CN106598856B - A kind of path detection method and device - Google Patents
A kind of path detection method and device Download PDFInfo
- Publication number
- CN106598856B CN106598856B CN201611153636.5A CN201611153636A CN106598856B CN 106598856 B CN106598856 B CN 106598856B CN 201611153636 A CN201611153636 A CN 201611153636A CN 106598856 B CN106598856 B CN 106598856B
- Authority
- CN
- China
- Prior art keywords
- route
- new route
- normal
- new
- huasdorff
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3688—Test management for test execution, e.g. scheduling of test suites
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention discloses a kind of path detection method and devices, solve current path detection all based on space characteristics, since the complexity of current path detection algorithm is high, result in the low technical problem of accuracy rate.Path detection method of the embodiment of the present invention includes: to get to acquire several normal route data, and determine proper motion corresponding with normal route data track；Proper motion track is trained, get corresponding Huasdorff distance, and to including that several Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, the normal route strategy of several determining normal route data of cluster is carried out by recursive algorithm to minimal graph partitioning algorithm treated Huasdorff figure；When getting the new route data of acquisition, space characteristics and/or velocity characteristic are carried out in conjunction with normal route strategy and/or curvature feature is compared new route data, it is the illegal route to the corresponding new route of new route data if new route data are unsatisfactory for normal route strategy.
Description
Technical field
The present invention relates to path detection technical field more particularly to a kind of path detection methods and device.
Background technique
Path testing (path testing) refers to a kind of technology according to path design test case, is frequently used for state
In conversion testing.Basis path testing method is on the basis of program control flowchart, and the loop constructed by analysis and Control is complicated
Property, substantially executable set of paths is exported, thus the method for design test case.The test case designed will guarantee testing
Each executable statement of middle program at least executes once.
Current path detection, since the complexity of current path detection algorithm is high, causes all based on space characteristics
Accuracy rate low technical problem.
Summary of the invention
A kind of path detection method and device provided in an embodiment of the present invention, solve current path detection all with space
Based on feature, since the complexity of current path detection algorithm is high, the low technical problem of accuracy rate is resulted in.
A kind of path detection method provided in an embodiment of the present invention, comprising:
It gets and acquires several normal route data, and determine proper motion corresponding with the normal route data
Track；
The proper motion track is trained, gets corresponding Huasdorff distance, and to including several
The Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, after minimal graph partitioning algorithm processing
The Huasdorff figure by recursive algorithm carry out cluster determination several normal route data normal route plan
Slightly；
When getting the new route data of acquisition, space characteristics and/or speed are carried out in conjunction with the normal route strategy
Feature and/or curvature feature are compared the new route data, if the new route data are unsatisfactory for the normal route
Strategy is then the illegal route to the corresponding new route of the new route data.
Optionally, it gets and acquires several normal route data, and determination is corresponding with the normal route data
Proper motion track specifically includes:
Several normal route numbers are carried out by several image acquisition devices for being set to several normal routes
According to acquisition；
According in the normal route object in the entire visual field of the image acquisition device with the normal road
The corresponding proper motion track of diameter data, and mobile mean filter processing is carried out to the proper motion track.
Optionally, the proper motion track is trained, gets corresponding Huasdorff distance, and to comprising
There are several described Huasdorff distance Huasdorff figures to carry out the processing of minimal graph partitioning algorithm, the minimal graph is divided and is calculated
Method treated Huasdorff figure carries out the normal of cluster determination several normal route data by recursive algorithm
Path policy specifically includes:
The proper motion track is trained, it is corresponding to get proper motion track spacing two-by-two
Huasdorff distance；
It include several described Huasdorff distance Huasdorff according to preset path spacing expanded range threshold value
Figure carries out the processing of minimal graph partitioning algorithm, and to the minimal graph partitioning algorithm, treated that the Huasdorff figure passes through recurrence
Algorithm carries out clustering the normal route strategy for determining several normal route data.
Optionally, when getting the new route data of acquisition, in conjunction with the normal route strategy carry out space characteristics and/
Or velocity characteristic and/or curvature feature are compared the new route data, if the new route data be unsatisfactory for it is described just
Normal path policy then specifically includes the corresponding new route of the new route data for the illegal route:
90% new route path point extends in the path spacing to be made whether to the new route data for getting acquisition
In range, and carry out the new route of the new route data and the intermediate value path Hausdorff of the path spacing expanded range
Whether distance is less than the Hausdorff distance in any edge path of the new route and the path spacing expanded range, if
It is to be, it is determined that the new route is normal route, and otherwise the new route is the illegal route；
And/or
Calculate get acquisition new route data new route speed, by Gaussian Profile to the new route speed into
Row modeling processing, and according to the first formulaThe new route is determined with mahalanobis distance again
It is whether similar to the normal route of the normal route strategy, if similar, it is determined that the new route is normal route, no
Then the new route is the illegal route；
Wherein, vi' be new route speed, mpIt is then the speed mean value of the normal route strategy, ∑ is path velocity point
The covariance matrix of cloth；
And/or
Calculate new route speed, new route acceleration and the new route position of the new route data for getting acquisition not
Continuous place passes through the second formulaNew route curvature is calculated, and by institute
Stating new route curvature, whether curvature mean value corresponding with the Gaussian Profile being fitted to the normal route strategy carries out mahalanobis distance
Similar comparison, if similar, it is determined that the new route is normal route, and otherwise the new route is the illegal route, wherein speed
Degree is vi', acceleration vi", x' and y' are respectively the first derivative of x and y.
Optionally, the path detection method further include:
When the new route is the illegal route, then the illegal route is monitored.
A kind of path detection device provided in an embodiment of the present invention, comprising:
Acquiring unit acquires several normal route data for getting, and determination and the normal route data phase
Corresponding proper motion track；
Training unit gets corresponding Huasdorff distance for being trained to the proper motion track, and
Including that several described Huasdorff distance Huasdorff figures carry out the processing of minimal graph partitioning algorithm, to the minimal graph
Partitioning algorithm treated Huasdorff figure carries out several described normal route data of cluster determination by recursive algorithm
Normal route strategy；
New route determination unit, for when getting the new route data of acquisition, in conjunction with the normal route strategy into
Row space characteristics and/or velocity characteristic and/or curvature feature are compared the new route data, if the new route data
It is unsatisfactory for the normal route strategy, then is the illegal route to the corresponding new route of the new route data.
Optionally, acquiring unit specifically includes:
Subelement is obtained, for carrying out several by several image acquisition devices for being set to several normal routes
The acquisition of the normal route data；
Handle subelement, for according in the normal route object in the entire visual field of the image acquisition device
Proper motion track corresponding with the normal route data, and mobile mean filter is carried out to the proper motion track
Processing.
Optionally, training unit specifically includes:
Huasdorff is got described normal two-by-two apart from subelement for being trained to the proper motion track
The corresponding Huasdorff distance of motion profile spacing；
Minimal graph divides subelement, described in including several according to preset path spacing expanded range threshold value
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, to the minimal graph partitioning algorithm treated institute
Huasdorff figure is stated to carry out clustering the normal route strategy for determining several normal route data by recursive algorithm.
Optionally, new route determination unit specifically includes:
Space characteristics subelement, for the new route data for getting acquisition to be made whether with 90% new route path point
In the path spacing expanded range, and carry out the new route and the path spacing expanded range of the new route data
Whether Hausdorff distance in intermediate value path is less than any edge path of the new route and the path spacing expanded range
Hausdorff distance, if being to be, it is determined that the new route is normal route, and otherwise the new route is the illegal route；
And/or
Velocity characteristic subelement passes through Gauss point for calculating the new route speed for getting the new route data of acquisition
Cloth carries out modeling processing to the new route speed, and according to the first formulaHorse is used again
Family name's distance determines whether the new route and the normal route of the normal route strategy are similar, if similar, it is determined that institute
Stating new route is normal route, and otherwise the new route is the illegal route；
Wherein, vi' be new route speed, mpIt is then the speed mean value of the normal route strategy, Σ is path velocity point
The covariance matrix of cloth；
And/or
Curvature feature subelement, for calculating the new route speed for the new route data for getting acquisition, new route accelerates
The discontinuous place of degree and new route position passes through the second formulaMeter
New route curvature is calculated, and new route curvature curvature corresponding with the Gaussian Profile being fitted to the normal route strategy is equal
Value carries out the whether similar comparison of mahalanobis distance, if similar, it is determined that the new route is normal route, otherwise the new route
For the illegal route, wherein speed vi', acceleration vi", x' and y' are respectively the first derivative of x and y.
Optionally, further includes:
Monitoring unit, for being monitored to the illegal route.
As can be seen from the above technical solutions, the embodiment of the present invention has the advantage that
A kind of path detection method and device provided in an embodiment of the present invention, wherein path detection method includes: to get
Several normal route data are acquired, and determine proper motion corresponding with normal route data track；To proper motion rail
Mark is trained, and gets corresponding Huasdorff distance, and to including several Huasdorff distance Huasdorff figure
The processing of minimal graph partitioning algorithm is carried out, minimal graph partitioning algorithm treated Huasdorff figure is gathered by recursive algorithm
Class determines the normal route strategy of several normal route data；When getting the new route data of acquisition, in conjunction with normal road
Diameter strategy carries out space characteristics and/or velocity characteristic and/or curvature feature is compared new route data, if new route data
It is unsatisfactory for normal route strategy, then is the illegal route to the corresponding new route of new route data.In the present embodiment, by normal
Motion profile is trained, and gets corresponding Huasdorff distance, and to including several Huasdorff distances
Huasdorff figure carries out the processing of minimal graph partitioning algorithm, and to minimal graph partitioning algorithm, treated that Huasdorff figure passes through recurrence
Algorithm carries out clustering the normal route strategy for determining several normal route data；When getting the new route data of acquisition,
Space characteristics and/or velocity characteristic are carried out in conjunction with normal route strategy and/or curvature feature is compared new route data, if
New route data are unsatisfactory for normal route strategy, then are the illegal route to the corresponding new route of new route data, solve at present
Path detection all based on space characteristics, since the complexity of current path detection algorithm is high, it is low to result in accuracy rate
Technical problem.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention without any creative labor, may be used also for those of ordinary skill in the art
To obtain other attached drawings according to these attached drawings.
Fig. 1 is a kind of flow diagram of one embodiment of path detection method provided in an embodiment of the present invention；
Fig. 2 is a kind of flow diagram of another embodiment of path detection method provided in an embodiment of the present invention；
Fig. 3 is a kind of structural schematic diagram of one embodiment of path detection device provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of another embodiment of path detection device provided in an embodiment of the present invention；
Fig. 5 and Fig. 6 is the application examples schematic diagram of Fig. 2.
Specific embodiment
A kind of path detection method and device provided in an embodiment of the present invention, solve current path detection all with space
Based on feature, since the complexity of current path detection algorithm is high, the low technical problem of accuracy rate is resulted in.
In order to make the invention's purpose, features and advantages of the invention more obvious and easy to understand, below in conjunction with the present invention
Attached drawing in embodiment, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that disclosed below
Embodiment be only a part of the embodiment of the present invention, and not all embodiment.Based on the embodiments of the present invention, this field
Those of ordinary skill's all other embodiment obtained without making creative work, belongs to protection of the present invention
Range.
Referring to Fig. 1, a kind of one embodiment of path detection method provided in an embodiment of the present invention includes:
101, it gets and acquires several normal route data, and determine proper motion corresponding with normal route data
Track；
In the present embodiment, when needing to carry out path detection, it is necessary first to it gets and acquires several normal route data,
And determine proper motion corresponding with normal route data track.
102, proper motion track is trained, gets corresponding Huasdorff distance, and to including several
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, and to minimal graph partitioning algorithm, treated
Huasdorff figure carries out clustering the normal route strategy for determining several normal route data by recursive algorithm；
Several normal route data are acquired when getting, and determine proper motion rail corresponding with normal route data
It after mark, needs to be trained proper motion track, gets corresponding Huasdorff distance, and to including several
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, and to minimal graph partitioning algorithm, treated
Huasdorff figure carries out clustering the normal route strategy for determining several normal route data by recursive algorithm.
103, when getting the new route data of acquisition, space characteristics and/or speed are carried out in conjunction with normal route strategy
Feature and/or curvature feature are compared new route data, if new route data are unsatisfactory for normal route strategy, the road Ze Duixin
The corresponding new route of diameter data is the illegal route.
It is trained when to proper motion track, gets corresponding Huasdorff distance, and to including several
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, and to minimal graph partitioning algorithm, treated
After Huasdorff figure carries out the normal route strategy of several normal route data of cluster determination by recursive algorithm, when obtaining
When getting the new route data of acquisition, space characteristics and/or velocity characteristic and/or curvature feature are carried out in conjunction with normal route strategy
New route data are compared, if new route data are unsatisfactory for normal route strategy, new road corresponding to new route data
Diameter is the illegal route.
In the present embodiment, by being trained to proper motion track, corresponding Huasdorff distance is got, and right
It include that several Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, at minimal graph partitioning algorithm
Huasdorff figure after reason carries out clustering the normal route strategy for determining several normal route data by recursive algorithm；When
When getting the new route data of acquisition, space characteristics and/or velocity characteristic are carried out in conjunction with normal route strategy and/or curvature is special
Sign is compared new route data, corresponding to new route data new if new route data are unsatisfactory for normal route strategy
Path is the illegal route, solves current path detection all based on space characteristics, due to current path detection algorithm
Complexity is high, results in the low technical problem of accuracy rate.
The above is the description carried out to the process of path detection method, will carry out detailed retouch to detailed process below
It states, referring to Fig. 2, a kind of another embodiment of path detection method provided in an embodiment of the present invention includes:
201, several normal route numbers are carried out by several image acquisition devices for being set to several normal routes
According to acquisition；
In the present embodiment, when needing to carry out path detection, it is necessary first to be set to several normal roads by several
The image acquisition device of diameter carries out the acquisition of several normal route data.
202, according in normal route object in the entire visual field of image acquisition device with normal route data phase
Corresponding proper motion track, and mobile mean filter processing is carried out to proper motion track；
Several normal route data are carried out when passing through several image acquisition devices for being set to several normal routes
Acquisition after, need according in normal route object in the entire visual field of image acquisition device with normal route data
Corresponding proper motion track, and mobile mean filter processing is carried out to proper motion track.
203, proper motion track is trained, gets the corresponding Huasdorff of proper motion track spacing two-by-two
Distance；
When object is opposite with normal route data in the entire visual field of image acquisition device in normal route for basis
The proper motion track answered, and proper motion track is carried out after moving mean filter processing, it needs to proper motion track
It is trained, gets the corresponding Huasdorff distance of proper motion track spacing two-by-two.
It 204, include several Huasdorff distances Huasdorff according to preset path spacing expanded range threshold value
Figure carries out the processing of minimal graph partitioning algorithm, passes through recursive algorithm progress to minimal graph partitioning algorithm treated Huasdorff figure
Cluster the normal route strategy for determining several normal route data；
Be trained when to proper motion track, get two-by-two the corresponding Huasdorff of proper motion track spacing away from
It include several according to preset path spacing expanded range threshold value when getting the new route data of acquisition from later
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, and to minimal graph partitioning algorithm, treated
Huasdorff figure carries out clustering the normal route strategy for determining several normal route data by recursive algorithm.
205, when getting the new route data of acquisition, space characteristics and/or speed are carried out in conjunction with normal route strategy
Feature and/or curvature feature are compared new route data, if new route data are unsatisfactory for normal route strategy, the road Ze Duixin
The corresponding new route of diameter data is the illegal route；
When according to preset path spacing expanded range threshold value include several Huasdorff distance Huasdorff figure
The processing of minimal graph partitioning algorithm is carried out, minimal graph partitioning algorithm treated Huasdorff figure is gathered by recursive algorithm
After class determines the normal route strategy of several normal route data, when getting the new route data of acquisition, in conjunction with just
Normal path policy carries out space characteristics and/or velocity characteristic and/or curvature feature is compared new route data, if new route
Data are unsatisfactory for normal route strategy, then are the illegal route to the corresponding new route of new route data.
Specifically, space characteristics and/or velocity characteristic and/or curvature feature are carried out to new route in conjunction with normal route strategy
Data are compared, and are illegal to the corresponding new route of new route data if new route data are unsatisfactory for normal route strategy
Path includes:
It is made whether 90% new route path point in path spacing expanded range the new route data for getting acquisition
In, and carry out new route data new route and path spacing expanded range intermediate value path Hausdorff apart from whether be less than
The Hausdorff distance in any edge path of new route and path spacing expanded range, if being to be, it is determined that new route is
Normal route, otherwise new route is the illegal route；
And/or
The new route speed for getting the new route data of acquisition is calculated, new route speed is built by Gaussian Profile
Mould processing, and according to the first formulaNew route and normal road are determined with mahalanobis distance again
Whether the normal route of diameter strategy is similar, if similar, it is determined that new route is normal route, and otherwise new route is the illegal route；
Wherein, vi' be new route speed, mpIt is then the speed mean value of normal path policy, Σ is path velocity distribution
Covariance matrix；
And/or
Calculate new route speed, new route acceleration and the new route position of the new route data for getting acquisition not
Continuous place passes through the second formulaNew route curvature is calculated, and will be new
Path curvatures curvature mean value corresponding with the Gaussian Profile being fitted to normal route strategy carries out the whether similar ratio of mahalanobis distance
Compared with if similar, it is determined that new route is normal route, and otherwise new route is the illegal route, wherein speed vi', acceleration be
vi", x' and y' are respectively the first derivative of x and y.
206, when new route is the illegal route, then the illegal route is monitored.
When combination normal route strategy carries out space characteristics and/or velocity characteristic and/or curvature feature to new route data
It is compared, is the illegal route to the corresponding new route of new route data if new route data are unsatisfactory for normal route strategy
Later, it needs to be monitored the illegal route.
It is described below with a concrete application scene, as shown in Figure 5 and Figure 6, application examples includes:
It can be applied to various situation, for example, video monitoring.In many public arenas, such as airport, it would be desirable to really
People are determined whether far from some regions, or on some road, if someone is drunk to walk.The system can also be used for instruction one
A new path, if we have found that often thering are many people to walk from some undeveloped route, which can recommend
Give road construction department one new construction route.Path detection is the new problem of comparison, and this patent proposes a kind of simple
, path detection and monitoring method based on multiple features, this method can be very good to distinguish different routing strategies.This is at present
System is mainly used in the monitoring scene of monocular vision, naturally it is also possible in the system of multi-vision visual.
One, path detection result training
Detector can give our motion paths of the object in movement visual field.We use the collected number of fixed camera
According to come the system of training us.According to motion profile of the object in entire visual field, it is assumed that from the i-th frame to n-th frame, object of which movement
Two-dimensional coordinate be
Ti={ (x1,y1),(x2,y2),...,(xn,yn)}
Due to position and the difference of speed, the length in everyone path is different, also, due to passing through detector
It is detecting the result is that with noisy, therefore, we carry out smoothly the path with mobile mean filter.
For the system for training us, we obtain the motion profile of people in the scene.For similar path, we
It is clustered using min-cut nomography recurrence.(figure in complete figure refers to the figure inside graph theory, is a kind of structure in figure.
For image segmentation.In this patent, exactly in order to find suitable path), a kind of each path policy of node on behalf, and it is every
A vertex is connected with other vertex, and obtained entirety is exactly a complete figure.Weight (the power at edge at each edge
Value is exactly two vertex according to the obtained distance value of some algorithm.Indicate the similarity on two vertex.This patent is two
The similarity in path) pass through the Hausdorff distance (a kind of measurement of similarity degree between two groups of point sets of description) on two vertex
It is calculated.For path A and B, Hausdorff distance D (A, B) is,
D (A, B)=max (d (A, B), d (B, A))
Wherein,
Benefit with Hausdorff distance is exactly that it can compare two not homoimerous sequences.So we can
To indicate two paths with the distance.Such as Fig. 5, in Hausdorff figure, if apart from small, weight also certain very little,
It is on the contrary then opposite.By minimal graph partitioning algorithm, which recursive can be divided into two parts by us, wherein each section generation
(path policy is exactly classifier to one group of table identical path policy in fact, and the classifier, this point are given in any path of subsequent input
Class device is according to pervious as a result, predicting whether given path is illegal path.).(minimal graph segmentation be exactly by data according to
Some threshold value is divided into two different parts), the figure of 5 nodes in Fig. 5.The weight at each edge is exactly Hausdorff distance,
Curve indicates the possible minimum cut about figure.Defining a range, (such as Fig. 6, intermediate dotted line are path policy, two realities
Line is then maximum range) as the spatial extention in path.Dynamic given threshold can be passed through.
Two, model of place
Such as the similar scene of some comparisons: scene is similar, scene is similar but speed is different, straight line or curve,
We need to establish a path model to distinguish these.In order to reach such target, we pass through figure partitioning algorithm first
Only conventional path is learnt.After these path learnings are complete, we carry out following three steps operation:
1, space characteristics
2, velocity characteristic
3, curvature feature
Since first step, when only current procedures meet the content of study, just go in next step.In the first step
In, pay close attention to the spatial position of each path policy.The path of each object can be compared with path already existing in database
Compared with.The similarity in space is judged with two conditions: first is the point in the path 90% tested all in the maximum magnitude in path
It is interior.Second is exactly that the Hausdorff distance in intermediate value path in path of the new route compared with is less than two of the path
The Hausdorff distance of rim path.If new route is unsatisfactory for the two conditions, it is abnormal for being considered as new route.
It should be noted that intermediate value path above-mentioned is exactly the path of that in Fig. 6 dotted line.The path of two solid lines is
We according to scene by dividing, and just as any road, solid line is two edges on road, in the road dotted line Ze Shi
Between.
In second step, need to distinguish the motion feature of Different Strategies.If the speed of new route and had existed just
The speed in some path of normal path policy is similar, then it is assumed that new route is similar.The speed P in pathi(xi,yi,ti) calculate
Formula is
The speed in path is modeled using Gaussian Profile, then with mahalanobis distance (the covariance distances of data) come
Judge whether the speed in two paths is similar.
vi' be new route speed, mpIt is then its mean value, Σ is path velocity point
The covariance matrix of cloth.
Third step then needs to find the speed, acceleration of new route and the discontinuous place of position, in this way it may determine that
Whether one people is to take the air line and whether a people walks is the path of mistake.Speed v can be passed throughi' and acceleration vi”
The curvature in path is obtained, calculation formula is as follows:
Wherein x' and y' is respectively the first derivative of x and y.We are exactly that minimal graph segmentation is obtained correct using model
Route result path policy K mean value and variance be fitted a Gaussian Profile.Then, we are by the song of new route
Rate is compared with our Gaussian Profile carries out mahalanobis distance.Pass through comparison, available some improper movements.For example, one
A drunk people walks the road of z font or a people slows down or turns around suddenly.
Therefore, initially, we detect some abnormal conditions by the consistency of space tracking.If space tracking is class
As, we are then compared by velocity characteristic.If velocity characteristic is also similar, finally, we confirm curvature again
Feature.
Object tracking is the most basic function in many systems such as video monitoring, behavioral value.The purpose of this patent is just
It is study object most probable behavior path in video scene, and registers some uncommon behaviors (such as someone is on the scene
Strange behavior path, automobile are run according to z font route, all run on the distance of normal walking in most people in scape).By
In there are some such as pavements, rest area or designed road, so behavior of the vast majority of people in the region is all
According to similar strategy, so can be generalized to other similar region to the algorithm of the regional study, also belonging to this patent
Protection scope.
The present embodiment carries out the detection of abnormal behaviour using space, speed and curvature feature.Abnormal behavior includes:
Path that one people never passes by until then, a people are walked according to habitually in the past different speed in the region or in this scenario
Do not have well-regulated unrest to walk.
In the present embodiment, by being trained to proper motion track, corresponding Huasdorff distance is got, and right
It include that several Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, at minimal graph partitioning algorithm
Huasdorff figure after reason carries out clustering the normal route strategy for determining several normal route data by recursive algorithm；When
When getting the new route data of acquisition, space characteristics and/or velocity characteristic are carried out in conjunction with normal route strategy and/or curvature is special
Sign is compared new route data, corresponding to new route data new if new route data are unsatisfactory for normal route strategy
Path is the illegal route, solves current path detection all based on space characteristics, due to current path detection algorithm
Complexity is high, results in the low technical problem of accuracy rate.
Referring to Fig. 3, a kind of one embodiment of path detection device provided in an embodiment of the present invention includes:
Acquiring unit 301 acquires several normal route data for getting, and determination is opposite with normal route data
The proper motion track answered；
Training unit 302 gets corresponding Huasdorff distance, and right for being trained to proper motion track
It include that several Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, at minimal graph partitioning algorithm
Huasdorff figure after reason carries out clustering the normal route strategy for determining several normal route data by recursive algorithm；
New route determination unit 303, for being carried out in conjunction with normal route strategy when getting the new route data of acquisition
Space characteristics and/or velocity characteristic and/or curvature feature are compared new route data, if new route data are unsatisfactory for normally
Path policy is then the illegal route to the corresponding new route of new route data.
In the present embodiment, by being trained to proper motion track, corresponding Huasdorff distance is got, and right
It include that several Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, at minimal graph partitioning algorithm
Huasdorff figure after reason carries out clustering the normal route strategy for determining several normal route data by recursive algorithm；When
When getting the new route data of acquisition, space characteristics and/or velocity characteristic are carried out in conjunction with normal route strategy and/or curvature is special
Sign is compared new route data, corresponding to new route data new if new route data are unsatisfactory for normal route strategy
Path is the illegal route, solves current path detection all based on space characteristics, due to current path detection algorithm
Complexity is high, results in the low technical problem of accuracy rate.
The above is the description carried out to each unit of path detection device, below interviews sub-unit, please refers to
Fig. 4, a kind of another embodiment of path detection device provided in an embodiment of the present invention include:
Acquiring unit 401 acquires several normal route data for getting, and determination is opposite with normal route data
The proper motion track answered；
Acquiring unit 401 specifically includes:
Subelement 4011 is obtained, if the image acquisition device for being set to several normal routes by several carries out
The acquisition of dry normal route data；
Handle subelement 4012, for according in normal route object in the entire visual field of image acquisition device with
The corresponding proper motion track of normal route data, and mobile mean filter processing is carried out to proper motion track.
Training unit 402 gets corresponding Huasdorff distance, and right for being trained to proper motion track
It include that several Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, at minimal graph partitioning algorithm
Huasdorff figure after reason carries out clustering the normal route strategy for determining several normal route data by recursive algorithm；
Training unit 402 specifically includes:
Huasdorff gets proper motion two-by-two for being trained to proper motion track apart from subelement 4021
The corresponding Huasdorff distance of track spacing；
Minimal graph divides subelement 4022, for including several according to preset path spacing expanded range threshold value
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, and to minimal graph partitioning algorithm, treated
Huasdorff figure carries out clustering the normal route strategy for determining several normal route data by recursive algorithm.
New route determination unit 403, for being carried out in conjunction with normal route strategy when getting the new route data of acquisition
Space characteristics and/or velocity characteristic and/or curvature feature are compared new route data, if new route data are unsatisfactory for normally
Path policy is then the illegal route to the corresponding new route of new route data.
New route determination unit 403 specifically includes:
Space characteristics subelement 4031, for the new route data for getting acquisition to be made whether with 90% new route road
Diameter point is in path spacing expanded range, and carries out the new route of new route data and the intermediate value path of path spacing expanded range
Whether Hausdorff distance is less than the Hausdorff distance in any edge path of new route and path spacing expanded range, if
It is to be, it is determined that new route is normal route, and otherwise new route is the illegal route；
And/or
Velocity characteristic subelement 4032 passes through height for calculating the new route speed for getting the new route data of acquisition
This distribution carries out modeling processing to new route speed, and according to the first formulaHorse is used again
Family name's distance determines whether new route is similar to the normal route of normal route strategy, if similar, it is determined that new route is normal road
Diameter, otherwise new route is the illegal route；
Wherein, vi' be new route speed, mpIt is then the speed mean value of normal path policy, Σ is path velocity distribution
Covariance matrix；
And/or
Curvature feature subelement 4033, for calculating the new route speed for the new route data for getting acquisition, new route adds
The discontinuous place of speed and new route position passes through the second formula
New route curvature is calculated, and new route curvature curvature mean value corresponding with the Gaussian Profile being fitted to normal route strategy is carried out
The whether similar comparison of mahalanobis distance, if similar, it is determined that new route is normal route, and otherwise new route is the illegal route,
In, speed vi', acceleration vi", x' and y' are respectively the first derivative of x and y.
Monitoring unit 404, for being monitored to the illegal route.
In the present embodiment, by being trained to proper motion track, corresponding Huasdorff distance is got, and right
It include that several Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, at minimal graph partitioning algorithm
Huasdorff figure after reason carries out clustering the normal route strategy for determining several normal route data by recursive algorithm；When
When getting the new route data of acquisition, space characteristics and/or velocity characteristic are carried out in conjunction with normal route strategy and/or curvature is special
Sign is compared new route data, corresponding to new route data new if new route data are unsatisfactory for normal route strategy
Path is the illegal route, solves current path detection all based on space characteristics, due to current path detection algorithm
Complexity is high, results in the low technical problem of accuracy rate.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided herein, it should be understood that disclosed system, device and method can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components
It can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, it is shown or
The mutual coupling, direct-coupling or communication connection discussed can be through some interfaces, the indirect coupling of device or unit
It closes or communicates to connect, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the present invention
Portion or part steps.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can store journey
The medium of sequence code.
The above, the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although referring to before
Stating embodiment, invention is explained in detail, those skilled in the art should understand that: it still can be to preceding
Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features；And these
It modifies or replaces, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (4)
1. a kind of path detection method characterized by comprising
It gets and acquires several normal route data, and determine proper motion rail corresponding with the normal route data
Mark；
The proper motion track is trained, gets corresponding Huasdorff distance, and including described in several
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, to the minimal graph partitioning algorithm treated institute
Huasdorff figure is stated to carry out clustering the normal route strategy for determining several normal route data by recursive algorithm；
When getting the new route data of acquisition, space characteristics and/or velocity characteristic are carried out in conjunction with the normal route strategy
And/or curvature feature is compared the new route data, if the new route data are unsatisfactory for the normal route strategy,
It is then the illegal route to the corresponding new route of the new route data；
It gets and acquires several normal route data, and determine proper motion track corresponding with the normal route data
It specifically includes:
Several normal route data are carried out by several image acquisition devices for being set to several normal routes
Acquisition；
According in the normal route object in the entire visual field of the image acquisition device with the normal route number
Mobile mean filter processing is carried out according to corresponding proper motion track, and to the proper motion track；
The proper motion track is trained, gets corresponding Huasdorff distance, and including described in several
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, to the minimal graph partitioning algorithm treated institute
It is specific to state the normal route strategy that Huasdorff figure cluster several determining normal route data by recursive algorithm
Include:
The proper motion track is trained, the corresponding Huasdorff of proper motion track spacing two-by-two is got
Distance；
According to preset path spacing expanded range threshold value include several described Huasdorff distance Huasdorff figures into
The processing of row minimal graph partitioning algorithm, to the minimal graph partitioning algorithm, treated that the Huasdorff figure passes through recursive algorithm
It carries out clustering the normal route strategy for determining several normal route data；
When getting the new route data of acquisition, space characteristics and/or velocity characteristic are carried out in conjunction with the normal route strategy
And/or curvature feature is compared the new route data, if the new route data are unsatisfactory for the normal route strategy,
Then the corresponding new route of the new route data is specifically included for the illegal route:
It is made whether 90% new route path point in the path spacing expanded range new route data for getting acquisition
In, and carry out the new route of the new route data and the intermediate value path Hausdorff distance of the path spacing expanded range
Whether the Hausdorff distance in any edge path of the new route and the path spacing expanded range is less than, if being
It is, it is determined that the new route is normal route, and otherwise the new route is the illegal route；
And/or
The new route speed for getting the new route data of acquisition is calculated, the new route speed is built by Gaussian Profile
Mould processing, and according to the first formulaThe new route and institute are determined with mahalanobis distance again
Whether the normal route for stating normal route strategy is similar, if similar, it is determined that the new route is normal route, otherwise institute
Stating new route is the illegal route；
Wherein, vi' be new route speed, mpIt is then the speed mean value of the normal route strategy, ∑ is path velocity distribution
Covariance matrix；
And/or
Calculate the discontinuous of the new route speed of new route data for getting acquisition, new route acceleration and new route position
Place passes through the second formulaNew route curvature is calculated, and will be described new
Whether path curvatures curvature mean value corresponding with the Gaussian Profile being fitted to the normal route strategy carries out mahalanobis distance similar
Comparison, if similar, it is determined that the new route is normal route, and otherwise the new route is the illegal route, wherein speed is
vi', acceleration vi", x' and y' are respectively the first derivative of x and y.
2. path detection method according to claim 1, which is characterized in that the path detection method further include:
When the new route is the illegal route, then the illegal route is monitored.
3. a kind of path detection device characterized by comprising
Acquiring unit acquires several normal route data for getting, and determination is corresponding with the normal route data
Proper motion track；
Training unit gets corresponding Huasdorff distance, and to packet for being trained to the proper motion track
Scheme to carry out the processing of minimal graph partitioning algorithm containing Huasdorff distance Huasdorff described in several, the minimal graph is divided
Huasdorff figure after algorithm process, which cluster by recursive algorithm, is determining several described normal route data just
Normal path policy；
New route determination unit, for being carried out in conjunction with the normal route strategy empty when getting the new route data of acquisition
Between feature and/or velocity characteristic and/or curvature feature the new route data are compared, if the new route data are discontented
The foot normal route strategy is then the illegal route to the corresponding new route of the new route data；
Acquiring unit specifically includes:
Subelement is obtained, for carrying out described in several by several image acquisition devices for being set to several normal routes
The acquisition of normal route data；
Handle subelement, for according in the normal route object in the entire visual field of the image acquisition device with
The corresponding proper motion track of the normal route data, and the proper motion track is carried out at mobile mean filter
Reason；
Training unit specifically includes:
Huasdorff gets the proper motion two-by-two for being trained to the proper motion track apart from subelement
The corresponding Huasdorff distance of track spacing；
Minimal graph divides subelement, described in including several according to preset path spacing expanded range threshold value
Huasdorff distance Huasdorff figure carries out the processing of minimal graph partitioning algorithm, to the minimal graph partitioning algorithm treated institute
Huasdorff figure is stated to carry out clustering the normal route strategy for determining several normal route data by recursive algorithm；
New route determination unit specifically includes:
Space characteristics subelement, for being made whether 90% new route path point in institute to the new route data for getting acquisition
It states in path spacing expanded range, and carries out the new route of the new route data and the intermediate value of the path spacing expanded range
Whether Hausdorff distance in path is less than any edge path of the new route and the path spacing expanded range
Hausdorff distance, if being to be, it is determined that the new route is normal route, and otherwise the new route is the illegal route；
And/or
Velocity characteristic subelement passes through Gaussian Profile pair for calculating the new route speed for getting the new route data of acquisition
The new route speed carries out modeling processing, and according to the first formulaAgain with geneva away from
It is whether similar from the determination new route and the normal route of the normal route strategy, if similar, it is determined that described new
Path is normal route, and otherwise the new route is the illegal route；
Wherein, v 'iFor the speed of new route, mpIt is then the speed mean value of the normal route strategy, ∑ is path velocity distribution
Covariance matrix；
And/or
Curvature feature subelement, for calculate new route speed, the new route acceleration of the new route data for getting acquisition with
And the discontinuous place of new route position passes through the second formulaIt calculates new
Path curvatures, and by new route curvature curvature mean value corresponding with the Gaussian Profile being fitted to the normal route strategy into
The whether similar comparison of row mahalanobis distance, if similar, it is determined that the new route is normal route, and otherwise the new route is non-
Method path, wherein speed is v 'i, acceleration be v "i, x' and y' are respectively the first derivative of x and y.
4. path detection device according to claim 3, which is characterized in that further include:
Monitoring unit, for being monitored to the illegal route.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611153636.5A CN106598856B (en) | 2016-12-14 | 2016-12-14 | A kind of path detection method and device |
| PCT/CN2017/093483 WO2018107748A1 (en) | 2016-12-14 | 2017-07-19 | Path detection method and apparatus |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611153636.5A CN106598856B (en) | 2016-12-14 | 2016-12-14 | A kind of path detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106598856A CN106598856A (en) | 2017-04-26 |
| CN106598856B true CN106598856B (en) | 2019-03-01 |
Family
ID=58802837
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611153636.5A Active CN106598856B (en) | 2016-12-14 | 2016-12-14 | A kind of path detection method and device |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN106598856B (en) |
| WO (1) | WO2018107748A1 (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106598856B (en) * | 2016-12-14 | 2019-03-01 | 威创集团股份有限公司 | A kind of path detection method and device |
| CN109241291B (en) * | 2018-07-18 | 2022-02-15 | 华南师范大学 | Knowledge graph optimal path query system and method based on deep reinforcement learning |
| CN110175633B (en) * | 2019-04-29 | 2020-12-15 | 杭州电子科技大学 | Fishing boat track abnormity detection method based on Beidou satellite transmission data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103517042A (en) * | 2013-10-17 | 2014-01-15 | 吉林大学 | Nursing home old man dangerous act monitoring method |
| CN103854027A (en) * | 2013-10-23 | 2014-06-11 | 北京邮电大学 | Crowd behavior identification method |
| CN103942533A (en) * | 2014-03-24 | 2014-07-23 | 河海大学常州校区 | Urban traffic illegal behavior detection method based on video monitoring system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8855361B2 (en) * | 2010-12-30 | 2014-10-07 | Pelco, Inc. | Scene activity analysis using statistical and semantic features learnt from object trajectory data |
| CN106598856B (en) * | 2016-12-14 | 2019-03-01 | 威创集团股份有限公司 | A kind of path detection method and device |
-
2016
- 2016-12-14 CN CN201611153636.5A patent/CN106598856B/en active Active
-
2017
- 2017-07-19 WO PCT/CN2017/093483 patent/WO2018107748A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103517042A (en) * | 2013-10-17 | 2014-01-15 | 吉林大学 | Nursing home old man dangerous act monitoring method |
| CN103854027A (en) * | 2013-10-23 | 2014-06-11 | 北京邮电大学 | Crowd behavior identification method |
| CN103942533A (en) * | 2014-03-24 | 2014-07-23 | 河海大学常州校区 | Urban traffic illegal behavior detection method based on video monitoring system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106598856A (en) | 2017-04-26 |
| WO2018107748A1 (en) | 2018-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mínguez et al. | Pedestrian path, pose, and intention prediction through gaussian process dynamical models and pedestrian activity recognition | |
| Wang et al. | Pedestrian recognition and tracking using 3D LiDAR for autonomous vehicle | |
| CN110378281A (en) | Group Activity recognition method based on pseudo- 3D convolutional neural networks | |
| US10037607B2 (en) | Topology determination for non-overlapping camera network | |
| CN106570453B (en) | Method, device and system for pedestrian detection | |
| CN112639873A (en) | Multi-object pose tracking device and method based on single-object pose estimator | |
| CN106774312B (en) | Method and device for determining moving track of mobile robot | |
| US9892326B2 (en) | Object detection in crowded scenes using context-driven label propagation | |
| CN102378992A (en) | Articulated region detection device and method for same | |
| CN106598856B (en) | A kind of path detection method and device | |
| CN107958260B (en) | Group behavior analysis method based on multi-feature fusion | |
| CN110264495A (en) | A kind of method for tracking target and device | |
| Murphy et al. | Incremental unsupervised topological place discovery | |
| CN114222986A (en) | Random trajectory prediction using social graph networks | |
| Henrio et al. | Anomaly detection in videos recorded by drones in a surveillance context | |
| Jingying | A survey on crowd counting methods and datasets | |
| Henschel et al. | Efficient multiple people tracking using minimum cost arborescences | |
| Akhter et al. | Abnormal action recognition in crowd scenes via deep data mining and random forest | |
| Mohamed et al. | Behavior analysis using a multilevel motion pattern learning framework | |
| CN116630888A (en) | Unmanned aerial vehicle monitoring method, unmanned aerial vehicle monitoring device, electronic equipment and storage medium | |
| Molina-Cabello et al. | Vehicle classification in traffic environments using the growing neural gas | |
| CN106991384B (en) | Pedestrian's group Activity recognition method and device | |
| Patino et al. | Multicamera trajectory analysis for semantic behaviour characterisation | |
| Zhang et al. | Online social behavior modeling for multi-target tracking | |
| Paduraru et al. | Pedestrian motion in simulation applications using deep learning |
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 | ||
| CB02 | Change of applicant information |
Address after: Kezhu road high tech Industrial Development Zone, Guangzhou city of Guangdong Province, No. 233 510670 Applicant after: Wei Chong group Limited by Share Ltd Address before: Kezhu road high tech Industrial Development Zone, Guangzhou city of Guangdong Province, No. 233 510670 Applicant before: Guangdong Weichuangshixun Science and Technology Co., Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |