background technology
Sensor network is research both at home and abroad in recent years and uses very popular field, on the development of the national economy and national defense and military, has important application.The progress of microelectric technique, computing technique and wireless communication technology has promoted the fast development of low-power consumption Multifunction Sensor, multiple functions such as making in small volume can the integrated information collection, data processing and radio communication.Sensor network is normally formed by being deployed in cheap microsensor nodes a large amount of in the monitored area; Through a multi-hop of communication formation, the network system of self-organizing; Purpose is the information of perceptive object in cooperation ground perception, collection and the processing network's coverage area, and sends to the observer.
Sensor network has very wide application prospect at numerous areas such as military and national defense, industrial or agricultural control, city management, biologic medical, environmental monitoring, rescue and relief work, anti-terrorism, deathtrap Long-distance Control.Transducer, perceptive object and observer constitute three key elements of sensor network.Sensor network merges information world in logic with objective physical world, changing human and natural interactive mode, through the direct perception of sensor network objective world, has expanded the function of existing network and the ability in the human knowledge world.
In the process of self-organizing, ubiquitousization, plyability development, the application terminal is the key of network interworking and fusion all the time at wireless network, and the terminal traffic technology provides colourful application function for network.Function from MANET; Each sensor node all has the dual-use function of information gathering and route; Except carrying out that local information is collected and handling, also store, manage and merge the data that other node forwards, accomplish some particular tasks with other node cooperation.If communication environment or other factors change; When causing certain or part of nodes of sensor network to lose efficacy; Before then reselected route automatically, guaranteed when network breaks down, can realize automatic healing that this was the characteristic feature of MANET by other node of their transmission data.
Surveying with discerning moving target is one of terminal traffic function of sensor network.Target classification identification is exactly the characteristic signal through analysis and processing target, target is classified as a certain type of prior delimitation.Basic way is normally confirmed certain decision rule on the basis of sample training collection, make by this rule being identified the error recognition rate minimum that object is classified and caused.The formal style sorting technique has smaller calculation than intelligent optimization methods such as neural net, genetic algorithms, is convenient to the quick execution of computer program, exports the result in real time.
The target identification method of current use mainly comprises based on modes such as image, induction coil, microwave, sound.Object recognition rate based on image can reach 90%, but based on the limitation of the target identification of image, and it is bigger to be that system performance receives the influence of environment and light.The installation cost and the maintenance cost of some target identification scheme are higher, are unfavorable for using on a large scale.Target identification scheme based on sensor network provides node arrangement very flexibly, system can portability with help large scale deployment.
Method provided by the invention belongs to typical formal style sorting technique, has the advantages that amount of calculation is little, energy consumption is low, real-time.This target classification identifying schemes adopts magnetic resistance and microseismic activity transducer to form the complex probe terminal, and single one-sidedness of planting the sensor probe Information Monitoring is avoided in cooperation ground perception and identification moving target.Through the result of MANET mode, carry out transferring to base station control center in the network with Classification and Identification.
In recent years along with the improving constantly of microsensor certainty of measurement, through extracting the characteristic quantity of target data, attribute that can evaluating objects.The moving target recognition technology has the important use value and significance, such as it is the important information source of traffic programme and administrative department, and the information of target type provide correct assessment can for the highway capacity predict.Militarily this is to scout a kind of important means of obtaining information, is used to discern the moving target type on the key road segment, is the detection and the automatic new approach that provides of discerning of battlefield target.Its advantage be to remedy the volume that traditional reconnaissance equipment exists big, be prone to promptly the battlefield perception information is provided for outpost detachment by defectives such as enemy's discoveries.
Summary of the invention
The purpose of this invention is to provide a kind of ground moving object classifying identification method based on combined type microsensor network system, characteristics are that amount of calculation is little, and real-time is good, and the target type of classification is many.
The invention is characterized in: magnetic resistance and microseismic activity signal through ground moving object is produced are gathered, and in conjunction with the communication function of MANET system, accomplish the target classification identification mission.
Fig. 1 is based on the organization chart of the moving object classification recognition system of compound sensor Ad Hoc network platform.Shown in this figure, whole system is made up of data acquisition assembly, self-organized network communication assembly, target classification recognizer component, power management assembly, back-stage management assembly.
The hardware development of sensor network nodes comprises probe node and the design that receives two circuit boards of Control Node (being the gateway aggregation node).Probe node is used for the data that perception target and preliminary treatment collect; Receive Control Node and be responsible for transmitting control commands, receive the data that probe node sends, and storage shows monitoring result.
By anisotropic magnetoresistive and microseismic activity sensor probe composition data acquisition component, the multi-source signal that it sends according to measured target, the event information that the detection of a target occurs.The self-organized network communication assembly is accomplished the message transmission task of multi-hop MANET.Target classification identification is the computing function of carrying out target classification.The power management assembly provides powered battery.The back-stage management assembly is responsible for accepting and passing on user instruction, starts the business service function of network.
Magnetoresistive transducer is a kind of microsensor of measuring the magnetic field of the earth.The bandwidth range of this type transducer is 1-5MHz, and response speed is fast, and the minimum vehicle that can guarantee travel speed 400km/h at the stroke up-sampling of every 0.1mm once satisfies the needs that moving target is surveyed fully.
For the complex probe sensing system, node circuit not only comprises magnetoresistive transducer but also comprised the microseismic activity transducer.The latter adopts the ADXL202 transducer, and it is the double-axel acceleration sensor of a kind of low cost, low-power consumption.ADXL202 exports digital signal, and pulse duty cycle is directly proportional with two acceleration that sensing shaft is experienced separately, and the output signal period is regulated with outer meeting resistance in the 0.5ms-10ms scope.ADXL202 measures positive negative acceleration, and maximum measuring range is ± 2g.In the sensor circuit of complex probe system, the accekeration that calculates according to duty ratio and resistor values, and get two on axially acceleration and numerical value as final useful signal data.
Twin shaft magnetoresistive transducer and twin shaft microseismic activity transducer are integrated on a terminal node, and it is consistent that requirement and road direction and vertical direction thereof are placed in the position.Sample frequency is set to 20Hz, i.e. two kinds of sensor signal data of per second collection respectively are 20 times.
Native system realizes distinguishing six type games targets: 1. free-hand personnel; 2. armed personnel (or the personnel that carry metallic object); 3. dilly; 4. passenger vehicle; 5. oversize vehicle; 6. caterpillar.Free-hand personnel are meant the Walking People that do not carry metallic object, and the armed personnel is meant the personnel that carry metal, gun.The vehicle that gross mass is not less than 4500kg is an oversize vehicle, otherwise belongs to dilly, and length is not less than the passenger vehicle that is of 6m in addition.Caterpillar is moved by the pedrail mechanism support.
Accompanying drawing 2 is FB(flow block)s of performing step.Sensor network system is after the deployment of detection location finishes, and network terminal node is implemented the target classification recognition function through to the magnetoresistance signal of the actual generation of ground moving object and the collection of microseismic activity signal.
The operating procedure of sorting technique is following:
Step 1: threshold design
The end-probing node of sensor network system need design different sensing voltage threshold values, and threshold value is to draw through test statistics in advance, and the selection of threshold method is following:
Wherein, Fiducial value be driftlessness through and glitch-free situation under, the output circuit voltage of network node; Undulating value be driftlessness through and under the noisy situation, the output circuit voltage of network node, counting at interval is the number of two undulating value data between the fiducial value.The sample calculation of threshold value is referring to embodiment.
Step 2: peak value conversion
The peak value transfer process is the actual signal that produces when moving target is passed through each network node, converts three binarization modes to: and+1,0 ,-1}.The rate of change and the threshold value of the actual signal that moving target is produced when each network node compare, if for just and greater than threshold value, then be characterized as+1; If for bearing and, then being characterized as-1 quantitatively greater than threshold value; If the numerical value of slope is less than or equal to threshold value, then be characterized as 0.
The magnetic resistance and the microseismic activity signal of peak valley modal representation target have the characteristic of " mountain peak " and " cheuch ".The peak value transfer process has high compression ratio, thereby makes the amount of calculation of target classification process little, real-time.
Step 3: sample storehouse structure
Gather the magnetoresistance signal of moving target generation and the sample of microseismic activity signal in advance through test, utilize the peak value transfer process to convert these sample signals to peak-mode respectively, form two class standard characteristic signals; With this sample storehouse of constructing target signature, realize the differentiation of six class targets: 1. free-hand personnel, 2. armed personnel (or the personnel that carry metallic object); 3. dilly; 4. passenger vehicle, 5. oversize vehicle, 6. caterpillar.
For twin shaft magnetic resistance and microseismic activity transducer, the stack features example signal in the target classification sample storehouse of foundation is as shown in the table:
Step 4: Characteristic Recognition
When sensor network system is carried out the target monitoring task; As long as moving target is through near each network node; The data acquisition assembly of this network node just carries out collection in worksite to magnetic resistance and these two kinds of signals of microseismic activity that moving target produces; Two class standard characteristic signals in two types of signal results of gathering and the sample storehouse are compared respectively, calculate the distance between them respectively, sue for peace apart from addition for two types; Which kind of target type minimum relatively again apart from summation, then with the target type of minimum range summation as recognition result.
Described Characteristic Recognition process details is operated as follows:
When adopting the standard feature signal to carry out the identification of target type, on-the-spot actual signal that produces of moving target and the standard feature signal in the sample storehouse are compared calculating:
R＝|c
1-c
b1|+|c
2-c
b2|+…+|c
10-c
b10|
Wherein, { c
1, c
2C
10Be three binarization mode data of the on-the-spot actual signal that produces of moving target of network node collection, { c
B1, c
B2C
B10Be the data of the standard feature signal in the sample storehouse, R is the vectorial difference between these two kinds of data.
When the data acquisition assembly of each network node collects magnetoresistance signal and the microseismic activity signal of target, calculate difference R.Suppose R
1The difference comparative result of expression magnetoresistance signal, R
2The difference comparative result of expression microseismic activity signal, the difference of these two kinds of signals is summed to (R apart from addition
1+ R
2).If the COMPREHENSIVE CALCULATING result (R that certain target type is corresponding
1+ R
2) numerical value minimum, then with the target type of its pairing standard feature signal result as Classification and Identification.Monitoring center's computer of sensor network system is responsible for collecting the situation of target classification identification, and is saved in database.
This target classification identification method is convenient to the real time execution of program code very much, and is low in energy consumption, at each sensor terminal node independent operation, reduced the transfer of data communication cost of sensor network with distributed way.
This method makes an experiment through the wireless sensor network hardware platform; Real data according to magnetic resistance of gathering and microseismic activity signal is accomplished target classification, the target classification recognition result the when computer of monitoring center shows the output movement target through the sensor network node.
Below in conjunction with the description of drawings specific embodiment.
Description of drawings
Fig. 1 is based on the organization chart of the moving object classification recognition system of compound sensor Ad Hoc network;
Fig. 2 is the FB(flow block) of native system performing step;
Fig. 3 is the principle process of target classification identification;
Fig. 4 is an example of peak value transfer process;
Fig. 5 is a sample example of armed personnel's (or the personnel that carry metallic object) magnetoresistance signal;
Fig. 6 is a sample example of armed personnel's (or the personnel that carry metallic object) microseismic activity signal.
Embodiment
Be described in detail with reference to the attached drawings some embodiment of the present invention.
The network terminal probe node of system comprises two kinds of transducers, microprocessor, radio communication and four modules of power supply, adopts the lithium ion battery power supply of sheet, is assemblied in the special circular duroplasts ghost.Whole probe node volume is little, and the probe node after the assembling has only the diameter of 10.3cm and the thickness of 3cm, thereby lays flexibly, and the inductance coil that can replace using is at present discerned moving target.
Accompanying drawing 3 is principle processes of target classification identification, and this PRS based on statistical method and formal style is partly formed by 5: sensor information is obtained, preliminary treatment, feature extraction and selection, classifier design, classified calculating process.
Consider to distinguish six class targets: (1) free-hand personnel; (2) armed personnel (or the personnel that carry metallic object); (3) dilly; (4) passenger vehicle; (5) oversize vehicle; (6) caterpillar.The standard of classification is the magnetic resistance that has according to moving target and the characteristic of microseismic activity signal.
The standard of target classification is following: free-hand personnel are the Walking People that do not carry metallic object, like the pedestrian of pad it, or the common people on the battlefield.The armed personnel is the personnel that carry gun and individual equipment, or carries the pedestrian of a certain amount of metallic object.
Because the current vehicle kind is many, can carry out different classification according to vehicle different specifications, structure, fuel, purposes and model.Here with reference to China GB/T 3730.1-1988 standard " term of automobile and semitrailer and definition-type of vehicle ", common vehicle is divided into oversize vehicle, dilly and passenger vehicle.Gross mass is not less than the oversize vehicle that is of 4500kg, like lorry, otherwise belongs to dilly, like car.In addition, length is not less than the passenger vehicle that is of 6m, like bus.Caterpillar is by pedrail mechanism support operation, and crawler belt is the flexible chain link round driving wheel, bogie wheel, inducer and carrier wheel that is driven by driving wheel, and it is made up of creeper tread and track pin etc.Caterpillar and wheeled vehicle are having significantly differently aspect the microseismic activity signal of ground, this class targets comprises crawler mounted excavator, panzer etc.
The sensing equipment that is used to survey and discern moving target at present is relatively heavier mostly, expensive, and the workload of system deployment and laying is big, like inductance coil measuring system, video camera head system, radar system and sonar detection system.
Native system adopts the microminiaturized Weir of Buddhist nun suddenly twin shaft magnetoresistive transducer and twin shaft microseismic activity acceleration transducer (ADXL) image data, effectively overcomes the problems referred to above that transducer type selecting aspect exists.When target was passed through the terminal node of sensor network, transducer detected various dipole moments, the especially vehicle target of target different parts, and the changes of magnetic field amount discloses the concrete magnetic resistance characteristic of target, utilized the magnetic resistance disturbing signal to distinguish different target types.The microseismic activity transducer is through to the detection of ground microseismic activity signal, and the target of telling process is personnel, wheeled vehicle or caterpillar, through with the analysis-by-synthesis of magnetoresistance signal, identify the particular type of target.
According to magnetoresistance signal and microseismic activity signal that moving target produces, development mixes the Circuits System of surveying.Individual packages is in independent task module respectively for the collection of two kinds of transducers and processing, and each task is independently programmed, parallel running.Because the sample frequency of system is lower, gets and makes 20Hz, it can satisfy the demand of identification mission.When target through out-of-date, two kinds of transducers can collect the signal that target produces reliably with this sample frequency.
The acquisition tasks module of two kinds of transducers is responsible for controlling sensor acquisition, then the data that collect is passed to data processing task, and in data processing task, initial data is through pretreatment operation and obtain final information.This information sends to the computer of Surveillance center with the Network Transmission form of single-hop or multi-hop.
For the acquisition tasks that on a circuit board, realizes two transducers; Verification experimental verification the real-time of whole compound sensor data acquisition assembly; Under the situation that target is passed through fast; When promptly passing through the end-probing node of sensor network system in 1-2 time second, system can return one group of valid data, is about 50-70 magnetoresistance signal and microseismic activity signal.Classification and identification algorithm is accomplished classified calculating based on these data, and multisensor complex probe system shows good real time performance.
After sensor network system was disposed, according to the signal that compound sensor is gathered, the step of target classification identification was following:
(1) threshold design
The selection of threshold method of design is following:
Wherein, Fiducial value be driftlessness through and glitch-free situation under, the output circuit voltage of network node; Undulating value be driftlessness through and under the noisy situation, the output circuit voltage of network node; Counting at interval is the number of two undulating value data between the fiducial value, generally is not less than 3.The sensor node that test shows good stability is got to count at interval and is 3-5, and the sensor node of less stable (for example between in use under the long partially situation) is got the interval and counted more than 5.
For example; Following table is the magnetic resistance output voltage signal data (mv) that magnetoresistive transducer collects in a period of time, and fiducial value is 64, and undulating value is 63; Here the undulating value 63mv between fiducial value 64mv has only two; Generally be not less than 3 principle according to counting at interval, thereby the interval here counts to get and do 3, calculating threshold value at last is 0.3mv.
| Output signal (mv) | 64 | 64 | 64 | 64 | 63 | 63 | 64 | 64 | 64 |
| Sampling instant | 15:26:4 8 | ?15:26:4 8 | ?15:26:4 9 | ?15:26:4 9 | ?15:26:4 9 | ?15:26:4 9 | ?15:26:4 9 | ?15:26:4 9 | ?15:26:4 9 |
(2) peak value conversion
The peak value transfer process is the signal that target is produced when the transducer, convert to three binarization modes+1,0 ,-1}, the process of peak value conversion is following:
According to pre-set threshold with echo signal convert into three values+1,0 ,-1} pattern:
Wherein, c (k) is for the rate of change of sampled data is a slope, when slope calculations the spaced points number less than 3 be designated as 0, promptly undulating value just occurs once in a while, the duration is shorter, thinks due to the interference signal; Δ k is to count in the interval in sampling time; Δ a (k) is the target output signal voltage changing value in counting at interval; C (k) is the ternary (digital) signal that the peak value conversion forms; C
ThresholdBe threshold value.
The example of a transfer process is shown in accompanying drawing 4.If the amplitude and the duration of spike or trough reach preset threshold, the value of spike or trough is considered to effectively.
The vectorial dimension unification that among the embodiment peak value conversion is formed confirms as 10, the peak signal of for example changing for+1 ,-1,0 ,-1 ,+1}, be transformed to+1 ,-1,0 ,-1 ,+1,0,0,0,0,0}, in the time of promptly not enough 10, back location is filled 0 value.Be convenient to like this calculate and unified vectorial dimension, also can confirm as other length.
(3) sample storehouse structure
Utilize the peak value translative mode, convert the target actual signal of gathering to peak-mode, set up the sample storehouse through test.For twin shaft magnetic resistance and microseismic activity transducer, the example such as the following table of the target sample characteristic signal of foundation:
(4) Characteristic Recognition
Characteristic Recognition is that actual signal and other standard feature signal of target class that target produces are compared:
R＝|c
1-c
b1|+|c
2-c
b2|+…+|c
10-c
b10|
When sampling the composite signal of target, this as recognition feature, is compared with corresponding vector in the standard feature signal, calculate the distance between them, again each is compared apart from sum.Which is minimum apart from sum, then with the target type of its pairing standard feature as recognition result.
For instance, as far as the magnetoresistance signal of target, the magnetoresistance signal of establishing a certain node collection after changing for {+1 ,-1 ,+1 ,-1,0 ,-1 ,+1,0,0,0} can get with the magnetoresistance signal master pattern comparison of following formula and each wheeled vehicle: R
Compact car=4, R
Passenger vehicle=11, R
Large car=1, R then
Large carValue is minimum, judges that then this target is a large car.
To the echo signal that two kinds of compound sensors produce, suppose R
1The result of calculation of expression magnetoresistance signal, R
2The result of calculation of expression microseismic activity signal, the numerical value sum (R that these two kinds of transducers of COMPREHENSIVE CALCULATING produce
1+ R
2).If the comprehensive numerical value sum (R that certain target type is corresponding
1+ R
2) minimum, then with the target type of its pairing standard feature as classification results.
In addition; The target actual signal that also can adopt simpler and more direct mode to come comprehensive these two kinds of transducers to produce; Realize target classification: after sensor node is accomplished the result of detection to a certain target, at first the data transaction that collects is become peak-mode (comprising magnetic resistance and microseismic activity signal), the microseismic activity signal of peak-mode and the target identification regular set of having set up are compared; Comparison through the microseismic activity signal; Directly determining this target is personnel, wheeled vehicle or caterpillar, the matching detection magnetoresistance signal of gathering then, thus further determine the specific category under the target.
In native system, transmit data between the baud rate of gateway aggregation node employing 11520Baud/s and the computer of monitoring center, preserve data with the form of Excel database table lattice file on the backstage.Twin shaft magnetoresistive transducer and microseismic activity transducer are integrated to be placed on the roadside at a terminal node with probe node, and placement location and road direction or its vertical direction are consistent, and the vertical range of node and target is no more than 3 meters.The gateway base-station node of sensor network and the supervisory computer of Surveillance center are positioned at same place.Sample frequency is set to 20Hz, i.e. the signal data of two kinds of transducers of per second collection is 20 times.Here consider single target is carried out Classification and Identification, do not consider a plurality of moving targets situation through probe node simultaneously.Real system is reliable, and the data sampling process is steady, in time the type of the output target of passing through.
Gathering the zero-time of moving target signal in the test confirms as follows:
1. when moving target proximity transducer probe node; Magnetoresistive transducer just can collect the containing metal body and move caused disturbance of magnetic field amount; The magnetic resistance perturbation process resets normally when target is left away; Thereby, confirm that the distance of target and sensor node is about the scope of 3-5 rice according to the sensitivity and the target type of Magnetic Sensor.Because the detection range of magnetoresistive transducer is limited, if the magnetoresistive transducer frequency acquisition is 20Hz, only in this limited distance, magnetoresistive transducer just can collect effective target magnetoresistance signal, and this also is effective detection time of magnetoresistive transducer.
2. because the operating distance of microseismic activity transducer is far away; The microseismic activity signal that for example on hard ground, transmits reaches 300 meters; Consider the acting in conjunction effect of two kinds of compound sensors; Only collect the microseismic activity signal that the moving target within the magnetoresistive transducer operating distance scope produces, this distance is just above-mentioned 3-5 rice usually.The microseismic activity transducer is with the frequency collection microseismic activity signal of 20Hz; Although when distance is far away, just perceived signal output; Here choose the time range of effective microseismic activity signal of collection; Be limited to moving target through near the node time, i.e. within the time of the effective perception of magnetoresistive transducer and the distance.Because the nodal distance target is near more, the microseismic activity signal is strong more, thereby this way has science.
Accompanying drawing 5 is magnetoresistance signal feature samples examples of armed personnel (or the personnel that carry metallic object); Accompanying drawing 6 is microseismic activity signal characteristic sample examples of armed personnel (or the personnel that carry metallic object), and these two examples intuitively and have clearly reflected the magnetoresistance signal of this class targets and the characteristic of microseismic activity signal.In these two accompanying drawings, abscissa is sampling instant (ms), and ordinate is sensor output voltage (mv).Two kinds of primary signals that target produces, after through conversion, produce and form three value forms+1,0, the peak signal of-1}.
The compact car of statistical sample car normally in part test, passenger vehicle mainly is a bus, large car is common landship, caterpillar with crawler dozer and excavator as subjects.The result is as shown in the table for the part statistical test, and accuracy can maintain more than 80%.
| Free-hand personnel | The armed personnel | Compact car | Passenger vehicle | Large car | Endless-track vehicle | |
| Number of targets | 50 | 50 | 50 | 30 | 30 | 20 |
| Correct class object number | 44 | 47 | 41 | 29 | 26 | 20 |
| Classification accuracy rate | 88％ | 94％ | 82％ | 96％ | 86％ | 100％ |
Show that through field trial this method can be used for discerning the target type on the key road segment, the accuracy of Classification and Identification is than higher.
Because the physical limitation of various transducers self, can not reach 100% target identification accuracy according to the little information source.Therefore when being necessary and the application scenario, also can gather the actual signal of target, multiple information source is carried out fusion treatment, can further improve the accuracy of target classification identification by the sound, other transducer such as infrared.
The above is the preferable main implementation process of the present invention, is not to be used for limiting practical range of the present invention; Every according to equivalence variation and modification that the present invention did, also contained by claim of the present invention.
Claims (1)
1. moving target classification identification method based on compound sensor Ad Hoc network; Be characterised in that: the wireless Ad Hoc sensor network system is after deployment finishes; Each network node provides the function of target classification identification through to the magnetoresistance signal of the actual generation of ground moving object and the collection of microseismic activity signal; According to the magnetic resistance of target and the compound characteristics of microseismic activity signal, common ground moving object is categorized as free-hand personnel, armed personnel or carries six types of personnel, dilly, passenger vehicle, oversize vehicle and the caterpillars of metallic object;
The operating procedure of objective classification method is following:
Step 1: threshold design
Threshold value is that computational methods are following through test statistics generation in advance:
Wherein, Fiducial value be driftlessness through and glitch-free situation under, the output circuit voltage of network node; Undulating value be driftlessness through and under the noisy situation, the output circuit voltage of network node, counting at interval is the number of two undulating value data between the fiducial value;
Step 2: peak value conversion
The peak value transfer process is the actual signal that produces when moving target is passed through each network node, converts three binarization modes to: and+1,0 ,-1}; The rate of change and the threshold value of the actual signal that moving target is produced when each network node compare, if for just and greater than threshold value, then be characterized as+1; If for bearing and, then being characterized as-1 quantitatively greater than threshold value; If the numerical value of rate of change is less than or equal to threshold value, then be characterized as 0;
Step 3: sample storehouse structure
Gather the magnetoresistance signal of moving target generation and the sample of microseismic activity signal in advance through test, utilize the peak value transfer process to convert these sample signals to three binarization modes respectively, form two class standard characteristic signals; With this sample storehouse of constructing target signature, realize the differentiation of six class targets: 1. free-hand personnel, 2. armed personnel or the personnel that carry metallic object; 3. dilly; 4. passenger vehicle, 5. oversize vehicle, 6. caterpillar;
Step 4: Characteristic Recognition
When sensor network system is carried out the target monitoring task; As long as moving target is through near each network node; The data acquisition assembly of this network node just carries out collection in worksite to magnetic resistance and these two kinds of signals of microseismic activity that moving target produces; Two class standard characteristic signals in two types of signal results of gathering and the sample storehouse are compared respectively, calculate the distance between them respectively, sue for peace apart from addition for two types; Which kind of target type minimum relatively again apart from summation, then with the target type of minimum range summation as recognition result;
Described Characteristic Recognition process details is operated as follows:
When adopting the standard feature signal to carry out the identification of target type, on-the-spot actual signal that produces of moving target and the standard feature signal in the sample storehouse are compared calculating:
R＝|c
1-c
b1|+|c
2-c
b2|+…+|c
10-c
b10|
Wherein, { c
1, c
2C
10Be three binarization mode data of the on-the-spot actual signal that produces of moving target of network node collection, { c
B1, c
B2C
B10Be the data of the standard feature signal in the sample storehouse, R is the vectorial difference between these two kinds of data;
When the data acquisition assembly of each network node collects magnetoresistance signal and the microseismic activity signal of target, calculate difference R; Suppose R
1The difference comparative result of expression magnetoresistance signal, R
2The difference comparative result of expression microseismic activity signal, the difference of these two kinds of signals is summed to (R apart from addition
1+ R
2); If the COMPREHENSIVE CALCULATING result (R that certain target type is corresponding
1+ R
2) numerical value minimum, then with the target type of its pairing standard feature signal result as Classification and Identification; Monitoring center's computer of sensor network system is responsible for collecting the situation of target classification identification, and is saved in database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009101850784A CN101868045B (en) | 2009-10-30 | 2009-10-30 | Moving target classification identification method based on compound sensor Ad Hoc network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009101850784A CN101868045B (en) | 2009-10-30 | 2009-10-30 | Moving target classification identification method based on compound sensor Ad Hoc network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101868045A CN101868045A (en) | 2010-10-20 |
| CN101868045B true CN101868045B (en) | 2012-04-18 |
Family
ID=42959543
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009101850784A Expired - Fee Related CN101868045B (en) | 2009-10-30 | 2009-10-30 | Moving target classification identification method based on compound sensor Ad Hoc network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101868045B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103914995B (en) * | 2014-03-28 | 2015-10-28 | 南京邮电大学 | A kind of traffic safety control method based on wireless network |
| CN106211139B (en) * | 2016-08-30 | 2019-04-30 | 单洪 | A kind of recognition methods encrypting MANET interior joint type |
| CN106650599A (en) * | 2016-10-14 | 2017-05-10 | 北京智眸科技有限公司 | A method for setting sparse sampling frequency regionally and selecting sampling points in stereo matching |
| CN110334734B (en) * | 2019-05-31 | 2024-08-09 | 宁波中车时代传感技术有限公司 | Intelligent sensing fusion method based on meta learning technology |
| CN110209281B (en) * | 2019-06-06 | 2022-03-15 | 瑞典爱立信有限公司 | Method, electronic device, and medium for processing motion signal |
| CN110488301B (en) * | 2019-07-17 | 2023-01-06 | 中国人民解放军91388部队 | Multi-source information fusion sonar comprehensive target identification method |
| CN116910690A (en) * | 2023-07-14 | 2023-10-20 | 中国兵器装备集团自动化研究所有限公司 | Target classification system based on data fusion |
-
2009
- 2009-10-30 CN CN2009101850784A patent/CN101868045B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN101868045A (en) | 2010-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101868045B (en) | Moving target classification identification method based on compound sensor Ad Hoc network | |
| Song et al. | A novel convolutional neural network based indoor localization framework with WiFi fingerprinting | |
| Cheung et al. | Traffic surveillance by wireless sensor networks | |
| CN102737510B (en) | Real-time traffic condition acquisition method based on mobile intelligent terminal | |
| CN102932812B (en) | Vehicle sensor concurrent monitoring method facing road conditions | |
| CN101556724B (en) | Safety management system of optical fiber perimeter and pattern recognition method thereof | |
| CN101865993B (en) | Target tracking method based on binary sensor Ad Hoc network | |
| CN105243844A (en) | Road state identification method based on mobile phone signal | |
| CN110542898A (en) | Radar group-based vehicle behavior continuous tracking detection system and method | |
| CN106856049B (en) | Key intersection demand aggregation analysis method based on bayonet number plate identification data | |
| CN102779410B (en) | Parallel implementation method of multi-source heterogeneous traffic data fusion | |
| CN108109423A (en) | Underground parking intelligent navigation method and system based on WiFi indoor positionings | |
| CN102646332A (en) | Traffic state estimation device and method based on data fusion | |
| CN103327091A (en) | System and method for obtaining passenger track and behavioral parameter | |
| CN101938832A (en) | Division and refinement-based node self-positioning method for wireless sensor network | |
| CN103344941B (en) | Based on the real-time target detection method of wireless sensor network | |
| CN108431837A (en) | Method and system for the stroke performance for evaluating driver | |
| CN102722987A (en) | Roadside parking space detection method | |
| CN102724751A (en) | Wireless indoor positioning method based on off-site survey | |
| CN112258850A (en) | Edge side multi-sensor data fusion system of vehicle-road cooperative system | |
| CN105101090A (en) | Node positioning method of wireless sensor network for environmental monitoring | |
| Prentow et al. | Towards indoor transportation mode detection using mobile sensing | |
| CN108369681A (en) | Method and system for the stroke performance for evaluating driver | |
| CN116092037B (en) | Vehicle type identification method integrating track space-semantic features | |
| Di | Investigation on the traffic flow based on wireless sensor network technologies combined with FA-BPNN models |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| ASS | Succession or assignment of patent right |
Owner name: PLA MILITARY COLLEGE Free format text: FORMER OWNER: PLA ARTILLERY COLLEGE Effective date: 20130325 |
|
| C41 | Transfer of patent application or patent right or utility model | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20130325 Address after: 230031 No. 451, Mount Huangshan Road, Shushan District, Anhui, Hefei Patentee after: PLA Military Academy Address before: 230031 No. 451, Mount Huangshan Road, Shushan District, Anhui, Hefei Patentee before: China PLA Artillery College |
|
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120418 Termination date: 20151030 |
|
| EXPY | Termination of patent right or utility model |