CN108805218A - Optical target association method based on deviation mapping clustering - Google Patents
Optical target association method based on deviation mapping clustering Download PDFInfo
- Publication number
- CN108805218A CN108805218A CN201810644970.3A CN201810644970A CN108805218A CN 108805218 A CN108805218 A CN 108805218A CN 201810644970 A CN201810644970 A CN 201810644970A CN 108805218 A CN108805218 A CN 108805218A
- Authority
- CN
- China
- Prior art keywords
- target
- sensor
- point
- association
- deviation
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses an optical target association method based on deviation mapping clustering, and aims to solve the problem of target association of an optical multi-sensor system under a multi-target complex scene. Firstly, converting target observation data of different sensor systems into a common imaging plane, and calculating the deviation of all possible target association pairs between two imaging systems in the plane; mapping the deviation of each group of associated matching pairs to a point on a two-dimensional plane; and calculating the density of a local neighborhood point set of each deviation mapping point, taking the maximum density value as a clustering center point, and repeatedly searching the mapping points with the closest spatial distribution around until the number of the mapping points is equivalent to the preset target number of the system, so as to obtain the final association ID contained in each mapping point in the clustering point set, namely the target association result. The method solves the problems that the traditional multi-target association is easily influenced by the interference target and is not suitable for complex scenes with a large number of targets, and the like, and improves the effectiveness and efficiency of the target association.
Description
Technical field
The invention belongs to optics multi-sensor information fusion fields, are related to the association of target observation between optics multisensor
Match, is based especially on the method that the Preference-Deviation Mapping cluster of target observation value realizes target association.
Background technology
In recent years, with the development of information technology, the multiple sensor integrated method suitable for complex scene has obtained extensively
Research and application.Its advantage is that system using the redundancy between multiple sensor measurement informations, improves the life of system
Ability and coverage area are deposited, by merging the independent measuring value of multiple sensors, may be implemented more better than Method for Single Sensor System
Effect.When multi-sensor data association refers to multiple sensors while observing multiple targets, the difference of same target will be belonged to
Measuring value between sensor is associated, it is simply that determining which sensor measuring value comes from same target.This is just
It is the core basic problem in modern sensor system --- target object figure (TOM) matches, and the quality of data correlation is very big
The quality of Multi-sensor Fusion performance is determined in degree.
When multiple sensors observe multiple corresponding targets simultaneously, due to target spatial distribution and sensing station it is opposite
Relationship, there is also differences for imaging on each sensor imaging plane, directly carry out target association at this time and have difficulties.Generally
When carrying out target association, need first based on the spatial relation between sensor by the measuring value in different sensors plane of vision
In conversion to public plane of vision.Ideally, these transformed measuring values should match with observation true value, same mesh
Target measuring value should be consistent.But during actual observation, influenced by sensing system difference and application environment, it is real
Corresponding to target distribution after the conversion of border can existence position deviation.Wherein often there is system deviation, this systems for sensor itself
Deviation is different from the random error that sensor measures, and can not be a kind of fixed deviation by repeatedly measuring elimination, when calculating
It can convert with the coordinate between sensor and change.In addition to deviation effects, seen by the falseness that ambient noise and clutter are formed
Target is surveyed, all brings many difficulties to related question.
Before being analyzed in principle and performance the prior art and method proposed by the invention, just generally answer first
Master-slave mode sensor information fusion system provides acquisition and the conversion method of target observation information:
(1) target observation obtains:
By optics multisensor syste observation space multiple target, it is imaged in two-dimensional focal plane array, and by target P in
The observed result (coordinate) of heart sensors A and sub- sensor B are denoted as respectivelyWith
(2) common image plane conversion:
Using the observation coordinate system of central sensor A as common coordinate system, three-dimensionals of a certain target P in space under the coordinate system
Space true coordinate is ZP=(xP,yP,zP)T, it is in the image plane imager coordinate of sensors AAssuming that
Three dimensions true coordinates of the sub- sensor B under the common coordinate system is ZB=(xB,yB,zB)T.It is now assumed that sensors A and B
The translation of spatial position is only existed, and optical axis direction remains unchanged, then three dimensions true coordinates of the target P at sensor B
It can be according to being above expressed asIt is corresponding can be obtained target P sensor B image plane at
As coordinate is
By image plane transfer principle, it is known that image plane coordinate (up,vp):
In formula:(θ is viewing field of camera angle, and N is image plane transverse and longitudinal resolution cell number) is sensor
Camera instantaneous field of view angle, i.e. field angle size corresponding to image plane unit pixel.
Coordinate under different sensors imaging plane can be derived to obtain according to formula 1WithCorrespondence is:
So far, the target observation value under different imaging planes is converted into two groups of target observations in public imaging plane
Value is denoted as observation set respectivelyWhereinTo belong to center sensing
The measurement collection of device A,It is converted to the collection table under sensors A coordinate system by space for the observation collection of sub- sensor B
Show.
To solve the problems, such as that target association, the prior art are mainly to form target by the statistical result of deviation between calculating target
The cost matrix (formula 3 and 4) of distribution, the problem of converting the related question between target to integral dispensing Least-cost, in turn
Solved using Hungary Algorithm, auction algorithm, synthesis clustering algorithm etc., handle the final result after cost matrix to get
To the distribution for target association between sensor.The essence of these methods is all merely with target observation offset distance value, and mesh
It is the one-dimension information that cannot embody direction character to mark observed deviation distance value.
It can be seen that the prior art is used as distribution cost merely with the space length between target, it is also easy to produce accidentally correlation;
It is vulnerable to the influence of jamming target under the situations such as target false-alarm, missing inspection;Under the more complex scene of number of targets, allocation algorithm
Complexity is high, and does not have adaptability to target mismatch.
When the present invention is in view of different sensors observation multiple target, correspond between sensor target observation from the distance of deviation and
On the direction of offset, all there is approximate consistent characteristic.In target observation 3D distributions shown in Fig. 2 (a), thick line type mark represents
The different sensors observation point of sensors A observation position, filament type mark representative sensor B observation positions, same target is shared
Same type mark;Fig. 2 (b) be sensors A and B to target observation result public image plane projection.Observation comes from same
The observation of the different sensors of one target, although the bias property of different sensors is different, while also by random error
When influencing, but being projected to public image plane, they have approximate consistency on the dispersive distribution in space.It therefore can profit
With the consistency of this deviation, a bit for two dimensional surface by each pair of Preference-Deviation Mapping, distribution is (star-like to represent coordinate as shown in Figure 3
Origin, round scatterplot represent the mapping point of deviation pair, are that dot density is distributed maximum region in oval frame), Fig. 3 (a)~(b)
It can be seen that unordered distribution at random is presented in mapping point in two dimensional surface, but there are dot density distributions in oval frame inner region
Maximum, wherein the matching with similar bias property is also almost the same to mapped point position, then local dot density
The maximum place of peak value can be considered correct object matching pattern.Therefore it using the method for point cluster in plane, finds
Mapping point center point set, so that it may to realize the efficient association to the data observation value of same target.
In conclusion the problem of target association is multi-sensor information fusion field urgent need to resolve, existing target association side
Method is confined to the processing to target association cost matrix, substantially the analysis to the one-dimension information of target deviation.Experiment is tested
Card and practical application, which show that the prior art exists, to be handled very well under actual complex background environment, and object space is not close to
Correlation is missed caused by more than easy differentiation, noise jamming target etc.;Simultaneously there is also increasing with destination number, it is associated under efficiency
Very fast problem drops.Therefore, it is this to study rapidly and efficiently and have the target association method of higher robustness to jamming target scene
The technical issues of field technology personnel extremely pay close attention to.There is no open source literature to relate to the use of deviation approximation one in existing technical research
Cause property and the method for taking two-dimentional deviation cluster to carry out target association, Method And Principle proposed by the present invention and experimental verification are feasible.
Technical solution
The problem to be solved in the present invention is the association matching of target observation between multisensor syste under multiple target complex scene
Problem.By the research to object space distribution character, target association is carried out based on target deviation mapping point cluster, solves tradition
The influence of jamming target is accidentally associated with, be vulnerable to existing for target association, and can not be suitable for the complex scene of a greater number target
The problem of.
The present invention basic principle be:
Since imaging of the multiple target under different sensors imaging plane has similar distribution, by non-thread between sensor
The space coordinate conversion of property, the certainty feature caused by sensing system deviation are preserved well, and relatively
The uncertain feature that small random error generates is blanked.Distance and offset of the target observation from deviation are corresponded between sensor
Direction on, all there is approximate consistent characteristic.Using the consistency of this deviation, different target is associated with to deviation by the present invention
A bit is mapped as in two-dimentional deviation plane, and each mapping point represents a kind of possible outcome of target association, close using search part
The method of degree peak value finds mapping point cluster centre.The cluster centre actually just represent target be correctly associated with it is matched as a result, from
The label information of each mapping point extraction target association of cluster centre, may finally obtain the target association letter between sensor
Breath.
The present invention includes three steps:
The first step:Target association is calculated to deviation.I.e. for existing two targets to be associated in public imaging plane
Collection is measured, coordinate is expressed asWithIt is calculated two in image plane by formula 5
The deviation bias of all possible target association pair between detection systemi→j, including the association pair for really belonging to same target,
Also include the non-false pairing for belonging to same target.
Second step, deviation Planar Mapping point cluster.It is indistinguishably two dimension by the Preference-Deviation Mapping of each group of association matching pair
Any in plane is calculated wherein the matching with similar bias property is also almost the same to mapped point position by formula 6
The local density ρ of each mapping pointi(wherein dijMatched offset distance bias is observed for i-th and j groupsi→j, block distance dcAnd χ
(x) selection such as formula 7 and formula 8), and it is cluster centre to choose the wherein maximum match point of density, searching for M around (is
Preset target association unite to number) a match point, it is contemplated that same sub-platform target point Xi(xi,yi) and central platform target
Point Yj(uj,vj) single correspondence, scanned for according to following steps:
(1) choosing has maximum part dot density max { ρi, i=1,2 ..., p × q cluster centre point Pcenter, and
It extracts it and matches ID { i, j };
(2) it centered on the point, is chosen successively closest to PiPoint Pi, examine its match ID, if not with the pairing of front
Conflict, then be fused to pairing set A () in;If matching ID conflicts, i.e.,
There are i → j1,i→j2Or i1→j,i2It is closer to retain optical strength information then by comparing objective optics intensity size by → j
Pairing;
(3) (2) are repeated, until that selects that M groups do not conflict nearest matches to until.
Third walks, and records final Fusion of Clustering point set A, extracts the association ID that each mapping point is included in point set, i.e.,
Obtain target association result.
Beneficial effects of the present invention are:
During the target association of multisensor syste, influenced by sensor itself difference and observing environment, Yi Cun
It is influenced in false jamming target, high requirement is proposed to target association algorithm at this time.Target observation between sensor is utilized merely
Offset distance, out of global scope choose target association total cost minimum associative combination, easily by around same pairing target
Other close observation interference are distributed, the phenomenon that being accidentally associated with is generated.
The present invention utilizes the approximate consistency of target observation image plane distribution between sensor, observing different sensors
Target deviation is clustered after being mapped as a little, is analyzed by the density peaks point set formed to cluster, to effectively
Immediate matching pair is isolated from be possible to target association pair, greatly reduces the problems such as being accidentally associated with.Meanwhile it can be effective
Using, in the geometry information of image plane distribution, when the multiple target under complex scene is matched, energy is more preferable between target
The influence of jamming target is avoided, while in multi-to-multi association and inconsistent object matching number, being had compared with the existing technology
Stronger adaptability.
Description of the drawings
Fig. 1 is the step figure of the present invention;
Fig. 2 is target profile;
Fig. 3 is that target association is matched to deviation profile figure;
Fig. 4 is the implementation process diagram of the present invention;
Fig. 5 is that Target Assignment emulates schematic diagram;
Fig. 6 is four kinds of method target association results contrast figures;
Specific implementation mode
This method is suitable for the multi-sensor Information Fusion System of each form, and different form system chooses benchmark imaging plane
May be variant, but implementation method is similar.For the sake of for convenience of description, it is generation that this, which sentences commonly used master-slave mode emerging system,
Table is chosen two sensor (central sensor A and sub- sensor B, multiple sub- sensor, methods are identical) systems progress steps and is said
It is bright.
The first step:Multiple target sensing system is imaged
By different sensing system observation space multiple targets, at observation picture on different system imaging plane.By 3D
On the target projection of spatial distribution to 2D imaging planes, projection of the target in sensors A and the focal planes B is formed, coordinate can
It is expressed asWith
Second step:Public imaging plane target observation conversion
On the basis of obtaining different sensors imaging, using formula 2, sub- sensor target imaging conversion to center is passed
Sensor plane.It can obtain observation setWhereinTo belong to center
The measurement collection of sensors A,It is converted to the collection under sensors A coordinate system by space for the observation collection of sub- sensor B
It closes and indicates, shown in imaging flow path switch figure such as Fig. 4 (a).
Third walks:Target association is calculated to deviation
On the basis of based on common image plane projection, all possible target is closed between two imaging systems in calculating image plane
The deviation of connection pair includes the association pair for really belonging to same target when calculating, also includes that the false of non-same category matches.It follows
To keeping unique principle, then all possible target association is m × n pairs shared, accordingly also will produce m for target association matching
× n groups are associated with deviation biasi→j, wherein biasi→jComputational methods shown in formula 5.So far, it can obtain all to be associated
Two-dimentional deviation coordinate when target is mutually matched.
4th step:Deviation Planar Mapping point clusters
Indistinguishably by the deviation bias of each group of association matching pairi→jA point being mapped as in two-dimentional deviation plane, profit
With the superiority of clustering algorithm, the closer mapping point of spatial distribution is gathered for one kind, then just m × n groups may be matched
To two-dimentional variance analysis problem be converted into the point clustering problem in plane.
The local neighborhood dot density ρ of each Preference-Deviation Mapping point is calculated by formula 6 firsti, with certain calculating point PiCentered on, by
Distance d is blocked in the calculating of formula 7c, when certain mapping point and point distance are less than dcWhen, dot density ρiAdd one, otherwise remain unchanged, directly
All mapping points in traversal plane, obtain the ρ of the pointiValue.Then selected point density piIt is cluster centre point to be worth maximum point
Pcenter, it is chosen to be a target association to being placed in the generic, and the pairing ID { i, j } corresponding to it is placed in and is matched
To in set A.Then centered on the point, into surrounding neighbors, search space is distributed immediate mapping point successively, then will
Its tag ID is compared with the tag ID in foregoing assemblage A.If the case where conflicting to ID in the presence of association, i.e. central sensor A
Or there is a target observation value in sub- sensor B observation and match and multiple corresponding observe that (symbolic indication is:i→j1,i→j2Or
i1→j,i2→ j), in order to screen, which association is true to being only under conflict situations, then compares it and detect optical strength information.
Consider that the optical strength information of detection is determined that the position of detector influences can be neglected on it, same by target build-in attribute
Target its optical strength under different detection systems should be sufficiently close to, and the closer observation of two strength informations is finally selected to make
To come from the observation matching of same target.If being associated with ID, there is no conflicts, it is directly fused in pairing set A.
5th step:Target association relationship is extracted in gathering from pairing.
The finally formed pairing set A of previous step is recorded, the relationship maps that each mapping point is included in the set are extracted
(matching ID), to obtain target association as a result, the deviation of wherein third to five steps clusters target association flow chart such as Fig. 4
(b) shown in.
Interpretation of result:
Association performance should now be invented and be compared analysis experiment.A kind of the case where considering target mismatch first (Fig. 5 (a)
It is shown), on public plane of vision, sensors A (circle expression) has 7 measuring values, sensor B (star-like expression) to have 5 amounts
Measured value, and for the part in oval frame, the measurement of each sensor B has nearby corresponded to the measuring value of two A, such case
Under may cause accidentally be associated with pairing the case where.Preference-Deviation Mapping cluster BMC (shown in Fig. 5 (b)) side proposed by the present invention is respectively adopted
Method and Hungary's distribution method (Hungarian methods, shown in Fig. 5 (c)) emulate it, it can be seen that BMC methods are to same target
Nearby there is multiple may match has very strong resolving power when measuring, it can be carried out effective using the distributions shift information of target
Matching;But optimized using offset distance as cost matrix solution Hungarian methods be easy by distance closer to
The interference of other measurements, generate accidentally be associated with the phenomenon that, as long as and interfere measure distribution it is spatially close enough when, this mistake
Association cannot avoid.Similarly analysis can obtain, the global arest neighbors matching method based on offset distance and the side apart from cluster
Method is not avoided that this accidentally association situation.I.e. in the case of target mismatch, and BMC adaptations of methods proposed by the present invention
It is significantly better than the prior art.
Further, we cluster target association method and the breast of (BMC) also directed to Preference-Deviation Mapping proposed by the invention
Tooth profit (Hungarian) matching process, synthesis cluster association method (ACB) and the nearest neighbor method (NN) for randomly selecting target starting
Etc. the prior arts association accuracy carry out more generally performance comparison test.Testing us every time all will carry out covering spy 1000 times
Carlow emulates.In the case where central sensor A and sub- sensor B observe 10 targets, the association of four kinds of methods is correct
Rate is with noise variation as shown in 6 (a), and association accuracy is with sensor bias variation as shown in 6 (b);There are target mismatches
In the case of, i.e., sub- sensor B observes 5 targets, the number of targets that central sensor A is observed from 5 change to 10 when, association
Shown in accuracy situation of change such as 6 (c).The experimental results showed that comparing other target association methods, the present invention proposes that BMC methods exist
It is associated with accuracy higher under various situations, and there is stronger robustness.Number is observed especially under complex background condition to differ
When causing to generate target mismatch, more other more have superiority using the method for range information, can relatively efficiently overcome sensor
Between the interference that measures of target falseness, improve sensing system to the adaptability of false-alarm and missing inspection.
Claims (3)
1. a kind of optical target correlating method based on Preference-Deviation Mapping cluster waits closing for existing two in public imaging plane
Join target and measure collection, coordinate is expressed asWithIt is characterized by comprising with
Lower step:
The first step, on the basis of public image plane, by all possible target between two detection systems in the calculating image plane of formula 1
The deviation bias of association pairi→j, also include the false pairing of non-same category including the association pair for really belonging to same target.
The Preference-Deviation Mapping of each group of association matching pair is indistinguishably a bit on two dimensional surface, wherein having phase by second step
Matching like bias property is also almost the same to mapped point position, and the local density ρ of each mapping point is calculated by formula 2i
(wherein dijMatched offset distance bias is observed for i-th and j groupsi→j, block distance dcSelection such as formula 3 and formula with χ (x)
4), and the selection wherein maximum match point of density is cluster centre, searches for M around (target association of systemic presupposition is to number)
A match point, it is contemplated that same sub-platform target point Xi(xi,yi) and central platform target point Yj(uj,vj) single corresponding close
System, scans for according to following steps:
(1) choosing has maximum part dot density max { ρi, i=1,2 ..., p × q cluster centre point Pcenter, and extract it
Match ID { i, j };
(2) it centered on the point, is chosen successively closest to PiPoint Pi, examine it to match ID, if do not liquidated with matching for front
It is prominent, then it is fused to pairing setIn;If matching ID conflicts, i.e.,
There are i → j1,i→j2Or i1→j,i2It is closer to retain optical strength information then by comparing objective optics intensity size by → j
Pairing;
(3) (2) are repeated, until that selects that M groups do not conflict nearest matches to until.
Third walks, and records final Fusion of Clustering point set A, extract the association ID that each mapping point is included in point set to get to
Target association result.
2. the target association method according to claim 1 based on Preference-Deviation Mapping cluster, it is characterised in that:Commonly used
Master-salve distributed sensor information fusion system in, sensor centered on A, B be sub- sensor, imaging and switch process such as
Under:
The first step is imaged in two-dimensional focal plane array by more optical sensor system observation space multiple targets, and by target P
It is denoted as respectively in the observed result (coordinate) of central sensor A and sub- sensor BWith
Second step obtains each sensor space location information, using central sensor A observation coordinates system as common coordinate system, target P
And coordinate minute marks of the sub- sensor B under common coordinate system is ZP=(xP,yP,zP)TAnd ZB=(xB,yB,zB)T.It will by formula 5
Sub- sensor observed object is converted to central sensor imaging plane, and observation set is denoted as after conversionWithWhereinTo belong to the measurement collection of central sensor A,For the sight of sub- sensor B
Collection is surveyed to convert to the set expression under sensors A coordinate system by space.
3. the target association method according to claim 2 based on Preference-Deviation Mapping cluster, it is characterised in that:Sub- sensor B
Can be multiple.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810644970.3A CN108805218B (en) | 2018-06-21 | 2018-06-21 | Optical target association method based on deviation mapping clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810644970.3A CN108805218B (en) | 2018-06-21 | 2018-06-21 | Optical target association method based on deviation mapping clustering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108805218A true CN108805218A (en) | 2018-11-13 |
CN108805218B CN108805218B (en) | 2021-12-28 |
Family
ID=64084071
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810644970.3A Active CN108805218B (en) | 2018-06-21 | 2018-06-21 | Optical target association method based on deviation mapping clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108805218B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112465065A (en) * | 2020-12-11 | 2021-03-09 | 中国第一汽车股份有限公司 | Sensor data association method, device, equipment and storage medium |
CN112529061A (en) * | 2020-12-03 | 2021-03-19 | 新奥数能科技有限公司 | Identification method and device for photovoltaic power abnormal data and terminal equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106796657A (en) * | 2014-11-14 | 2017-05-31 | 英特尔公司 | For the automatic target selection of multiple target Object tracking |
CN107239788A (en) * | 2017-04-20 | 2017-10-10 | 浙江工业大学 | The optimal selection Spectral Clustering of characteristic vector group based on degree adaptive |
WO2018067158A1 (en) * | 2016-10-06 | 2018-04-12 | Strobe, Inc. | Lidar system |
-
2018
- 2018-06-21 CN CN201810644970.3A patent/CN108805218B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106796657A (en) * | 2014-11-14 | 2017-05-31 | 英特尔公司 | For the automatic target selection of multiple target Object tracking |
WO2018067158A1 (en) * | 2016-10-06 | 2018-04-12 | Strobe, Inc. | Lidar system |
CN107239788A (en) * | 2017-04-20 | 2017-10-10 | 浙江工业大学 | The optimal selection Spectral Clustering of characteristic vector group based on degree adaptive |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112529061A (en) * | 2020-12-03 | 2021-03-19 | 新奥数能科技有限公司 | Identification method and device for photovoltaic power abnormal data and terminal equipment |
CN112529061B (en) * | 2020-12-03 | 2024-04-16 | 新奥数能科技有限公司 | Photovoltaic power abnormal data identification method and device and terminal equipment |
CN112465065A (en) * | 2020-12-11 | 2021-03-09 | 中国第一汽车股份有限公司 | Sensor data association method, device, equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN108805218B (en) | 2021-12-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9816939B2 (en) | Virtual inspection systems with multiple modes | |
CN104700404B (en) | A kind of fruit positioning identifying method | |
Pollock | The automatic recognition of individual trees in aerial images of forests based on a synthetic tree crown image model | |
Krotosky et al. | Mutual information based registration of multimodal stereo videos for person tracking | |
CN106485740B (en) | A kind of multidate SAR image registration method of combination stable point and characteristic point | |
CN105957082A (en) | Printing quality on-line monitoring method based on area-array camera | |
US20110025834A1 (en) | Method and apparatus of identifying human body posture | |
CN105874350A (en) | Calibration device, calibration method, and calibration program | |
CN109341668A (en) | Polyphaser measurement method based on refraction projection model and beam ray tracing method | |
CN105741379A (en) | Method for panoramic inspection on substation | |
CN103591940B (en) | Method of evaluating confidence of matching signature of hyperspectral image | |
CN106780551B (en) | A kind of Three-Dimensional Moving Targets detection method and system | |
Nyaruhuma et al. | Verification of 2D building outlines using oblique airborne images | |
CN102494663A (en) | Measuring system of swing angle of swing nozzle and measuring method of swing angle | |
CN111046877A (en) | Millimeter wave image suspicious article detection method and system | |
CN108805218A (en) | Optical target association method based on deviation mapping clustering | |
Liu et al. | Infrared-visible image registration for augmented reality-based thermographic building diagnostics | |
Zhao et al. | Detecting insulators in the image of overhead transmission lines | |
US9256786B2 (en) | Method of identification from a spatial and spectral object model | |
Bostanci et al. | Feature coverage for better homography estimation: an application to image stitching | |
CN109064496A (en) | A kind of change detecting method of remote sensing images object hierarchy | |
CN109242900A (en) | Locating focal plane method, processing unit, locating focal plane system and storage medium | |
CN113744222A (en) | Three-dimensional modeling calibration method for yarn evenness based on inscribed circle | |
CN112700544A (en) | Three-dimensional live-action generation method for mine pit terrain | |
Xu et al. | Large-range reconstruction with non-pre-calibrated camera via MBM of laser referenced by radial-collinear-features |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |