CN109146936A - A kind of image matching method, device, localization method and system - Google Patents
A kind of image matching method, device, localization method and system Download PDFInfo
- Publication number
- CN109146936A CN109146936A CN201810783004.XA CN201810783004A CN109146936A CN 109146936 A CN109146936 A CN 109146936A CN 201810783004 A CN201810783004 A CN 201810783004A CN 109146936 A CN109146936 A CN 109146936A
- Authority
- CN
- China
- Prior art keywords
- image
- tensor
- matching
- pixel
- directional diagram
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of image matching method, device, localization method and system, matching process includes the tensor direction for calculating each pixel in the first image, obtains the first tensor directional diagram of the first image;The tensor direction of each pixel in calculating benchmark image, and the benchmark image is divided into the multiple and equal-sized subgraph of the first image, determine the second tensor directional diagram of each subgraph;The matching value of the second tensor directional diagram of the first tensor directional diagram and the subgraph is calculated one by one, and determines the highest subgraph of matching degree as matching result according to the matching value.The advantages that matching process and device have the matching being widely portable between a variety of different heterologous images, and matching accuracy is high, and matching speed is fast.Localization method and system, which have, can be achieved high-precision, low cost, affected by environment small, the advantages such as low to external signal dependency degree.
Description
Technical field
The present invention relates to images match field of locating technology more particularly to a kind of image matching method, device, localization methods
And system.
Background technique
Current unmanned aircraft industry development is swift and violent, and the target detecting based on unmanned plane, unmanned inspecting hole device platform is extensive
For the every field such as military affairs, industry, agricultural, rescue, clear image, the video of target can not only be shot, obtained, moreover it is possible to
Intellectual analysis and automatic positioning are carried out to target.
Ground target of taking photo by plane location technology is always the hot technology of all kinds of researchers and engineering staff's concern, the technology
It is based primarily upon GPS, inertial navigation, laser radar range and photogrammetric technology, it is usually comprehensive to be completed finally using these technologies
Target positioning.The relevant technologies mainly have based on GPS/ inertial navigation/laser radar range combined orientation technology and are based at this stage
The combined orientation technology of GPS/ inertial navigation/photogrammetric.The latter can continue to be subdivided into based on GPS/ inertial navigation/intersection measurement group again
Close location technology and based on GPS/ inertial navigation/back projection measurement combined orientation technology.
(1) based on GPS/ inertial navigation/laser radar range combined orientation technology, the position of airborne platform is provided using GPS,
Posture and the course of platform are provided using inertial navigation, using the distance between lidar measurement target and airborne platform, then root
The position of target is calculated according to the distance between airborne platform position, posture, course, laser radar direction, target and airborne platform.
It include: in the prior art to make tranquil in the summer to the research achievement of the technology, Jiang Lixing, Fan Xiaozhong: airborne laser radar positioning for ground error
Analyze [J] Surveying and mapping Technology, 2011,28 (5): 365-368;And Wang Jianjun, Xu Lijun, Li little Lu: attitude angle with
Influence [C] of the machine measurement error to airborne laser radar point cloud positioning accuracy and three-dimensional imaging precision, national laser radar is over the ground
Observe advanced scientific seminar, 2010.Based on the positioning with higher of GPS/ inertial navigation/laser radar range combined orientation technology
Precision is able to achieve sub_meter position, but still has the following deficiencies: laser radar needs photoelectric nacelle that can accurately control direction simultaneously
The angle that target is directed toward between flying platform is calculated, inertial navigation component is additionally needed to be capable of the course of precise measurement flying platform
And posture, otherwise under the conditions of telemeasurement, small angular deviation may cause very big position deviation.Therefore the technology needs
Want high cost, high-precision photoelectric nacelle and inertial navigation component, laser radar hardware cost itself is high in addition, so system integral into
This is higher, is unfavorable for using in some low-cost platforms such as unmanned plane.In addition laser radar range mainly measures point target,
Use, image that method based on Aerial Images measurement can take photo by plane to one convenient not as good as the method measured based on Aerial Images
The target of upper any position measures.
(2) based on GPS/ inertial navigation/intersection measurement combined orientation technology, the position of airborne platform is provided using GPS, is utilized
Inertial navigation provides posture and the course of platform, carries out shooting intersection measurement target and aircraft to target from multiple angles using camera
Then relative position calculates target according to relative position between airborne platform position, posture, course, target and airborne platform
Absolute position.It include: in the prior art easy side, Xiao Zhiming, Li Yuxiong: automatic sorting measurement data to the research achievement of the technology
Multiple target crossing location method [J] TV tech, 2014,38 (11): 219-223;Sun Hui, Li Zhiqiang, Zhang Jianhua: airborne
Photoelectric platform target crossing location [J] Chinese Optical, 2015,8 (6): 988-996;Yu Qifeng, Shangyang: videographic measurment principle
With application study [M] Science Press, 2009..The technology still has following deficiency: since intersection measurement needs multi-angle
Photographic subjects, therefore limitation is generated to flight path, biggish intersection angle sometimes can not be often obtained, to influence to measure
Precision.In addition, longshot measurement is easy to be interfered by atmospheric refraction, this also can seriously lower the positioning accurate of intersection measurement
Degree.The about tens of rice of target location error outside usual logarithm km.
(3) based on GPS/ inertial navigation/back projection measurement combined orientation technology, the position of airborne platform is provided using GPS,
Posture and the course of platform are provided using inertial navigation, is pressed from both sides using the direction of camera shooting measurement target and aircraft on photoelectric nacelle
Then angle calculates the exhausted of target according to the angular separation of airborne platform position, posture, course, flying height and target and aircraft
To position.It include: in the prior art Merino L, Caballero F, Mart í nez-de Dios to the research achievement of the technology
J R, et al.A cooperative perception system for multiple UAVs:Application to
Automatic detection of forest fires [J] .Journal of Field Robotics, 2006,23 (3-
4): 165-184;Merino L, Caballero F, Martinez-de Dios J R, et al.Cooperative fire
Detection using unmanned aerial vehicles [C] //Robotics and Automation,
2005 IEEE International Conference on.IEEE of 2005.ICRA 2005.Proceedings of the,
2005:1884-1889.The technology still has following deficiency: the technology utilizes measurement data: posture, camera are directed toward, flight is high
Degree carries out directly calculation to target position, and therefore, attitude error, error in pointing, height error can have very target positioning result
It is big to influence.Target location error outside usual logarithm km is up to hundred meters or more.
In the prior art, carry out images match (especially heterologous images match) method mainly include PQ-HOG,
HOPC, MI, MTM, NCC, GO, ImpGO etc., but these matching process are correct in the adaptability of different peculiar smell source images, matching
There are still the space improved in rate, it is especially applied to the positioning field based on images match, the height for matching accuracy is straight
The height for being related to positioning accuracy is connect, in order to improve positioning accuracy, improves and the accuracy for improving images match also has very
Important meaning.
Summary of the invention
The technical problem to be solved in the present invention is that, for technical problem of the existing technology, the present invention provides one
Kind is widely used in the matching between a variety of different heterologous images, and matching accuracy is high, the fast images match side of matching speed
Method, device, also, on the basis of image matching method, realize high-precision, low cost is affected by environment small, to external signal
The low localization method of dependency degree and system.
In order to solve the above technical problems, technical solution proposed by the present invention are as follows: a kind of image matching method calculates the first figure
The tensor direction of each pixel, obtains the first tensor directional diagram of the first image as in;Each pixel in calculating benchmark image
The tensor direction of point, and the benchmark image is divided into the multiple and equal-sized subgraph of the first image, determine each institute
State the second tensor directional diagram of subgraph;The second tensor directional diagram of the first tensor directional diagram and the subgraph is calculated one by one
Matching value, and determine the highest subgraph of matching degree as matching result according to the matching value.
Further, tensor direction formula according to formula (1), which calculates, determines:
In formula (1), θ is the tensor direction value for the pixel being calculated, t11、t12、t22It is the tensor of the pixel
Value.
Further, tensor value formula according to formula (2), which calculates, determines:
T in formula (2)11、t12、t22It is the tensor value of the pixel, GσThe Gaussian filter for being σ for preset standard deviation
Parameter, IxFor the local derviation of default local image region in the X direction comprising the pixel, IyIt is default comprising the pixel
The local derviation of local image region in the Y direction.
Further, the formula according to formula (3) calculates matching value:
In formula (3), S (O (t), O (w)) is the matching value of the first tensor directional diagram O (t) and the second tensor directional diagram O (w),
θiFor the tensor direction of pixel i in the first tensor directional diagram, θ 'iFor the tensor direction of pixel i in the second tensor directional diagram, n
It is respectively the pixel number of X-direction and Y-direction in the first tensor directional diagram with m.
A kind of image matching apparatus, including processor and memory, the processor are deposited on the memory for executing
The program of storage is stored with the program for being performed and any one of as above the method can be achieved on the memory.
A kind of localization method of images match, comprising: obtain the first image, the first image is to including mesh to be positioned
Area to be targeted in being marked on take pictures the image of acquisition;
By the first image and the benchmark image with coordinate predefined according to any one of such as Claims 1-4
The matching process is matched, and determines matching result of the first image in the benchmark image;
It determines the coordinate of the first image according to the matching result, and determines the seat of target to be positioned in the first image
Mark.
It further, further include being corrected to the image of the acquisition of taking pictures;It is described correction include to described image into
Row is just lower to regard correction process.
It further, further include being modified to the image of the acquisition of taking pictures;The amendment includes to the photo
Direction is modified, so that the direction of described image is consistent with the direction of the benchmark image, and/or: the photo is divided
Resolution is adjusted, so that the resolution ratio of described image is consistent with the resolution ratio of the benchmark image.
A kind of positioning system of images match, including image collection module and matching module and locating module;
It is to including comprising target to be positioned that described image, which obtains module for obtaining the first image, the first image,
Area to be targeted take pictures the image of acquisition;
The matching locating module be used for by the first image and the benchmark image with coordinate that has predefined according to
Described in any item matching process as above are matched, and determine matching result of the first image in the benchmark image;
The locating module is used to determine the coordinate of the first image according to the matching result, and determines the first image
In target to be positioned coordinate.
Further, described image obtains module and is also used to: the image of the acquisition of taking pictures is corrected and/or is repaired
Just;The correction includes that just lower view correction process is carried out to described image;The amendment includes carrying out to the direction of the photo
Amendment, so that the direction of described image is consistent with the direction of the benchmark image, and/or: the resolution ratio of the photo is carried out
Adjustment, so that the resolution ratio of described image is consistent with the resolution ratio of the benchmark image.
Compared with the prior art, the advantages of the present invention are as follows:
1, matching process of the invention obtains tensor direction by the tensor direction of calculating the first image and benchmark image
Scheme, and realize the matching between the first image and benchmark image by tensor directional diagram, the first image and benchmark may be implemented
Fast and accurately matching when image is heterologous image, further, it is possible to widely be adapted between a variety of different heterologous images
Matching.
2, the present invention is matched by aerial photograph (the first image) with the benchmark image for being previously determined coordinate, is passed through
The coordinate information for determining aerial photograph is matched, the coordinate of target to be positioned in aerial photograph is determined further according to the coordinate, on the one hand,
Position fixing process does not depend on Global Satellite Navigation System (GNSS), does not depend on the external informations such as road sign point, passes through the photograph taken photo by plane
Piece and pre-stored benchmark image can complete position fixing process, small to external information interdependency, by external factor such as environment
Influence it is small, be limited small, good reliability;On the other hand, there is high-precision, high-resolution benchmark image by being stored in advance,
Aerial Images are matched with the benchmark image again, so that it may realize high-precision positioning.
Detailed description of the invention
Fig. 1 is the flow diagram of the specific embodiment of the invention.
Fig. 2 is the matching schematic diagram of the specific embodiment of the invention.
Fig. 3 is the matching test use-case schematic diagram of the specific embodiment of the invention.
Fig. 4 is the matching result comparative analysis figure of the specific embodiment of the invention.
Fig. 5 is that the specific embodiment of the invention positions schematic diagram.
Fig. 6 is the image flame detection schematic diagram of the specific embodiment of the invention.
Fig. 7 is the technical effect of the specific embodiment of the invention and the effect contrast figure of the prior art.
Specific embodiment
Below in conjunction with Figure of description and specific preferred embodiment, the invention will be further described, but not therefore and
It limits the scope of the invention.
As shown in figure 1 shown in dotted line frame, the image matching method of the present embodiment, comprising: calculate each pixel in the first image
The tensor direction of point, obtains the first tensor directional diagram of the first image;The tensor direction of each pixel in calculating benchmark image, and
Benchmark image is divided into the multiple and equal-sized subgraph of the first image, determines the second tensor directional diagram of each subgraph;One by one
The matching value of the second tensor directional diagram of the first tensor directional diagram and subgraph is calculated, and determines that matching degree is highest according to matching value
Subgraph is as matching result.In the present embodiment, the first image is preferably the rectangular image that side length is 64 pixels to 96 pixels.
In the present embodiment, tensor direction formula according to formula (1), which calculates, determines:
In formula (1), θ is the tensor direction value for the pixel being calculated, t11、t12、t22It is the tensor of the pixel
Value.
Tensor value formula according to formula (2), which calculates, to be determined:
T in formula (2)11、t12、t22It is the tensor value of the pixel, GσThe Gaussian filter for being σ for preset standard deviation
Parameter, IxFor the local derviation of default local image region in the X direction comprising the pixel, IyIt is default comprising the pixel
The local derviation of local image region in the Y direction.In the present embodiment, the default local image region comprising the pixel is with this
Centered on pixel, side length is the region of default pixel, and such as 3 pixels multiply the region of 3 pixels or 5 pixels multiply the areas of 5 pixels
Domain.
Matching value is calculated according to formula shown in formula (3):
In formula (3), S (O (t), O (w)) is the matching value of the first tensor directional diagram O (t) and the second tensor directional diagram O (w),
θiFor the tensor direction of pixel i in the first tensor directional diagram, θ 'iFor the tensor direction of pixel i in the second tensor directional diagram, n
It is respectively the pixel number of X-direction and Y-direction in the first tensor directional diagram with m.Due to the first tensor directional diagram size with
The size of second tensor directional diagram be it is identical, i.e. the first tensor directional diagram and the second tensor directional diagram are all by n × m pixel
The tensor directional diagram that the image of point is calculated.According to formula shown in formula (3), matching value is smaller, illustrates the first tensor directional diagram
Difference between the second tensor directional diagram is smaller, and matching degree is higher.
In the present embodiment, as shown in Fig. 2, by the realtime graphic of certain ground region of acquisition of taking photo by plane, i.e., realtime graphic is
The first image in this method, benchmark image are the bases that coordinate position has just been obtained and had been determined by processing before taking photo by plane
Quasi- image, the benchmark image can be the image of satellite remote sensing acquisition, and have passed through and be accurately positioned the coordinate for determining image (such as
Longitude and latitude).In the present embodiment, the first image is the image of n × n pixel.It is counted by above-mentioned formula (1) and formula (2)
Calculate the first tensor directional diagram of available first image.Each picture in benchmark image can also be calculated in the same way
The tensor direction of vegetarian refreshments, also, benchmark image is divided into the region of n × n pixel, the corresponding subgraph in each region,
As shown in the dashed rectangle in Fig. 2 in benchmark image, available one of the corresponding subgraph in region of each n × n pixel
Corresponding second tensor directional diagram.Again by the first tensor directional diagram of the first image and subgraph corresponding second each in benchmark image
Tensor directional diagram is compared, i.e., carries out matching value calculating according to formula shown in formula (3), benchmark image in Fig. 2 can be obtained
Matching degree highest between the corresponding second tensor directional diagram of region subgraph shown in middle solid line boxes and the first tensor directional diagram, i.e.,
Can by the first images match into benchmark image region corresponding to solid line boxes, complete the matching of image.After the completion of matching,
Since the coordinate of benchmark image can determine that the coordinate in the first image upper left corner is (rx, ry) it has been determined that then passing through matching, that
Any point (x, y) in first image, it can determine that its coordinate is (rx+x, ry+y).
In the present embodiment, by nearly hundred kinds of different images by matching process (being denoted as PG) of the invention and the prior art
In the matching algorithms such as PQ-HOG, HOPC, MI, GO, ImpGO compare and analyze, image used by comparative analysis include but
Several heterologous images of Fig. 3 displaying are not limited to, Mean match accuracy comparing result is as shown in figure 4, Mean match accuracy
Refer to and the matching analysis is carried out to used image, and calculates using the average value for matching accuracy when different images.By comparing
It was determined that the Mean match accuracy of the method for the present invention is above matching process in the prior art, especially it is substantially better than
GO, MI, PQ-HOG scheduling algorithm is for side length in template size (size of the first image) relative to HOPC, ImpGO scheduling algorithm
The rectangular image of 64 pixels to 96 pixel ranges is also significantly improved.The numerical value of template size refers to that the length and width of template take accordingly
The size of the square area of magnitude pixels.Since the accuracy of images match can seriously affect the positioning based on images match
Accurate rate, therefore, the raising of images match accuracy are still of great significance to.Wherein, PQ-HOG matching process is joined
See existing technical literature: A.Sibiryakov, " Fast and high-performance template matching
Method, " in CVPR, 2011, pp.1417-1424.HOPC matching process is referring to existing technical literature: Y.Ye and
L.Shen, " Hopc:a Novel Similarity Metric Based on Geometric Structural
Properties for Multi-Modal Remote Sensing Image Matching, " in ISPRS Annals of
Photogrammetry, Remote Sensing and Spatial Information Sciences, 2016, vol.3,
pp.9-16.MI matching process is referring to existing technical literature: P.Viola and W.Wells, " Alignment by
Maximization of mutual information, " International journal of computer vision,
Vol.24, no.2, pp.137-154,1977.
The image matching apparatus of the present embodiment, including processor and memory, processor store on memory for executing
Program, the program for being performed and any one of as above method can be achieved is stored on memory.
As shown in Figure 1, the localization method of the images match of the present embodiment, comprising: obtain the first image, the first image is pair
Take pictures the image of acquisition comprising the area to be targeted including target to be positioned;
First image is carried out with the benchmark image with coordinate predefined according to matching process any one of as above
Matching, determines matching result of first image in benchmark image;The coordinate of the first image is determined according to matching result, and is determined
The coordinate of target to be positioned in first image.
It in the present embodiment, further include being corrected to the image for acquisition of taking pictures;Correction includes that just lower view is carried out to image
Correction process.It further include being modified to the image for acquisition of taking pictures;Amendment includes being modified to the direction of photo, so that image
Direction it is consistent with the direction of benchmark image, and/or: the resolution ratio of photo is adjusted, so that the resolution ratio and base of image
The resolution ratio of quasi- image is consistent.
In the present embodiment, the specific implementation process of this localization method is described by a specific position fixing process.Such as
Shown in Fig. 5, target to be positioned is vehicle A, and for vehicle A along road driving, there are building and other topography and landform characters in both sides of the road
(such as tree, river, hillside).The position shown in the vehicle driving to Fig. 5 actual ground situation, when being positioned.It is logical
It crosses aircraft (such as unmanned plane) and carries photographing device (camera) lift-off, boat of the shooting comprising the area to be targeted including vehicle A
It takes a picture, area to be targeted is as shown in the dotted line in practical surface state figure.Obtain aerial photograph shown in fig. 5 (i.e. the first figure
Picture).Also equipped with the equipment such as attitude transducer and altimeter on aircraft, records photographing photo while shooting aerial photograph
The state parameter of Shi Xiangji, camera status parameter include posture, direction, height.
In the present embodiment, it since the setting angle of state of flight and photographing device by aircraft is influenced, claps
It might not be directly suitable for being matched with benchmark image according to aerial photograph captured by equipment.Therefore, it is necessary to set according to taking pictures
Standby inner parameter (such as focal length, principal point) carries out image flame detection processing to Aerial Images, and aerial photograph is corrected the lower view that is positive
Picture.Image flame detection processing can be carried out by the image processing algorithm of existing maturation.It, can be by left side in Fig. 6 by correcting process
Image flame detection is image right in Fig. 6, so as to further increase matched accuracy.Meanwhile according to the side of benchmark image
To parameters such as, resolution ratio, further the direction of Aerial Images, resolution ratio, scaling size etc. are adjusted, so that after adjustment
The direction of Aerial Images, the direction of resolution ratio and benchmark image, resolution ratio it is close or consistent, to facilitate subsequent progress image
Match.The adjustment process can also be carried out by existing mature image processing algorithm.
In the present embodiment, the benchmark image with coordinate predefined is as shown in the benchmark image in Fig. 5, the benchmark
Image is that pre- first pass through is taken photo by plane or the modes such as satellite remote sensing obtain, and coordinate has been determined for the benchmark image in advance.As adopted in Fig. 5
With the form (straight dashed line in Fig. 5 benchmark image indicates) of longitude and latitude, the coordinate of benchmark image accurately draw
Point.Certainly, coordinate information can also be described in the form of other coordinate systems.Certainly, since benchmark image is to obtain in advance
, therefore, the ground attachment (such as woods, building) in benchmark image may be different with current actual ground situation
It causes, namely also currently inconsistent by aerial photograph obtained of taking photo by plane, still, still can there is identical feature between the two
(such as landform, there is no the building of variation, roads etc.), therefore, through the invention in image matching method, pass through meter
Aerial photograph can be accurately matched in benchmark image by the tensor of nomogram picture, as shown in dashed rectangle in Fig. 5 benchmark image
Region, since benchmark image has determining coordinate information, accordingly it is also possible to accurately determine the coordinate information of aerial photograph.
In the present embodiment, identify that target (i.e. vehicle A) to be positioned is being navigated in aerial photograph by image recognition technology
It position in taking a picture can by simply calculating since the coordinate information of aerial photograph has been obtained by matching
The coordinate of target (vehicle A) to be positioned is calculated.To complete the positioning to target to be positioned.
The comparison of the localization method of the present embodiment and conventional method as shown in fig. 7, the present embodiment above-mentioned position fixing process
In, completely without the positioning signal for relying on Global Satellite Navigation System (such as GPS, Beidou etc.), also need not rely on road sign point
Deng.As long as the benchmark image of region has been stored in advance, by after taking photo by plane and obtaining aerial photograph, so that it may easily and fast
Complete the positioning to target.Also, on the basis of using high coordinate precision and high-resolution benchmark image, pass through image
It also can achieve accordingly with the precision positioned very high, positioning accuracy can achieve sub-meter grade.Moreover, the position fixing process by
In not depending on external information, therefore strong antijamming capability, stability is high, good reliability.
The positioning system of the images match of the present embodiment, including image collection module and matching module and locating module;Figure
It is to take pictures to comprising the area to be targeted including target to be positioned as obtaining module for obtaining the first image, the first image
The image of acquisition;Matching locating module is used for the first image and the benchmark image with coordinate that has predefined according to such as taking up an official post
One matching process is matched, and determines matching result of first image in benchmark image;Locating module is used for basis
It determines the coordinate of the first image with result, and determines the coordinate of target to be positioned in the first image.Image collection module is also used to:
The image for acquisition of taking pictures is corrected and/or is corrected;Correction includes that just lower view correction process is carried out to image;Amendment includes pair
The direction of photo is modified so that the direction of image is consistent with the direction of benchmark image, and/or: to the resolution ratio of photo into
Row adjustment, so that the resolution ratio of image is consistent with the resolution ratio of benchmark image.
Above-mentioned only presently preferred embodiments of the present invention, is not intended to limit the present invention in any form.Although of the invention
It has been disclosed in a preferred embodiment above, however, it is not intended to limit the invention.Therefore, all without departing from technical solution of the present invention
Content, technical spirit any simple modifications, equivalents, and modifications made to the above embodiment, should all fall according to the present invention
In the range of technical solution of the present invention protection.
Claims (10)
1. a kind of image matching method, it is characterised in that: the tensor direction for calculating each pixel in the first image obtains described the
First tensor directional diagram of one image;The tensor direction of each pixel in calculating benchmark image, and the benchmark image is divided
For the multiple and equal-sized subgraph of the first image, the second tensor directional diagram of each subgraph is determined;Institute is calculated one by one
The matching value of the second tensor directional diagram of the first tensor directional diagram and the subgraph is stated, and matching degree is determined according to the matching value
Highest subgraph is as matching result.
2. image matching method according to claim 1, it is characterised in that: tensor direction public affairs according to formula (1)
Formula, which calculates, to be determined:
In formula (1), θ is the tensor direction value for the pixel being calculated, t11、t12、t22It is the tensor value of the pixel.
3. image matching method according to claim 2, it is characterised in that: tensor value formula according to formula (2)
It calculates and determines:
T in formula (2)11、t12、t22It is the tensor value of the pixel, GσThe Gaussian filter parameter for being σ for preset standard deviation,
IxFor the local derviation of default local image region in the X direction comprising the pixel, IyFor the default Local map comprising the pixel
As the local derviation of region in the Y direction.
4. image matching method according to claim 3, it is characterised in that: calculate matching value according to formula shown in formula (3):
In formula (3), S (O (t), O (w)) is the matching value of the first tensor directional diagram O (t) and the second tensor directional diagram O (w), θiFor
The tensor direction of pixel i, θ ' in first tensor directional diagramiFor the tensor direction of pixel i in the second tensor directional diagram, n and m
The pixel number of X-direction and Y-direction in respectively the first tensor directional diagram.
5. a kind of image matching apparatus, it is characterised in that: including processor and memory, the processor is for executing described deposit
The program stored on reservoir, being stored on the memory to be performed can be achieved such as any one of Claims 1-4 the method
Program.
6. a kind of localization method of images match, it is characterised in that:
The first image is obtained, the first image is to carry out acquisition of taking pictures to comprising the area to be targeted including target to be positioned
Image;
By the first image and the benchmark image with coordinate predefined according to as described in any one of Claims 1-4
Matching process matched, determine matching result of the first image in the benchmark image;
It determines the coordinate of the first image according to the matching result, and determines the coordinate of target to be positioned in the first image.
7. the localization method of images match according to claim 6, it is characterised in that: further include to the acquisition of taking pictures
Image is corrected;The correction includes that just lower view correction process is carried out to described image.
8. the localization method of images match according to claim 7, it is characterised in that: further include to the acquisition of taking pictures
Image is modified;The amendment includes being modified to the direction of the photo, so that the direction of described image and the base
The direction of quasi- image is consistent, and/or: the resolution ratio of the photo is adjusted so that the resolution ratio of described image with it is described
The resolution ratio of benchmark image is consistent.
9. a kind of positioning system of images match, it is characterised in that: including image collection module and matching module and locating module;
It is to comprising undetermined including target to be positioned that described image, which obtains module for obtaining the first image, the first image,
Position region take pictures the image of acquisition;
The matching locating module is used for the first image and the benchmark image with coordinate predefined according to such as power
Benefit requires 1 to 4 described in any item matching process to be matched, and determines matching of the first image in the benchmark image
As a result;
The locating module is used to determine the coordinate of the first image according to the matching result, and determine in the first image to
Position the coordinate of target.
10. the positioning system of images match according to claim 9, it is characterised in that: described image obtains module and also uses
In: the image of the acquisition of taking pictures is corrected and/or is corrected;The correction includes that just lower view correction is carried out to described image
Processing;The amendment includes being modified to the direction of the photo, so that the direction of described image and the benchmark image
Direction is consistent, and/or: the resolution ratio of the photo is adjusted, so that the resolution ratio of described image and the benchmark image
Resolution ratio it is consistent.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810783004.XA CN109146936B (en) | 2018-07-17 | 2018-07-17 | Image matching method, device, positioning method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810783004.XA CN109146936B (en) | 2018-07-17 | 2018-07-17 | Image matching method, device, positioning method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109146936A true CN109146936A (en) | 2019-01-04 |
CN109146936B CN109146936B (en) | 2021-04-27 |
Family
ID=64800745
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810783004.XA Active CN109146936B (en) | 2018-07-17 | 2018-07-17 | Image matching method, device, positioning method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109146936B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111238488A (en) * | 2020-03-18 | 2020-06-05 | 湖南云顶智能科技有限公司 | Aircraft accurate positioning method based on heterogeneous image matching |
CN114612555A (en) * | 2022-03-17 | 2022-06-10 | 杭州弥深智能科技有限公司 | Positioning method, positioning device, electronic equipment and storage medium |
CN115932823A (en) * | 2023-01-09 | 2023-04-07 | 中国人民解放军国防科技大学 | Aircraft ground target positioning method based on heterogeneous region feature matching |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060045380A1 (en) * | 2004-08-26 | 2006-03-02 | Jones Graham R | Data processing |
WO2006099339A2 (en) * | 2005-03-14 | 2006-09-21 | The Board Of Trustees Of The University Of Illinois | A fiber coherence index apparatus and method for imaging and characterizing fibrous structures |
CN102609949A (en) * | 2012-02-16 | 2012-07-25 | 南京邮电大学 | Target location method based on trifocal tensor pixel transfer |
CN104463098A (en) * | 2014-11-04 | 2015-03-25 | 中国矿业大学(北京) | Method for recognizing coal and rock through structure tensor direction column diagram features of image |
CN106407315A (en) * | 2016-08-30 | 2017-02-15 | 长安大学 | Vehicle self-positioning method based on street view image database |
CN107609507A (en) * | 2017-09-08 | 2018-01-19 | 哈尔滨工业大学 | Feature based tensor sum supports the Remote Sensing Target recognition methods of tensor machine |
-
2018
- 2018-07-17 CN CN201810783004.XA patent/CN109146936B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060045380A1 (en) * | 2004-08-26 | 2006-03-02 | Jones Graham R | Data processing |
WO2006099339A2 (en) * | 2005-03-14 | 2006-09-21 | The Board Of Trustees Of The University Of Illinois | A fiber coherence index apparatus and method for imaging and characterizing fibrous structures |
CN102609949A (en) * | 2012-02-16 | 2012-07-25 | 南京邮电大学 | Target location method based on trifocal tensor pixel transfer |
CN104463098A (en) * | 2014-11-04 | 2015-03-25 | 中国矿业大学(北京) | Method for recognizing coal and rock through structure tensor direction column diagram features of image |
CN106407315A (en) * | 2016-08-30 | 2017-02-15 | 长安大学 | Vehicle self-positioning method based on street view image database |
CN107609507A (en) * | 2017-09-08 | 2018-01-19 | 哈尔滨工业大学 | Feature based tensor sum supports the Remote Sensing Target recognition methods of tensor machine |
Non-Patent Citations (1)
Title |
---|
董晶: "模板图像快速可靠匹配技术研究", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111238488A (en) * | 2020-03-18 | 2020-06-05 | 湖南云顶智能科技有限公司 | Aircraft accurate positioning method based on heterogeneous image matching |
CN114612555A (en) * | 2022-03-17 | 2022-06-10 | 杭州弥深智能科技有限公司 | Positioning method, positioning device, electronic equipment and storage medium |
CN115932823A (en) * | 2023-01-09 | 2023-04-07 | 中国人民解放军国防科技大学 | Aircraft ground target positioning method based on heterogeneous region feature matching |
Also Published As
Publication number | Publication date |
---|---|
CN109146936B (en) | 2021-04-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103822635B (en) | The unmanned plane during flying spatial location real-time computing technique of view-based access control model information | |
CN106408601B (en) | A kind of binocular fusion localization method and device based on GPS | |
CN105698742B (en) | A kind of quick land area measuring device and measurement method based on unmanned plane | |
JP4232167B1 (en) | Object identification device, object identification method, and object identification program | |
US20110282580A1 (en) | Method of image based navigation for precision guidance and landing | |
CN106155081B (en) | A kind of a wide range of target monitoring of rotor wing unmanned aerial vehicle and accurate positioning method | |
CN103093459B (en) | Utilize the method that airborne LiDAR point cloud data assisted image mates | |
CN108007474A (en) | A kind of unmanned vehicle independent positioning and pose alignment technique based on land marking | |
TWI444593B (en) | Ground target geolocation system and method | |
CN109146936A (en) | A kind of image matching method, device, localization method and system | |
US10186046B2 (en) | Navigation based on downward facing sensors | |
CN107607091A (en) | A kind of method for measuring unmanned plane during flying flight path | |
CN110515110B (en) | Method, device, equipment and computer readable storage medium for data evaluation | |
CN107065895A (en) | A kind of plant protection unmanned plane determines high-tech | |
CN106971408A (en) | A kind of camera marking method based on space-time conversion thought | |
CN114004977A (en) | Aerial photography data target positioning method and system based on deep learning | |
EP3183712A1 (en) | Determining compass orientation of imagery | |
CN116625354A (en) | High-precision topographic map generation method and system based on multi-source mapping data | |
CN108801225A (en) | A kind of unmanned plane tilts image positioning method, system, medium and equipment | |
CN113819904B (en) | polarization/VIO three-dimensional attitude determination method based on zenith vector | |
CN101793517B (en) | Online quick method for improving accuracy of attitude determination of airborne platform | |
CN113340272B (en) | Ground target real-time positioning method based on micro-group of unmanned aerial vehicle | |
JP2011112556A (en) | Search target position locating device, method, and computer program | |
CN105424059B (en) | Wide baseline near infrared camera position and orientation estimation method | |
CN113284239B (en) | Method and device for manufacturing electronic sand table of smart city |
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 |