CN107610177B - The method and apparatus of characteristic point is determined in a kind of synchronous superposition - Google Patents
The method and apparatus of characteristic point is determined in a kind of synchronous superposition Download PDFInfo
- Publication number
- CN107610177B CN107610177B CN201710909238.XA CN201710909238A CN107610177B CN 107610177 B CN107610177 B CN 107610177B CN 201710909238 A CN201710909238 A CN 201710909238A CN 107610177 B CN107610177 B CN 107610177B
- Authority
- CN
- China
- Prior art keywords
- point
- information
- spatial position
- characteristic features
- position change
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The embodiment of the invention provides the methods and apparatus that characteristic point is determined in a kind of synchronous superposition, the method comprise the steps that acquiring image information using image acquisition equipment;Image information based on acquisition obtains the fisrt feature point set in described image information;Analyze the non-static characteristic point in the fisrt feature point set, wherein the non-static characteristic point is the corresponding characteristic point of mobile object;The non-static characteristic point is filtered, in the fisrt feature point set to generate second feature point set.The present invention can simply and easily remove non-static characteristic point, improve the accuracy and speed of map structuring.
Description
Technical field
The present embodiments relate to the field of cloud map structuring, in particular to determined in a kind of synchronous superposition
The method and apparatus of characteristic point.
Background technique
Synchronizing and being positioned at the core of map structuring is extracted by camera collection image, and by feature point extraction algorithm
Characteristic point in image, then with characteristic point and key frame building point cloud map, subsequent positioning is then completed by point cloud and is led
Boat.Thus characteristic point be V-SLAM map basic element, a good characteristic point should have be it is static, can detect again.
And in actual scene, existing a large amount of non-static characteristic points will affect a little if not rejecting these non-static characteristic points
The precision of cloud.V-SLAM, using strategies such as various optimizations, consumes a large amount of computing resource to reject these characteristic points.Cause
This, if it is possible to it proposes a kind of method that can easily extract static nature point, the building speed of cloud map can be optimized.
Summary of the invention
The embodiment of the invention provides one kind being capable of a kind of synchronous positioning that is simple and convenient and can optimizing resource and map structure
The method and apparatus for building middle determining characteristic point.
In order to solve the above-mentioned technical problem, the embodiment of the invention provides following technical solutions:
The method of characteristic point is determined in a kind of synchronous superposition comprising:
Image information is acquired using image acquisition equipment;
Image information based on acquisition obtains the fisrt feature point set in described image information;
Analyze the non-static characteristic point in the fisrt feature point set, wherein the non-static characteristic point is motive objects
The corresponding characteristic point of body;
The non-static characteristic point is filtered, in the fisrt feature point set to generate second feature point set.
In a preferred embodiment, the non-static characteristic point in the analysis fisrt feature point set includes:
The fisrt feature point set in the described image information of different moments is obtained respectively;
Determine the same characteristic features point in the fisrt feature point set at each moment;
Equipment is obtained in the movement state information minute at each moment based on the determining same characteristic features point and described image
Analyse the non-static characteristic point.
In a preferred embodiment, described to be existed based on the determining same characteristic features point and described image acquisition equipment
The movement state information at each moment analyzes the non-static characteristic point
Based on the depth information of the determining same characteristic features point, the spatial position change letter of the same characteristic features point is calculated
Breath;
Obtain the spatial position change information that described image obtains equipment;
The spatial position that spatial position change information and described image based on the same characteristic features point obtain equipment becomes
Change information, analysis determines the non-static characteristic point.
In a preferred embodiment, the spatial position change information based on the same characteristic features point and the figure
Spatial position change information as obtaining equipment, analysis determine that the non-static characteristic point includes:
The spatial position change information that equipment is obtained based on described image, to the spatial position change of the same characteristic features point
Operation is normalized in information, to obtain normed space change in location information;
Based on the normed space change in location information, cluster operation is carried out to the same characteristic features point, described in obtaining
Non-static characteristic point.
In a preferred embodiment, the depth information based on the determining same characteristic features point calculates described identical
The spatial position change information of characteristic point includes:
The depth information of the same characteristic features point based on each moment, determines the location information of the same characteristic features point;
The spatial position change information is determined based on the difference between the location information of each same characteristic features point at each moment.
In a preferred embodiment, the spatial position change information that equipment is obtained based on described image, to the phase
Operation is normalized in spatial position change information with characteristic point, includes: to obtain normed space change in location information
Described image is obtained to the spatial position change information of equipment and the spatial position change letter of same characteristic features point
Breath carries out difference operation, obtains the normed space change in location information.
In a preferred embodiment, described to be based on the normed space change in location information, the same characteristic features are clicked through
Row cluster operation includes: to obtain the non-static characteristic point
Using the normed space change in location information as core, cluster operation is carried out to the same characteristic features point;
Calculate the sum of the class spacing of all kinds of same characteristic features points;
The biggish a kind of characteristic point of the sum of the class spacing is judged as non-static characteristic point.
A kind of equipment the embodiment of the invention also provides characteristic point is determined in synchronous superposition comprising:
Image collection module is configured to acquire the image information of different moments respectively;
Processor is configured to the image information of acquisition, obtains the fisrt feature point set in described image information;
Analyze the non-static characteristic point in the fisrt feature point set, wherein the non-static characteristic point is that mobile object is corresponding
Characteristic point;The non-static characteristic point is filtered, in the fisrt feature point set to generate second feature point set.
In a preferred embodiment, the processor is further configured to obtain the described image information of different moments respectively
In fisrt feature point set;Determine the same characteristic features point in the fisrt feature point set at each moment;Based on determining institute
It states same characteristic features point and described image obtains movement state information of the equipment at each moment and analyzes the non-static characteristic point.
In a preferred embodiment, the processor is further configured to the depth based on the determining same characteristic features point
Information calculates the spatial position change information of the same characteristic features point;
Obtain the spatial position change information that described image obtains equipment;
The spatial position that spatial position change information and described image based on the same characteristic features point obtain equipment becomes
Change information, analysis determines the non-static characteristic point.
Based on disclosed above, can know the embodiment of the present invention have it is following the utility model has the advantages that
1, it since method provided in an embodiment of the present invention can directly remove non-static characteristic point, then can effectively reduce
To the consumption of computing resource when SLAM back-end processing, improves point cloud and construct speed;
2, in position fixing process, cloud map construction device can determine the current location of itself by multipoint positioning, non-
The presence of static nature point will increase the probability of erroneous matching, lead to the reduction of positioning accuracy, and the embodiment of the present invention can be in structure
Cloud of laying foundations before is filtered characteristic point, removes behavioral characteristics point therein, significantly improves positioning accuracy;
3, to the filtering of behavioral characteristics point, the sparsity of some clouds can be improved, and does not influence the positioning accuracy of a cloud, together
When in positioning stage, can significantly reduce the calculation amount of positioning, promote locating speed.
Detailed description of the invention
Fig. 1 is the principle stream that the method for characteristic point is determined in one of embodiment of the present invention synchronous superposition
Cheng Tu;
Fig. 2 is the process that the method for the non-static characteristic point in the fisrt feature point set is analyzed in the embodiment of the present invention
Figure;
Fig. 3 is to obtain equipment each based on the determining same characteristic features point and described image in the embodiment of the present invention
The movement state information at moment analyzes the method flow diagram of the non-static characteristic point;
Fig. 4 is the method flow diagram that non-static characteristic point is further determined that in the embodiment of the present invention;
Fig. 5 is the Method And Principle flow chart for determining non-static characteristic point in the embodiment of the present invention by clustering processing analysis;
Fig. 6 be the embodiment of the present invention in synchronous superposition in determine characteristic point equipment theory structure
Figure;
Fig. 7 is the theory structure schematic diagram of the processor in the embodiment of the present invention.
Specific embodiment
In the following, specific embodiments of the present invention are described in detail in conjunction with attached drawing, but not as the limitation of the invention.
It should be understood that various modifications can be made to disclosed embodiments.Therefore, description above should not regard
To limit, and only as the example of embodiment.Those skilled in the art will expect within the scope and spirit of this
Other modifications.
The attached drawing being included in the description and forms part of the description shows embodiment of the disclosure, and with it is upper
What face provided is used to explain the disclosure together to substantially description and the detailed description given below to embodiment of the disclosure
Principle.
It is of the invention by the description of the preferred form with reference to the accompanying drawings to the embodiment for being given as non-limiting example
These and other characteristic will become apparent.
Although being also understood that invention has been described referring to some specific examples, those skilled in the art
Member realizes many other equivalents of the invention in which can determine, they have feature as claimed in claim and therefore all
In the protection scope defined by whereby.
When read in conjunction with the accompanying drawings, in view of following detailed description, above and other aspect, the feature and advantage of the disclosure will become
It is more readily apparent.
The specific embodiment of the disclosure is described hereinafter with reference to attached drawing;It will be appreciated, however, that the disclosed embodiments are only
Various ways implementation can be used in the example of the disclosure.Known and/or duplicate function and structure and be not described in detail to avoid
Unnecessary or extra details makes the disclosure smudgy.Therefore, specific structural and functionality disclosed herein is thin
Section is not intended to restrictions, but as just the basis of claim and representative basis be used to instructing those skilled in the art with
Substantially any appropriate detailed construction diversely uses the disclosure.
This specification can be used phrase " in one embodiment ", " in another embodiment ", " in another embodiment
In " or " in other embodiments ", it can be referred to one or more of the identical or different embodiment according to the disclosure.
In the following, the embodiment of the present invention is described in detail in conjunction with attached drawing, the embodiment of the invention provides a kind of synchronous positioning with
The method that characteristic point is determined in map structuring, this method can apply the extraction in V-SLAM map for static nature point
Cheng Zhong can greatly promote the speed and precision of component cloud map by the non-static characteristic point in removal image characteristic point.Its
In, synchronous superposition is (SLAM or Simultaneous localization and mapping).
As shown in Figure 1, for the method for determining characteristic point in one of embodiment of the present invention synchronous superposition
Principle flow chart, synchronizing in the embodiment of the present invention be positioned in map structuring the method for determining characteristic point and may include:
Image information is acquired using image acquisition equipment;
Image information based on acquisition obtains the fisrt feature point set in described image information;
Analyze the non-static characteristic point in the fisrt feature point set, wherein the non-static characteristic point is motive objects
The corresponding characteristic point of body;
The non-static characteristic point is filtered, in the fisrt feature point set to generate second feature point set.
In the embodiment of the present invention, the technical field in map structuring can be applied, wherein can obtain first with image
Equipment acquires image information, which may include depth camera or cameras people etc., be obtained by image
Take equipment that can acquire image information, which can be the image information with depth information.Meanwhile it being obtained in image
Also there is corresponding temporal information, i.e., each image is associated with corresponding temporal information in image information acquired in equipment.
In addition, in the embodiment of the present invention after the image information for obtaining each moment, image can be obtained based on the image information of acquisition
Fisrt feature point set in information, it can feature point extraction is carried out to image information according to preset algorithm, it includes more for obtaining
The fisrt feature point set of a characteristic point.Wherein, the preset algorithm of the embodiment of the present invention may include that the detection of FAST characteristic point is calculated
Method or SIFT (scale invariant feature conversion) algorithm can execute the operation of features described above point extraction by above-mentioned algorithm.Example
Such as, the Tm moment is acquired, the image information f (t) with depth information, which can be indicated with RGB data, by above-mentioned pre-
Imputation method (FAST/SIFT) extracts the fisrt feature point set P ∈ { pm0, pm1, pm2 ... pmn } in image information, wherein collecting
Each characteristic point closed in P can be indicated with location information, such as pixel coordinate information.Preferably, also when extracting characteristic point
The characterization information about characteristic point each in fisrt feature point set can be generated, this feature description can be expressed as D ∈
{ dm0, dm1, dm2 ... dmn }, this feature description information may include the characteristic information of characteristic point, with for identification and distinguish
Each characteristic point.After obtaining above-mentioned fisrt feature point set and the description of corresponding feature, can it is corresponding with time Tm into
Row storage.The fisrt feature point set of image information based on the available each moment acquisition of the above process and corresponding spy
Levy the characterization information of point.
After obtaining the fisrt feature point set of image information at each moment, each fisrt feature point set can be divided
Analysis, to identify the non-static characteristic point in fisrt feature point set.Wherein, the non-static characteristic point is that mobile object is corresponding
Characteristic point.Also, the non-static feature can be filtered in the fisrt feature point set after identifying non-static characteristic point
Point, to generate second feature point set.Each characteristic point in the second feature point set is static nature point.Utilize each moment
Second feature point set used as map structuring, the accuracy and speed of the process can be improved.
It can directly be removed due to method provided in an embodiment of the present invention non-static in the image information that each moment obtains
Characteristic point, then consumption when can effectively reduce SLAM back-end processing to computing resource improve point cloud and construct speed;In addition exist
In position fixing process, cloud map construction device can determine the current location of itself by multipoint positioning, non-static characteristic point
In the presence of the probability that will increase erroneous matching, lead to the reduction of positioning accuracy, the embodiment of the present invention can be right before building point cloud
Characteristic point is filtered, and is removed behavioral characteristics point therein, is significantly improved positioning accuracy;Filtering to behavioral characteristics point simultaneously,
The sparsity of some clouds can be improved, and not influence the positioning accuracy of a cloud, while in positioning stage, can significantly reduce positioning
Calculation amount, promoted locating speed.
Further, as shown in Fig. 2, to analyze the non-static spy in the fisrt feature point set in the embodiment of the present invention
The flow chart of the method for point is levied, which may include:
The fisrt feature point set in the described image information of different moments is obtained respectively;
Determine the same characteristic features point in the fisrt feature point set at each moment;
Equipment is obtained in the movement state information minute at each moment based on the determining same characteristic features point and described image
Analyse the non-static characteristic point.
As described in above-described embodiment, by the image information of image acquisition equipment available different moments, and being based on should
Image information can further extract fisrt feature point set therein, therefore, the image information institute at available each moment
Corresponding fisrt feature point set, such as the fisrt feature point set at Tm moment and the fisrt feature point set at Tm+1 moment.It is obtaining
The same characteristic features point that can also be obtained after the fisrt feature point set at each moment after P in each fisrt feature point set is taken, such as may be used
It is clicked through with the feature in the fisrt feature point set of each characteristic point and Tm moment in the fisrt feature point set by the Tm+1 moment
Row characteristic point Similarity measures find out matched characteristic point, i.e. same characteristic features point in Tm and Tm+1 moment.Here, the present invention is real
Each fisrt feature can be determined according to for characterization information corresponding to each characteristic point in fisrt feature point set by applying example
Same characteristic features point in point set, since the characterization information for characteristic point is the letter that can indicate characteristic point stationary state
It ceases (information that will not generally change), therefore the matched characteristic point of characterization information can be judged as identical feature
Point.
Further, it after the same characteristic features point in the fisrt feature point set for judging each moment, is also based on really
The fixed same characteristic features point and described image obtain movement state information of the equipment at each moment and analyze the non-static spy
Sign point.
It is illustrated in figure 3 in the embodiment of the present invention and is set based on the determining same characteristic features point and described image acquisition
The standby movement state information at each moment analyzes the method flow diagram of the non-static characteristic point, wherein institute in the embodiment of the present invention
It states and equipment is obtained described in the movement state information analysis at each moment based on the determining same characteristic features point and described image
Non-static characteristic point may include:
Based on the depth information of the determining same characteristic features point, the spatial position change letter of the same characteristic features point is calculated
Breath;
Obtain the spatial position change information that described image obtains equipment;
The spatial position that spatial position change information and described image based on the same characteristic features point obtain equipment becomes
Change information, analysis determines the non-static characteristic point.
It in embodiments of the present invention, can be based on determination after determining the same characteristic features point in each fisrt feature point set
Same characteristic features point depth information, calculate same characteristic features point spatial position change information.
As described above, the image acquisition equipment in the embodiment of the present invention can obtain the letter of the image with depth information
It ceases, each characteristic point in the fisrt feature point set of extraction also can have depth information, then herein can be based on depth
The location information (such as coordinate information) of each characteristic point of information, determines the spatial position change information of each characteristic point.
Specifically, based on the depth information of the determining same characteristic features point described in the embodiment of the present invention, described in calculating
The spatial position change information of same characteristic features point may include:
The depth information of the same characteristic features point based on each moment, determines the location information of the same characteristic features point;
The spatial position change information is determined based on the difference between the location information of each same characteristic features point at each moment.
That is, the location information of each same characteristic features point can be determined based on the depth information of each same characteristic features point, it should
Location information is indicated with space coordinate, later can be by the coordinate value of each same characteristic features point at Tm+1 moment and Tm moment
The coordinate value of each same characteristic features point carries out the difference operation of space vector, to obtain the spatial position change of each same characteristic features point
Information.For example, Tm the and Tm+1 moment can be calculated based on the intrinsic parameter of the camera of image acquisition equipment, such as depth information
The coordinate value (Xm, Ym, Zm) and (Xm+1, Ym+1, Zm+1) under coordinate system of each same characteristic features point, then the Tm+1 moment relative to
The spatial variations information of the same characteristic features point at Tm moment can be expressed as space vector (Xm+1-Xm, Ym+1-Ym, Zm+1-Zm).
In addition, since image acquisition equipment is during executing image acquisition, the position of camera, orientation, angle
Changes will occur, and the image acquisition equipment in the embodiment of the present invention can also obtain it in the mobile speed of different moments in real time
The parameter informations such as degree, direction, angle, location information, and the spatial position change information of corresponding determining image acquisition equipment.Example
Such as available Tm to Tm+1 moment code-disc (Odometry) data (v, θ), which indicates that image obtains in 1 time interval
The speed and drift angle for taking the camera of equipment mobile.That is, in the embodiment of the present invention also it is available at various moments
The location information (being indicated using space coordinate) of image acquisition equipment, and the relative position obtained between each moment can also be corresponded to
Variation, i.e. spatial position change information.For example, calculating the location information of Tm and Tm+1 moment image acquisition equipment, such as sitting
Respectively (Xc, Yc, Zc) and (Xc+1, Yc+1, Zc+1), then the Tm+1 moment is identical relative to the Tm moment for the lower coordinate value of mark system
The spatial variations information of characteristic point can be expressed as space vector (Xc+1-Xc, Yc+1-Yc, Zc+1-Zc).
In the sky for the spatial position change information and each same characteristic features point for obtaining the image acquisition equipment between different moments
Between position change information when, can determine non-static characteristic point based on above- mentioned information.
It is illustrated in figure 4 the method flow diagram that non-static characteristic point is further determined that in the embodiment of the present invention, wherein this hair
In bright embodiment, the spatial position change information and described image based on the same characteristic features point obtains the sky of equipment
Between position change information, analysis determines that the non-static characteristic point includes:
The spatial position change information that equipment is obtained based on described image, to the spatial position change of the same characteristic features point
Operation is normalized in information, to obtain normed space change in location information;
Based on the normed space change in location information, cluster operation is carried out to the same characteristic features point, described in obtaining
Non-static characteristic point.
In the sky for the spatial position change information and each same characteristic features point for obtaining the image acquisition equipment between different moments
Between position change information when, can based on described image obtain equipment spatial position change information, to the same characteristic features point
Spatial position change information operation is normalized, the normalization operation may include: by described image obtain equipment sky
Between the spatial position change information of position change information and same characteristic features point carry out difference operation, obtain the normed space
Change in location information.
Can by synchronization calculated image acquisition equipment spatial position change information and the moment phase
Spatial position change information with characteristic point carries out difference operation, such as can calculate image of the Tm+1 moment relative to the Tm moment
The spatial position change information for obtaining equipment is (Xc+1-Xc, Yc+1-Yc, Zc+1-Zc), and the space bit of same characteristic features point
Setting change information is (Xm+1-Xm, Ym+1-Ym, Zm+1-Zm), then above-mentioned for the Tm+1 moment, the normalization of each same characteristics is grasped
The result of work is the difference between (Xc+1-Xc, Yc+1-Yc, Zc+1-Zc) and (Xm+1-Xm, Ym+1-Ym, Zm+1-Zm), should
Difference can be expressed as normed space change in location information.
It is illustrated with following formula, (Lx, Ly, Lz) can be expressed as standard control change in location information, (Xc, Yc,
Zc) indicate that the spatial position change information of image acquisition equipment, and (Xm+1, Ym+1, Zm+1) can be expressed as same characteristic features
The spatial position change information of point, then the relationship between three can be with are as follows:
It is possible to further carry out cluster behaviour to the same characteristic features point based on the normed space change in location information
Make, to obtain the non-static characteristic point.As shown in figure 5, to be non-quiet by clustering processing analysis determination in the embodiment of the present invention
The Method And Principle flow chart of state characteristic point, wherein it is described to be based on the normed space change in location information, to the same characteristic features
Point carries out cluster operation, may include: to obtain the non-static characteristic point
Using the normed space change in location information as core, cluster operation is carried out to the same characteristic features point
Calculate the sum of the class spacing of all kinds of same characteristic features points;
The big a kind of same characteristic features point of the sum of described class spacing is judged as non-static characteristic point.
In the embodiment of the present invention, using the k-means clustering method of k=2, circulating repetition K time interval, to same identical
The standard control change in location information of characteristic point P carries out clustering, wherein can be respectively by the same characteristic features point at a certain moment
Normed space change in location information as feature vector, to the normed space change in location information of all same characteristic features point P into
The sum of row clustering processing, and calculate between class distance, the feature point set for taking the sum of between class distance small is combined into static nature point, class spacing
The sum of the big non-static characteristic point of feature point.The analysis and determination that non-static characteristic point can be thus achieved by the process, from
Above-mentioned non-static characteristic point is removed in fisrt feature point set can obtain second feature point set.It is adopted in the embodiment of the present invention
Non-static characteristic point is determined with the mode of above-mentioned clustering, improvement is not made for clustering algorithm, only by normed space
Change in location information carries out clustering, and obtains cluster result, and non-static characteristic point is realized according to the cluster result really
It is fixed, no longer clustering algorithm is repeated herein.
Based on disclosed above, it is following the utility model has the advantages that due to the embodiment of the present invention can to know that the embodiment of the present invention has
The method of offer can directly remove non-static characteristic point, then to computing resource when can effectively reduce SLAM back-end processing
Consumption improves point cloud and constructs speed;In position fixing process, cloud map construction device can determine itself by multipoint positioning
The presence of current location, non-static characteristic point will increase the probability of erroneous matching, lead to the reduction of positioning accuracy, and the present invention is implemented
Example can be filtered characteristic point before building point cloud, remove behavioral characteristics point therein, significantly improve positioning accuracy;It is right
The filtering of behavioral characteristics point can improve the sparsity of some clouds, and not influence the positioning accuracy of a cloud, while in positioning rank
Section can significantly reduce the calculation amount of positioning, promote locating speed.
In addition, the equipment the embodiment of the invention also provides characteristic point is determined in a kind of synchronous superposition, it should
Equipment can be using the method for determining characteristic point in the synchronous superposition as described in above-described embodiment.And pass through removal
Non-static characteristic point in image characteristic point can greatly promote the speed and precision of component cloud map.Wherein, synchronous positioning with
Map structuring is (SLAM or Simultaneous localization and mapping).
As shown in fig. 6, the original of the equipment to determine characteristic point in the synchronous superposition in the embodiment of the present invention
Manage structure chart, wherein may include:
Image collection module 1 is configured to acquire the image information of different moments respectively;
Processor 2 is configured to the image information of acquisition, obtains the fisrt feature point set in described image information
It closes;Analyze the non-static characteristic point in the fisrt feature point set, wherein the non-static characteristic point is corresponding for mobile object
Characteristic point;The non-static characteristic point is filtered, in the fisrt feature point set to generate second feature point set.
Determine that the equipment of characteristic point can be applied in map structuring in the embodiment of the present invention, in synchronous superposition
Technical field, the equipment can use image collection module 1 acquire image information, which may include depth
Camera or cameras people etc. are spent, image information can be acquired by image collection module 1, which can be
Image information with depth information.Meanwhile also there is the corresponding time in the image information acquired in image collection module
Information, i.e., each image are associated with corresponding temporal information.
In addition, it is illustrated in figure 7 the theory structure schematic diagram of processor in the embodiment of the present invention in the embodiment of the present invention,
Wherein the processor 2 in the embodiment of the present invention can be structured as a data processing device, while also may include multiple subnumbers
According to processing module, to realize different functions.
As shown in Fig. 2, the processor 2 in the embodiment of the present invention may include: extraction module 21 and analysis module 22, wherein
Extraction module 21 can from the reading of image collection module 1 each moment of acquisition image information, and can be based on the acquisition
Image information extracts the fisrt feature point set in image information, it can carries out feature to image information according to preset algorithm
Point extracts, and obtains the fisrt feature point set including multiple characteristic points.Wherein, the preset algorithm of the embodiment of the present invention may include
FAST feature point detection algorithm or SIFT (scale invariant feature conversion) algorithm, can be held by above-mentioned algorithm extraction module 21
The operation that row features described above point extracts.For example, the acquisition Tm moment, the image information f (t) with depth information, which can
To be indicated with RGB data, the fisrt feature point set P ∈ in image information is extracted by above-mentioned preset algorithm (FAST/SIFT)
{ pm0, pm1, pm2 ... pmn }, wherein each characteristic point in set P can be indicated with location information, as pixel coordinate is believed
Breath.Preferably, extraction module 21, can be with corresponding generation about each spy in fisrt feature point set when extracting each characteristic point
The characterization information of point is levied, this feature description can be expressed as D ∈ { dm0, dm1, dm2 ... dmn }, this feature description information
May include the characteristic information of characteristic point, with for identification with distinguish each characteristic point.Obtaining above-mentioned fisrt feature point set
And it after corresponding feature description, corresponding can be stored with time Tm.It can obtain each moment based on above-mentioned and adopt
The fisrt feature point set of the image information of collection and the characterization information of corresponding characteristic point.
After fisrt feature point set of the extraction module 21 from the image information for extracting each moment in image information, mould is analyzed
Block 22 can analyze each fisrt feature point set, to identify the non-static characteristic point in fisrt feature point set.Wherein,
The non-static characteristic point is the corresponding characteristic point of mobile object.Also, analysis module 22 is after identifying non-static characteristic point
The non-static characteristic point can be filtered, from the fisrt feature point set further to generate second feature point set.It should
Each characteristic point in second feature point set is static nature point.Using the second feature point set at each moment as map structure
Use is built, the accuracy and speed of the process can be improved.
It can directly be removed due to equipment provided in an embodiment of the present invention non-static in the image information that each moment obtains
Characteristic point, then consumption when can effectively reduce SLAM back-end processing to computing resource improve point cloud and construct speed;In addition exist
In position fixing process, cloud map construction device can determine the current location of itself by multipoint positioning, non-static characteristic point
In the presence of the probability that will increase erroneous matching, lead to the reduction of positioning accuracy, the embodiment of the present invention can be right before building point cloud
Characteristic point is filtered, and is removed behavioral characteristics point therein, is significantly improved positioning accuracy;Filtering to behavioral characteristics point simultaneously,
The sparsity of some clouds can be improved, and not influence the positioning accuracy of a cloud, while in positioning stage, can significantly reduce positioning
Calculation amount, promoted locating speed.
Further, the analytic process that analysis module 22 executes above-mentioned non-static characteristic point in the embodiment of the present invention can wrap
It includes: respectively from the fisrt feature point set in the described image information that extraction module 21 obtains different moments;Determine each moment
Same characteristic features point in the fisrt feature point set;Mould is obtained based on the determining same characteristic features point and described image
Movement state information of the block at each moment analyzes the non-static characteristic point.
As described in above-described embodiment, by the image information of the available different moments of image collection module 1, and extract
Module 21, which is based on the image information, can further extract the corresponding generation feature description letter of fisrt feature point set merging therein
Breath.Equipment in the preferred embodiment of the present invention can further include memory 3, which can obtain with image
Module 1 and processor 2 connect, generated for storing image information acquired in image collection module 1 and processor 2
Fisrt feature point set and relevant feature point description information etc..
Analysis module 22 in the embodiment of the present invention obtains available each moment from extraction module 21 or memory 3
Image information corresponding to fisrt feature point set, such as the fisrt feature point set and the fisrt feature at Tm+1 moment at Tm moment
Point set.Analysis module 22 can also obtain each fisrt feature point set after P after the fisrt feature point set for obtaining each moment
Same characteristic features point in conjunction, such as can be special by first of each characteristic point and Tm moment in the fisrt feature point set at Tm+1 moment
The characteristic point levied in point set carries out characteristic point Similarity measures, finds out matched characteristic point in Tm and Tm+1 moment, i.e., identical
Characteristic point.Here, the embodiment of the present invention can be according to for the description of feature corresponding to each characteristic point in fisrt feature point set
Information determines the same characteristic features point in each fisrt feature point set, due to the characterization information for characteristic point be being capable of table
Show the information (information that will not generally change) of characteristic point stationary state, therefore the matched characteristic point of characterization information is
It may determine that as identical characteristic point.
Further, analysis module 22 is after the same characteristic features point in the fisrt feature point set for judging each moment, also
Movement state information of the module at each moment can be obtained based on the determining same characteristic features point and described image analyzes institute
State non-static characteristic point.
Specifically, wherein in the embodiment of the present invention analysis module 22 based on the determining same characteristic features point and described
Movement state information of the image collection module at each moment analyzes the non-static characteristic point and may include:
Based on the depth information of the determining same characteristic features point, the spatial position change letter of the same characteristic features point is calculated
Breath;Obtain the spatial position change information that described image obtains module;Spatial position change letter based on the same characteristic features point
Breath and described image obtain the spatial position change information of module, and analysis determines the non-static characteristic point.
It in embodiments of the present invention, can after the same characteristic features point that analysis module 22 determines in each fisrt feature point set
With the depth information based on determining same characteristic features point, the spatial position change information of same characteristic features point is calculated.
As described above, the image collection module in the embodiment of the present invention can obtain the letter of the image with depth information
It ceases, each characteristic point in the fisrt feature point set of extraction also can have depth information, then herein can be based on depth
The location information (such as coordinate information) of each characteristic point of information, determines the spatial position change information of each characteristic point.
Specifically, depth information of the analysis module 22 based on the determining same characteristic features point, meter in the embodiment of the present invention
The spatial position change information for calculating the same characteristic features point may include:
The depth information of the same characteristic features point based on each moment, determines the location information of the same characteristic features point;Base
Difference between the location information of each same characteristic features point at each moment determines the spatial position change information.
That is, analysis module 22 obtains the depth information of image module 1, i.e. the depth information of characteristic point, and can be with
Based on the depth information of each same characteristic features point, determine the location information of each same characteristic features point, the location information with space coordinate into
Row expression, later can be by the coordinate value of the coordinate value of each same characteristic features point at Tm+1 moment and each same characteristic features point at Tm moment
The difference operation for carrying out space vector, to obtain the spatial position change information of each same characteristic features point.For example, can be based on figure
As obtain module camera intrinsic parameter, such as depth information, calculate Tm and Tm+1 moment each same characteristic features point in coordinate
Coordinate value (Xm, Ym, Zm) and (Xm+1, Ym+1, Zm+1) under system, then same characteristic features point of the Tm+1 moment relative to the Tm moment
Spatial variations information can be expressed as space vector (Xm+1-Xm, Ym+1-Ym, Zm+1-Zm).
In addition, position, orientation, angle can all occur since image collection module 1 is during executing image acquisition
Variation, the image collection module in the embodiment of the present invention can also obtain in real time its movement speed of different moments, direction,
The parameter informations such as angle, location information, and the spatial position change information of corresponding determining image collection module.Such as analysis mould
Available Tm to Tm+1 moment code-disc (Odometry) data (v, θ) of block 22, the data indicate the image in 1 time interval
Obtain the camera of module mobile speed and drift angle.That is, in the embodiment of the present invention also it is available at various moments
Image collection module location information (being indicated using space coordinate), and the opposite position obtained between each moment can also be corresponded to
Set variation, i.e. spatial position change information.For example, calculating the location information of Tm and Tm+1 moment image collection module, such as exist
Coordinate value under coordinate system is respectively (Xc, Yc, Zc) and (Xc+1, Yc+1, Zc+1), then phase of the Tm+1 moment relative to the Tm moment
Spatial variations information with characteristic point can be expressed as space vector (Xc+1-Xc, Yc+1-Yc, Zc+1-Zc).
In the sky for the spatial position change information and each same characteristic features point for obtaining the image collection module between different moments
Between position change information when, can determine non-static characteristic point based on above- mentioned information.
Spatial position change information of the analysis module 22 based on the same characteristic features point and described in the embodiment of the present invention
The spatial position change information of image collection module, analysis determine that the non-static characteristic point may include:
The spatial position change information that module is obtained based on described image, to the spatial position change of the same characteristic features point
Operation is normalized in information, to obtain normed space change in location information;
Based on the normed space change in location information, cluster operation is carried out to the same characteristic features point, described in obtaining
Non-static characteristic point.
The spatial position change information of image collection module between different moments and each identical is obtained in analysis module 22
When the spatial position change information of characteristic point, the spatial position change information of module can be obtained based on described image, to described
Operation is normalized in the spatial position change information of same characteristic features point, which may include: to obtain described image
The spatial position change information of modulus block and the spatial position change information of same characteristic features point carry out difference operation, obtain institute
State normed space change in location information.
Can by synchronization calculated image collection module spatial position change information and the moment phase
Spatial position change information with characteristic point carries out difference operation, such as can calculate image of the Tm+1 moment relative to the Tm moment
The spatial position change information for obtaining module is (Xc+1-Xc, Yc+1-Yc, Zc+1-Zc), and the space bit of same characteristic features point
Setting change information is (Xm+1-Xm, Ym+1-Ym, Zm+1-Zm), then above-mentioned for the Tm+1 moment, the normalization of each same characteristics is grasped
The result of work is the difference between (Xc+1-Xc, Yc+1-Yc, Zc+1-Zc) and (Xm+1-Xm, Ym+1-Ym, Zm+1-Zm), should
Difference can be expressed as normed space change in location information.
It is illustrated with following formula, (Lx, Ly, Lz) can be expressed as standard control change in location information, (Xc, Yc,
Zc) indicate that the spatial position change information of image collection module, and (Xm+1, Ym+1, Zm+1) can be expressed as same characteristic features
The spatial position change information of point, then the relationship between three can be with are as follows:
It is possible to further carry out cluster behaviour to the same characteristic features point based on the normed space change in location information
Make, to obtain the non-static characteristic point.In the embodiment of the present invention, analysis module 22 is believed based on the normed space change in location
Breath carries out cluster operation to the same characteristic features point, may include: to obtain the non-static characteristic point
Using the normed space change in location information as core, cluster operation is carried out to the same characteristic features point
Calculate the sum of the class spacing of all kinds of same characteristic features points;
The big a kind of same characteristic features point of the sum of described class spacing is judged as non-static characteristic point.
In the embodiment of the present invention, analysis module 22 can use the k-means clustering method of k=2, circulating repetition K time
Interval carries out clustering to the standard control change in location information of same same characteristic features point P, wherein can be respectively by certain a period of time
The normed space change in location information of the same characteristic features point at quarter is as feature vector, to the normed space of all same characteristic features point P
The sum of change in location information carries out clustering processing, and calculate between class distance, the feature point set for taking the sum of between class distance small is combined into quiet
State characteristic point, the big non-static characteristic point of feature point of the sum of class spacing.Non-static characteristic point can be thus achieved by the process
Analysis and determination, above-mentioned non-static characteristic point is removed from fisrt feature point set can obtain second feature point set.
Non-static characteristic point is determined in the embodiment of the present invention by the way of above-mentioned clustering, clustering algorithm is not made and is changed
Into normed space change in location information is only carried out clustering, and obtains cluster result, is realized according to the cluster result
The determination of non-static characteristic point, herein no longer repeats clustering algorithm.
Based on disclosed above, it is following the utility model has the advantages that due to the embodiment of the present invention can to know that the embodiment of the present invention has
The equipment of offer can directly remove non-static characteristic point, then to computing resource when can effectively reduce SLAM back-end processing
Consumption improves point cloud and constructs speed;In position fixing process, cloud map construction device can determine itself by multipoint positioning
The presence of current location, non-static characteristic point will increase the probability of erroneous matching, lead to the reduction of positioning accuracy, and the present invention is implemented
Example can be filtered characteristic point before building point cloud, remove behavioral characteristics point therein, significantly improve positioning accuracy;It is right
The filtering of behavioral characteristics point can improve the sparsity of some clouds, and not influence the positioning accuracy of a cloud, while in positioning rank
Section can significantly reduce the calculation amount of positioning, promote locating speed.
It is apparent to those skilled in the art that for convenience and simplicity of description, the data of foregoing description
The electronic equipment that processing method is applied to, can be with reference to the corresponding description in before-mentioned products embodiment, and details are not described herein.
Above embodiments are only exemplary embodiment of the present invention, are not used in the limitation present invention, protection scope of the present invention
It is defined by the claims.Those skilled in the art can within the spirit and scope of the present invention make respectively the present invention
Kind modification or equivalent replacement, this modification or equivalent replacement also should be regarded as being within the scope of the present invention.
Claims (8)
1. determining the method for characteristic point in a kind of synchronous superposition comprising:
Image information is acquired using image acquisition equipment;
Image information based on acquisition obtains the fisrt feature point set in described image information;
Analyze the non-static characteristic point in the fisrt feature point set, wherein the non-static characteristic point is mobile object pair
The characteristic point answered;
The non-static characteristic point is filtered, in the fisrt feature point set to generate second feature point set;
Wherein, the non-static characteristic point in the analysis fisrt feature point set includes:
The fisrt feature point set in the described image information of different moments is obtained respectively;
Determine the same characteristic features point in the fisrt feature point set at each moment;
Movement state information of the equipment at each moment, which is obtained, based on the determining same characteristic features point and described image analyzes institute
State non-static characteristic point.
2. described based on the determining same characteristic features point and described image according to the method described in claim 1, wherein
Movement state information of the acquisition equipment at each moment analyzes the non-static characteristic point and includes:
Based on the depth information of the determining same characteristic features point, the spatial position change information of the same characteristic features point is calculated;
Obtain the spatial position change information that described image obtains equipment;
Spatial position change information and described image based on the same characteristic features point obtain the spatial position change letter of equipment
Breath, analysis determine the non-static characteristic point.
3. according to the method described in claim 2, the spatial position change information based on the same characteristic features point, Yi Jisuo
The spatial position change information of image acquisition equipment is stated, analysis determines that the non-static characteristic point includes:
The spatial position change information that equipment is obtained based on described image, to the spatial position change information of the same characteristic features point
Operation is normalized, to obtain normed space change in location information;
Based on the normed space change in location information, cluster operation is carried out to the same characteristic features point, it is described non-quiet to obtain
State characteristic point.
4. according to the method described in claim 2, wherein, the depth information based on the determining same characteristic features point is counted
The spatial position change information for calculating the same characteristic features point includes:
The depth information of the same characteristic features point based on each moment, determines the location information of the same characteristic features point;
The spatial position change information is determined based on the difference between the location information of each same characteristic features point at each moment.
5. according to the method described in claim 3, wherein, the spatial position change for obtaining equipment based on described image is believed
Breath, is normalized operation to the spatial position change information of the same characteristic features point, to obtain normed space change in location letter
Breath includes:
By described image obtain equipment spatial position change information and same characteristic features point spatial position change information into
Row difference operation obtains the normed space change in location information.
6. it is described to be based on the normed space change in location information according to the method described in claim 3, wherein, to the phase
Cluster operation is carried out with characteristic point, includes: to obtain the non-static characteristic point
Using the normed space change in location information as core, cluster operation is carried out to the same characteristic features point;
Calculate the sum of the class spacing of all kinds of same characteristic features points;
The biggish a kind of characteristic point of the sum of the class spacing is judged as non-static characteristic point.
7. determining the equipment of characteristic point in a kind of synchronous superposition comprising:
Image collection module is configured to acquire the image information of different moments respectively;
Processor is configured to the image information of acquisition, obtains the fisrt feature point set in described image information;Analysis
Non-static characteristic point in the fisrt feature point set, wherein the non-static characteristic point is the corresponding feature of mobile object
Point;The non-static characteristic point is filtered, in the fisrt feature point set to generate second feature point set;
Wherein, the processor is further configured to obtain the fisrt feature point set in the described image information of different moments respectively
It closes;Determine the same characteristic features point in the fisrt feature point set at each moment;Based on the determining same characteristic features point, and
Described image obtains movement state information of the equipment at each moment and analyzes the non-static characteristic point.
8. equipment according to claim 7, wherein the processor is further configured to based on the determining identical spy
The depth information for levying point, calculates the spatial position change information of the same characteristic features point;
Obtain the spatial position change information that described image obtains equipment;
Spatial position change information and described image based on the same characteristic features point obtain the spatial position change letter of equipment
Breath, analysis determine the non-static characteristic point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710909238.XA CN107610177B (en) | 2017-09-29 | 2017-09-29 | The method and apparatus of characteristic point is determined in a kind of synchronous superposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710909238.XA CN107610177B (en) | 2017-09-29 | 2017-09-29 | The method and apparatus of characteristic point is determined in a kind of synchronous superposition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107610177A CN107610177A (en) | 2018-01-19 |
CN107610177B true CN107610177B (en) | 2019-10-29 |
Family
ID=61067238
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710909238.XA Active CN107610177B (en) | 2017-09-29 | 2017-09-29 | The method and apparatus of characteristic point is determined in a kind of synchronous superposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107610177B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108460779B (en) * | 2018-02-12 | 2021-09-24 | 浙江大学 | Mobile robot image visual positioning method in dynamic environment |
CN110309240A (en) * | 2018-03-14 | 2019-10-08 | 北京京东尚科信息技术有限公司 | The method and apparatus for removing dynamic object |
CN110533055B (en) * | 2018-05-25 | 2022-04-26 | 北京京东尚科信息技术有限公司 | Point cloud data processing method and device |
CN111860057A (en) * | 2019-04-29 | 2020-10-30 | 北京眼神智能科技有限公司 | Face image blurring and living body detection method and device, storage medium and equipment |
CN112767409A (en) * | 2019-11-05 | 2021-05-07 | 珠海格力电器股份有限公司 | Image processing method and device before positioning, storage medium and computer equipment |
CN112484738B (en) * | 2020-11-24 | 2023-04-28 | 深圳市优必选科技股份有限公司 | Robot mapping method and device, computer readable storage medium and robot |
CN112435278B (en) * | 2021-01-26 | 2021-05-04 | 华东交通大学 | Visual SLAM method and device based on dynamic target detection |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7200282B2 (en) * | 2001-12-12 | 2007-04-03 | Sony Corporation | Implementation of Hough transform and its application in line detection and video motion analysis |
CN103154666B (en) * | 2011-06-14 | 2015-03-18 | 日产自动车株式会社 | Distance measurement device and environment map generation apparatus |
CN104424635A (en) * | 2013-08-30 | 2015-03-18 | 联想(北京)有限公司 | Information processing method, system and equipment |
CN104424481B (en) * | 2013-09-02 | 2018-11-09 | 联想(北京)有限公司 | The method and smart machine of information processing |
CN103456022B (en) * | 2013-09-24 | 2016-04-06 | 中国科学院自动化研究所 | A kind of high-resolution remote sensing image feature matching method |
CN103955948B (en) * | 2014-04-03 | 2016-10-05 | 西北工业大学 | A kind of space movement target detection method under dynamic environment |
CN105303554B (en) * | 2015-09-16 | 2017-11-28 | 东软集团股份有限公司 | The 3D method for reconstructing and device of a kind of image characteristic point |
CN107146238B (en) * | 2017-04-24 | 2019-10-11 | 西安电子科技大学 | Based on the preferred motion target tracking method of characteristic block |
-
2017
- 2017-09-29 CN CN201710909238.XA patent/CN107610177B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN107610177A (en) | 2018-01-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107610177B (en) | The method and apparatus of characteristic point is determined in a kind of synchronous superposition | |
CN108549873B (en) | Three-dimensional face recognition method and three-dimensional face recognition system | |
US10534957B2 (en) | Eyeball movement analysis method and device, and storage medium | |
CN105404886B (en) | Characteristic model generation method and characteristic model generating means | |
WO2017024963A1 (en) | Image recognition method, measure learning method and image source recognition method and device | |
US8855363B2 (en) | Efficient method for tracking people | |
CN103310194B (en) | Pedestrian based on crown pixel gradient direction in a video shoulder detection method | |
WO2016150240A1 (en) | Identity authentication method and apparatus | |
US20130251246A1 (en) | Method and a device for training a pose classifier and an object classifier, a method and a device for object detection | |
JP5959093B2 (en) | People search system | |
Kaâniche et al. | Recognizing gestures by learning local motion signatures of HOG descriptors | |
CN101131728A (en) | Face shape matching method based on Shape Context | |
CN107368770B (en) | Method and system for automatically identifying returning passenger | |
CN107766864B (en) | Method and device for extracting features and method and device for object recognition | |
Sang et al. | Pose-invariant face recognition via RGB-D images | |
Whitehill et al. | A discriminative approach to frame-by-frame head pose tracking | |
Andiani et al. | Face recognition for work attendance using multitask convolutional neural network (MTCNN) and pre-trained facenet | |
CN112668383A (en) | Attendance checking method and system based on face recognition, electronic equipment and storage medium | |
CN116958724A (en) | Training method and related device for product classification model | |
CN108664919A (en) | A kind of Activity recognition and detection method based on single sample | |
KR100711223B1 (en) | Face recognition method using Zernike/LDA and recording medium storing the method | |
Duan et al. | Bio-inspired visual attention model and saliency guided object segmentation | |
Sarfraz et al. | An efficient front-end facial pose estimation system for face recognition | |
Chen et al. | Gender recognition from gait using radon transform and relevant component analysis | |
BERISTAIN | Human-robot interface using face detection and recognition, for the service robot," donaxi |
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 |