CN108491498A - A kind of bayonet image object searching method based on multiple features detection - Google Patents
A kind of bayonet image object searching method based on multiple features detection Download PDFInfo
- Publication number
- CN108491498A CN108491498A CN201810227605.2A CN201810227605A CN108491498A CN 108491498 A CN108491498 A CN 108491498A CN 201810227605 A CN201810227605 A CN 201810227605A CN 108491498 A CN108491498 A CN 108491498A
- Authority
- CN
- China
- Prior art keywords
- image
- characteristic point
- bayonet
- region
- target signature
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Abstract
The invention discloses a kind of bayonet image object searching methods based on multiple features detection.The present invention includes the following steps:Target vehicle image is obtained, chooses more apparent feature on target vehicle manually, interception this feature image is as target signature image;The matching characteristic point in target signature image and bayonet image is detected, and screens the higher characteristic point of matching degree;The correlated characteristic point group and the candidate matches image-region in bayonet image for determining matching characteristic point;Respectively according to the quantity of characteristic point in correlated characteristic group, the similarity R of perceptual hash algorithm, greyscale color histogram calculation target signature image and candidate matches image-region1、R2、R3;To similarity R1、R2、R3Target signature image and the final similarity R of candidate matches image-region is weighted;Retrieve bayonet image successively according to above-mentioned steps, the bayonet optical sieving where candidate matches image-region by similarity value more than P comes out and shows.The present invention effectively reduces the investigation range of target vehicle.
Description
Technical field
The present invention relates to ITS Information technical fields, more particularly, to a kind of bayonet image based on multiple features detection
Target Searching Method.
Background technology
Car tracing is strike traffic offence, hit-and-run, the effective means for relating to the behaviors such as vehicle crime.Therefore, at some
Traffic block port can all install camera and acquire and preserve traffic image, to be checked when needed.
General bayonet platform by softwares such as Car license recognition, vehicle cab recognitions, vehicle by when acquire the vehicle
License plate number, vehicle model, preserve the vehicle by section position and the information such as specific time in case subsequent searches.But
It is that the identification of car plate and vehicle can have mistake, and common deck, false-trademark block the even more delinquent personnel of the behaviors such as car plate
Escape the conventional means of Plate searching.In addition, it is also more coarse for the means such as vehicle cab recognition classification are current, it can not be careful
Type of vehicle is divided, needs manually to be screened, it is time-consuming and laborious.
Invention content
In order to make up for the deficiencies of the prior art, the present invention provides a kind of bayonet image objects based on multiple features detection to search
Suo Fangfa.
The technical solution adopted by the present invention is:
A kind of bayonet image object searching method based on multiple features detection, includes the following steps:
Step 1:Target vehicle image is obtained, and chooses more apparent feature on target vehicle manually, intercepts this feature
Image is as target signature image;
Step 2:The matching characteristic point in target signature image and bayonet image is detected by SIFT feature matching algorithm, and
Screen the higher characteristic point of matching degree;
Step 3:It determines the correlated characteristic point group of matching characteristic point, and is determined in bayonet image according to correlated characteristic point group
Candidate matches image-region;
Step 4:Target signature image and candidate matches image-region are calculated according to the quantity of characteristic point in correlated characteristic group
Similarity R1;
Step 5:The similarity R of target signature image and candidate matches image-region is calculated with perceptual hash algorithm2;
Step 6:The greyscale color histogram of target signature image and candidate matches image-region is calculated, and according to gray scale face
Color Histogram calculates the similarity R of target signature image and candidate matches image-region3;
Step 7:To similarity R1、R2、R3It is weighted, calculates target signature image and candidate matches image-region is final
Similarity R;
Step 8:It retrieves bayonet image successively according to above-mentioned steps, similarity value is more than to the candidate matches image-region of P
The bayonet optical sieving at place comes out and is arranged in order display according to the descending sequence of similarity value.
Preferably, when in the step 1 to target vehicle feature selecting, the multiple targets chosen manually on target vehicle are special
Levy image.Wherein, choosing multiple target vehicle features manually significantly more efficient can utilize known target information of vehicles, with more
Distinct feature increases the accuracy of search.
Preferably, in the step 2, the method for screening the higher characteristic point of matching degree includes the following steps:
Step a:The matching characteristic point in target signature image and bayonet image, system are detected by SIFT feature matching algorithm
Meter obtains the quantity N of matching characteristic point1Minimum euclidean distance L between all matching characteristic pointsmin;
Step b:According to the texture feature of target to be searched characteristic image and bayonet image, size and clarity, set respectively
Set a factor alpha more than 11With a factor alpha less than 12, statistics Euclidean distance is less than α1×LminMatching characteristic point, such as
The quantity of fruit matching characteristic point is less than α2×N1, then it is assumed that two images mismatch, and then proceed to screen next bayonet image,
Otherwise step c is carried out;
Step c:One matching characteristic point maximum of setting retains amount threshold N2If meeting the matching of Euclidean distance condition
The quantity of characteristic point is more than α2×N1, continue to judge whether its quantity is more than N2If the quantity of matching characteristic point is more than N2, then
Only choose the preceding N of Euclidean distance minimum2A characteristic point carries out subsequent processing, and otherwise reservation matching characteristic is counted out constant.
Suitable higher characteristic point of matching degree of choosing can reduce the calculation amount of subsequent algorithm, while to being unsatisfactory for feature
The bayonet picture of point quantitative requirement is excluded, and picture retrieval speed is increased.
Preferably, the correlated characteristic point group in the step 3 refers to the collection being made of the correlated characteristic point more than or equal to 3
It closes, determines that method is as follows:
Step a:3 characteristic points to match in arbitrary connects bayonet socket image and target signature image, composition one is by spy
Sign point is the triangle on vertex;
Step b:Calculate the angle of the correspondence vertex angle for the characteristic point that matches in bayonet image and target signature image
Difference, if the difference of the angle of each corresponding apex angle is both less than threshold value T1, then it is assumed that these three points are correlated characteristic point, and all
Correlated characteristic point is known as correlated characteristic point group.
It finds correlated characteristic point group and is conducive to exclude the erroneous judgement because caused by similar with target image characteristics, and keep
Target image invariable rotary property, effect are preferable.
Preferably, in the step 3, the candidate matches image-region in bayonet image is determined according to correlated characteristic point group
Method includes the following steps:
Step a:The external spin moment of minimum of all correlated characteristic points in bayonet image and target signature image is found out respectively
Shape simultaneously calculates separately its perimeter, and the minimum of the perimeter and target signature image of the external rotation rectangle of the minimum of bayonet image is external
Rotate the length of sides of the ratio s of the perimeter of rectangle as candidate matches image-region in bayonet image relative to target signature image
Scale multiple;
Step b:Target signature image is fitted on bayonet image, the center phase mutual respect of minimum rotation rectangle each other is made
Close, and four sides for rotating rectangle correspond to respectively it is parallel;
Step c:Equal proportion s expands the target signature image length of side, and the image-region after will be enlarged by is got the bid in bayonet image
Go out as candidate matches image-region.
It determines that candidate matches image-region can accurately determine position of the object region in bayonet image, reduces
Deviations, while other region interference on bayonet image are eliminated indirectly, contribute to the calculation amount for reducing subsequent processing, improves
Accuracy.
Preferably, in the step 4, target signature image and candidate matches image are calculated according to the quantity of correlated characteristic point
The similarity R in region1Calculation formula it is as follows:
Wherein, N2Retain quantity for matching characteristic point maximum, N is phase in the correlated characteristic point group of candidate matches image-region
Close the quantity of characteristic point.
Preferably, in the step 6, according to greyscale color histogram calculation target signature image and candidate matches image district
The similarity R in domain3It is as follows:
Step a:The greyscale color histogram of target signature image and candidate matches image-region is total to from 0-255 respectively
256 gray levels are divided into 64 regions, and each region is continuous 4 gray levels;
Step b:Calculate separately the sum of the pixel quantity of each gray level region, and by target signature image and candidate matches
The sum of pixel quantity in 64 regions divided in image-region greyscale color histogram is according to ascending suitable of gray level
Sequence arranges, and is respectively seen as 64 dimensional vectors, is counted as M and H;
Step c:Two folder cosine of an angles between vector M and H are calculated according to the cosine formula of vector, cosine value represents mesh
Mark the similarity R of characteristic image and candidate matches image-region3:
Wherein, M is [M1,M2,...,Mn], H is [H1,H2,...,Hn]。
Preferably, in the step 7, the calculation formula of target signature image and the final similarity R of candidate matches image
It is as follows:
R=α × R1+β×R2+γ×R3
Wherein, the value of α, β, γ meet the value of alpha+beta+γ=1, α, β, γ and can be made according to the difference of actual conditions
Go out adjustment.
It is detected using sift characteristic points, the spy of the methods of perceptual hash algorithm and greyscale color histogram similarity calculation
Point, according to actual conditions proper transformation proportionality coefficient, candidate matches image-region and target image in COMPREHENSIVE CALCULATING bayonet image
Similarity, enhance the judgement precision of picture search to a certain extent.
Compared with prior art, the present invention advantage is:
1, for the present invention by choosing target vehicle feature manually, sift characteristic matchings position candidate image area, comprehensive
Sift correlated characteristic point quantity compares, perceptual hash algorithm and greyscale color histogram Similarity algorithm are completed bayonet image object and searched
Rope.The present invention proposes that correlated characteristic point group and candidate matches image-region determine according to the rotational invariance of sift characteristic matchings
Method reduces the calculation amount of processing procedure in conjunction with the manual method chosen target image notable feature and scanned for, can be compared with
Fast and accurately to position possible target vehicle image.By sift correlated characteristic point quantity compare, perceptual hash algorithm and
The various ways such as greyscale color histogram Similarity algorithm consider, and calculate candidate image area and target vehicle characteristic image
Similarity point-device can identify the target vehicle image in magnanimity bayonet image;
2, the present invention utilizes a variety of image similarity computational algorithms, considers calculating matching image phase according to actual conditions
Like degree, target image can be more accurately determined;Meanwhile the characteristic image region for choosing target image manually scans for,
It can efficiently be scanned for for target vehicle, plain speed and accuracy rate are searched in raising.
Description of the drawings
Fig. 1:The present invention is based on the flow charts of the bayonet image object searching method of multiple features detection;
Fig. 2:Target vehicle image in the present invention;
Fig. 3:Target signature image in the present invention;
Fig. 4:Target signature image and bayonet image high-precision matching characteristic point line schematic diagram in the present invention;
Fig. 5:Target signature image correlated characteristic point and its minimum external rotation rectangle schematic diagram in the present invention;
Fig. 6:Bayonet image correlated characteristic point and its minimum external rotation rectangle schematic diagram in the present invention;
Fig. 7:Candidate matches image-region in the present invention.
Specific implementation mode
A kind of bayonet image object searching method based on multiple features detection of the present invention, this method flow is as shown in Figure 1, tool
Body is implemented according to the following steps:
Step 1:Target vehicle image is obtained first, and chooses more apparent feature on target vehicle manually, and interception should
Characteristic image is as target signature image.In the present embodiment, target vehicle image on target vehicle as shown in Fig. 2, intercept
Target signature image it is as shown in Figure 3.
Step 2:The matching characteristic point in target signature image and bayonet image is detected by SIFT feature matching algorithm, and
The higher characteristic point of matching degree is screened according to following method:
Step a:The matching characteristic point in target signature image and bayonet image, system are detected by SIFT feature matching algorithm
Meter obtains the quantity N of matching characteristic point1Minimum euclidean distance L between all matching characteristic pointsmin;
Step b:According to the texture feature of target to be searched characteristic image and bayonet image, size and clarity, set respectively
Set a factor alpha more than 11With a factor alpha less than 12, statistics Euclidean distance is less than α1×LminMatching characteristic point, such as
The quantity of fruit matching characteristic point is less than α2×N1, then it is assumed that two images mismatch, and then proceed to screen next bayonet image,
Otherwise step c is carried out;
Step c:One matching characteristic point maximum of setting retains amount threshold N2If meeting the matching of Euclidean distance condition
The quantity of characteristic point is more than α2×N1, continue to judge whether its quantity is more than N2If the quantity of matching characteristic point is more than N2, then
Only choose the preceding N of Euclidean distance minimum2A characteristic point carries out subsequent processing, and otherwise reservation matching characteristic is counted out constant.
In the present embodiment, the quantity of selected target signature image and the matching characteristic point of bayonet image is 190, is counted
The minimum euclidean distance gone out between all matching characteristic points is 63.1427, chooses factor alpha1It is 2, α2It is 0.1, statistics obtains European
The quantity of matching characteristic point of the distance less than 2 × 63.1427 is 64, is much larger than 0.1 × 190.Meanwhile setting matching characteristic point
It is 10 that maximum, which retains quantity, i.e., first 10 that Euclidean distance is minimum are selected from 64 matching characteristic for meeting required precision points
Characteristic point carries out subsequent processing.Fig. 4 is of target signature image and highest preceding 10 characteristic points of matching degree of bayonet image
With image.
Step 3:The correlated characteristic point group in bayonet image is calculated as follows:
Step a:3 characteristic points to match in arbitrary connects bayonet socket image and target signature image, composition one is by spy
Sign point is the triangle on vertex;
Step b:Calculate the angle of the correspondence vertex angle for the characteristic point that matches in bayonet image and target signature image
Difference, if the difference of the angle of each corresponding apex angle is both less than threshold value T1, then it is assumed that these three points are correlated characteristic point, and all
Correlated characteristic point is known as correlated characteristic point group.
The threshold value that the difference of the angle of each corresponding apex angle is arranged in the present embodiment is 10 degree, and can find the point met the requirements has
8.
Then, candidate matches image-region is determined as follows:
Step a:The external spin moment of minimum of all correlated characteristic points in bayonet image and target signature image is found out respectively
Shape simultaneously calculates separately its perimeter, and the minimum of the perimeter and target signature image of the external rotation rectangle of the minimum of bayonet image is external
Rotate the length of sides of the ratio s of the perimeter of rectangle as candidate matches image-region in bayonet image relative to target signature image
Scale multiple;
Step b:Target signature image is fitted on bayonet image, the center phase mutual respect of minimum rotation rectangle each other is made
Close, and four sides for rotating rectangle correspond to respectively it is parallel;
Step c:Equal proportion s expands the target signature image length of side, and the image-region after will be enlarged by is got the bid in bayonet image
Go out as candidate matches image-region.
8 correlated characteristic points of target signature image and bayonet image in the present embodiment, minimum rotation rectangle and minimum rotation
The center of torque shape marks in Fig. 5, Fig. 6 respectively, and it is 0.39 that the ratio between the perimeter of rotation rectangle s, which is calculated, above-mentioned according to this
Method intercepts candidate matches image-region, as shown in Figure 7.
Step 4:The similarity of target signature image and candidate matches image-region is calculated according to the quantity of correlated characteristic point
R1:
Wherein, N2Retain quantity for matching characteristic point maximum, N is phase in the correlated characteristic point group of candidate matches image-region
Close the quantity of characteristic point.In the present embodiment, the maximum of matching characteristic point retains quantity N2It is 10, the quantity N of correlated characteristic point
It is 8, the similarity R of target signature image and candidate matches image-region can be calculated1It is 0.8.
Step 5:The similarity R of target signature image and candidate matches image-region is calculated with perceptual hash algorithm2:This reality
Apply example by image according to the method and step of perceptual hash algorithm be reduced into 8 × 8 size and calculate its fingerprint, by comparison can
With know target signature image and candidate matches image-region identical fingerprints quantity be 49, can calculate its similarity is
0.76。
Step 6:According to the similarity R of greyscale color histogram calculation target signature image and candidate matches image-region3
The step of it is as follows:
Step a:The greyscale color histogram of target signature image and candidate matches image-region is total to from 0-255 respectively
256 gray levels are divided into 64 regions, and each region is continuous 4 gray levels;
Step b:Calculate separately the sum of the pixel quantity of each gray level region, and by target signature image and candidate matches
The sum of pixel quantity in 64 regions divided in image-region greyscale color histogram is according to ascending suitable of gray level
Sequence arranges, and is respectively seen as 64 dimensional vectors, is counted as M and H;
Step c:Two folder cosine of an angles between vector M and H are calculated according to the cosine formula of vector, cosine value represents mesh
Mark the similarity R of characteristic image and candidate matches image-region3:
Wherein, M is [M1,M2,...,Mn], H is [H1,H2,...,Hn]。
The similarity of target signature image and candidate matches image-region in the present embodiment can be calculated to obtain by the above method
R3Equal to 0.32.
Step 7:Since hash algorithm is to the no robustness of the rotation of image, the image based on greyscale color histogram is similar
Property judge to be looked after be affected, the present embodiment be arranged α, β, γ be respectively 0.7,0.2,0.1, according to formula R=α × R1+β
×R2+γ×R3Its final similarity R can be calculated and be equal to 0.744.In the present invention, the image when similarity is more than 0.45
It can be assumed that being similar, actually searching multiple plain images can obtain;It can judge searched plain card when similarity is more than 0.6
Include target image in mouth image.
Step 8:It retrieves bayonet image successively according to above-mentioned steps, similarity value is more than to 0.5 candidate matches image district
Bayonet optical sieving where domain comes out and is arranged in order display according to the descending sequence of similarity value.
The foregoing is merely the preferable embodiments of the present invention, are not intended to limit the invention, all spirit in the present invention
Within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (8)
1. a kind of bayonet image object searching method based on multiple features detection, it is characterised in that:The method includes following steps
Suddenly:
Step 1:Target vehicle image is obtained, chooses more apparent feature on target vehicle manually, interception this feature image is made
For target signature image;
Step 2:The matching characteristic point in target signature image and bayonet image is detected by SIFT feature matching algorithm, and is screened
The higher characteristic point of matching degree;
Step 3:It determines the correlated characteristic point group of matching characteristic point, and the candidate in bayonet image is determined according to correlated characteristic point group
Match image-region;
Step 4:The phase of target signature image and candidate matches image-region is calculated according to the quantity of characteristic point in correlated characteristic group
Like degree R1;
Step 5:The similarity R of target signature image and candidate matches image-region is calculated with perceptual hash algorithm2;
Step 6:The greyscale color histogram of target signature image and candidate matches image-region is calculated, and straight according to greyscale color
Side's figure calculates the similarity R of target signature image and candidate matches image-region3;
Step 7:To similarity R1、R2、R3It is weighted, calculates target signature image and the final phase of candidate matches image-region
Like degree R;
Step 8:Bayonet image is retrieved successively according to above-mentioned steps, by similarity value more than where the candidate matches image-region of P
Bayonet optical sieving out and be arranged in order display according to the descending sequence of similarity value.
2. a kind of bayonet image object searching method based on multiple features detection according to claim 1, it is characterised in that:
In the step 1, when to target vehicle feature selecting, multiple target signature images on target vehicle are chosen manually.
3. a kind of bayonet image object searching method based on multiple features detection according to claim 1, it is characterised in that:
The method that the higher characteristic point of matching degree is screened in the step 2, includes the following steps:
Step a:The matching characteristic point in target signature image and bayonet image is detected by SIFT feature matching algorithm, is counted
Go out the quantity N of matching characteristic point1Minimum euclidean distance L between all matching characteristic pointsmin;
Step b:According to the texture feature of target to be searched characteristic image and bayonet image, size and clarity, it is respectively set one
It is a be more than 1 factor alpha1With a factor alpha less than 12, statistics Euclidean distance is less than α1×LminMatching characteristic point, if
Quantity with characteristic point is less than α2×N1, then it is assumed that two images mismatch, and then proceed to screen next bayonet image, otherwise
Carry out step c;
Step c:One matching characteristic point maximum of setting retains amount threshold N2If meeting the matching characteristic of Euclidean distance condition
The quantity of point is more than α2×N1, continue to judge whether its quantity is more than N2If the quantity of matching characteristic point is more than N2, then only select
Take the preceding N of Euclidean distance minimum2A characteristic point carries out subsequent processing, and otherwise reservation matching characteristic is counted out constant.
4. a kind of bayonet image object searching method based on multiple features detection according to claim 1, it is characterised in that:
Correlated characteristic point group in the step 3 refers to the set being made of the correlated characteristic point more than or equal to 3, determines method such as
Under:
Step a:3 characteristic points to match in arbitrary connects bayonet socket image and target signature image, composition one is by characteristic point
For the triangle on vertex;
Step b:The difference of the angle of the correspondence vertex angle for the characteristic point that matches in bayonet image and target signature image is calculated,
If the difference of the angle of each corresponding apex angle is both less than threshold value T1, then it is assumed that these three points are correlated characteristic point, and all correlations
Characteristic point is known as correlated characteristic point group.
5. a kind of bayonet image object searching method based on multiple features detection according to claim 1, it is characterised in that:
In the step 3, the method that the candidate matches image-region in bayonet image is determined according to correlated characteristic point group, including following step
Suddenly:
Step a:The external rotation rectangle of minimum of all correlated characteristic points in bayonet image and target signature image is found out respectively simultaneously
Its perimeter is calculated separately, by the external rotation of minimum of the perimeter of the external rotation rectangle of the minimum of bayonet image and target signature image
The ratio s of the perimeter of rectangle is as scaling of the candidate matches image-region relative to the length of side of target signature image in bayonet image
Multiple;
Step b:Target signature image is fitted on bayonet image, the center of minimum rotation rectangle each other is made to overlap, and
And rotate rectangle four sides correspond to respectively it is parallel;
Step c:Equal proportion s expands the target signature image length of side, and the image-region after will be enlarged by marks work in bayonet image
For candidate matches image-region.
6. a kind of bayonet image object searching method based on multiple features detection according to claim 3, it is characterised in that:
In the step 4, the similarity R of target signature image and candidate matches image-region is calculated according to the quantity of correlated characteristic point1
Calculation formula it is as follows:
Wherein, N2Retain quantity for matching characteristic point maximum, N is related special in the correlated characteristic point group of candidate matches image-region
Levy the quantity of point.
7. a kind of bayonet image object searching method based on multiple features detection according to claim 1, it is characterised in that:
In the step 6, according to the similarity R of greyscale color histogram calculation target signature image and candidate matches image-region3's
It is as follows:
Step a:Respectively by the greyscale color histogram of target signature image and candidate matches image-region from 0-255 totally 256
Gray level is divided into 64 regions, and each region is continuous 4 gray levels;
Step b:Calculate separately the sum of the pixel quantity of each gray level region, and by target signature image and candidate matches image
The sum of the pixel quantity in 64 regions divided in area grayscale color histogram is arranged according to the ascending sequence of gray level
Row, are respectively seen as 64 dimensional vectors, are counted as M and H;
Step c:Two folder cosine of an angles between vector M and H are calculated according to the cosine formula of vector, cosine value represents target spy
Levy the similarity R of image and candidate matches image-region3:
Wherein, M is [M1,M2,...,Mn], H is [H1,H2,...,Hn]。
8. a kind of bayonet image object searching method based on multiple features detection according to claim 1, it is characterised in that:
In the step 7, the calculation formula of target signature image and the final similarity R of candidate matches image-region are as follows:
R=α × R1+β×R2+γ×R3
Wherein, the value of α, β, γ meet alpha+beta+γ=1.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810227605.2A CN108491498B (en) | 2018-03-20 | 2018-03-20 | Bayonet image target searching method based on multi-feature detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810227605.2A CN108491498B (en) | 2018-03-20 | 2018-03-20 | Bayonet image target searching method based on multi-feature detection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108491498A true CN108491498A (en) | 2018-09-04 |
CN108491498B CN108491498B (en) | 2021-06-25 |
Family
ID=63318505
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810227605.2A Active CN108491498B (en) | 2018-03-20 | 2018-03-20 | Bayonet image target searching method based on multi-feature detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108491498B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109447023A (en) * | 2018-11-08 | 2019-03-08 | 北京奇艺世纪科技有限公司 | Determine method, video scene switching recognition methods and the device of image similarity |
CN110009000A (en) * | 2019-03-11 | 2019-07-12 | 东北大学 | The grain heap object detection method of sorting algorithm is improved based on ADABOOST+SVM |
CN110807473A (en) * | 2019-10-12 | 2020-02-18 | 浙江大华技术股份有限公司 | Target detection method, device and computer storage medium |
CN111222571A (en) * | 2020-01-06 | 2020-06-02 | 腾讯科技(深圳)有限公司 | Image special effect processing method and device, electronic equipment and storage medium |
CN111241332A (en) * | 2020-01-17 | 2020-06-05 | 珠海博明视觉科技有限公司 | Method for searching corresponding measuring program through automatic identification of workpiece |
CN111428122A (en) * | 2020-03-20 | 2020-07-17 | 南京中孚信息技术有限公司 | Picture retrieval method and device and electronic equipment |
CN112015935A (en) * | 2020-09-03 | 2020-12-01 | 科大讯飞股份有限公司 | Image searching method and device, electronic equipment and storage medium |
CN113077410A (en) * | 2020-01-03 | 2021-07-06 | 上海依图网络科技有限公司 | Image detection method, device and method, chip and computer readable storage medium |
CN117011553A (en) * | 2023-07-19 | 2023-11-07 | 苏州旭智设计营造有限公司 | Multi-scene conference free conversion control system |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7949186B2 (en) * | 2006-03-15 | 2011-05-24 | Massachusetts Institute Of Technology | Pyramid match kernel and related techniques |
US20110206237A1 (en) * | 2010-02-25 | 2011-08-25 | Canon Kabushiki Kaisha | Recognition apparatus and method thereof, and computer program |
US8738633B1 (en) * | 2012-01-31 | 2014-05-27 | Google Inc. | Transformation invariant media matching |
CN103927511A (en) * | 2014-02-25 | 2014-07-16 | 华北电力大学(保定) | Image identification method based on difference feature description |
CN105574063A (en) * | 2015-08-24 | 2016-05-11 | 西安电子科技大学 | Image retrieval method based on visual saliency |
-
2018
- 2018-03-20 CN CN201810227605.2A patent/CN108491498B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7949186B2 (en) * | 2006-03-15 | 2011-05-24 | Massachusetts Institute Of Technology | Pyramid match kernel and related techniques |
US20110206237A1 (en) * | 2010-02-25 | 2011-08-25 | Canon Kabushiki Kaisha | Recognition apparatus and method thereof, and computer program |
US8738633B1 (en) * | 2012-01-31 | 2014-05-27 | Google Inc. | Transformation invariant media matching |
CN103927511A (en) * | 2014-02-25 | 2014-07-16 | 华北电力大学(保定) | Image identification method based on difference feature description |
CN105574063A (en) * | 2015-08-24 | 2016-05-11 | 西安电子科技大学 | Image retrieval method based on visual saliency |
Non-Patent Citations (2)
Title |
---|
MICHAL BERETA: "Local descriptors and similarity measures for frontal face recognition: A comparative analysis", 《JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION》 * |
何敬: "基于特征点群相似度计算模型的图像表示方法", 《浙江大学学报》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109447023A (en) * | 2018-11-08 | 2019-03-08 | 北京奇艺世纪科技有限公司 | Determine method, video scene switching recognition methods and the device of image similarity |
CN110009000A (en) * | 2019-03-11 | 2019-07-12 | 东北大学 | The grain heap object detection method of sorting algorithm is improved based on ADABOOST+SVM |
CN110009000B (en) * | 2019-03-11 | 2022-09-02 | 东北大学 | Grain pile target detection method based on ADABOOST + SVM improved classification algorithm |
CN110807473A (en) * | 2019-10-12 | 2020-02-18 | 浙江大华技术股份有限公司 | Target detection method, device and computer storage medium |
CN110807473B (en) * | 2019-10-12 | 2023-01-03 | 浙江大华技术股份有限公司 | Target detection method, device and computer storage medium |
CN113077410A (en) * | 2020-01-03 | 2021-07-06 | 上海依图网络科技有限公司 | Image detection method, device and method, chip and computer readable storage medium |
CN111222571A (en) * | 2020-01-06 | 2020-06-02 | 腾讯科技(深圳)有限公司 | Image special effect processing method and device, electronic equipment and storage medium |
CN111222571B (en) * | 2020-01-06 | 2021-12-14 | 腾讯科技(深圳)有限公司 | Image special effect processing method and device, electronic equipment and storage medium |
CN111241332A (en) * | 2020-01-17 | 2020-06-05 | 珠海博明视觉科技有限公司 | Method for searching corresponding measuring program through automatic identification of workpiece |
CN111241332B (en) * | 2020-01-17 | 2023-08-18 | 珠海博明视觉科技有限公司 | Method for searching corresponding measurement program by automatically identifying workpiece |
CN111428122A (en) * | 2020-03-20 | 2020-07-17 | 南京中孚信息技术有限公司 | Picture retrieval method and device and electronic equipment |
CN111428122B (en) * | 2020-03-20 | 2023-09-01 | 南京中孚信息技术有限公司 | Picture retrieval method and device and electronic equipment |
CN112015935A (en) * | 2020-09-03 | 2020-12-01 | 科大讯飞股份有限公司 | Image searching method and device, electronic equipment and storage medium |
CN112015935B (en) * | 2020-09-03 | 2022-12-06 | 科大讯飞股份有限公司 | Image searching method and device, electronic equipment and storage medium |
CN117011553A (en) * | 2023-07-19 | 2023-11-07 | 苏州旭智设计营造有限公司 | Multi-scene conference free conversion control system |
Also Published As
Publication number | Publication date |
---|---|
CN108491498B (en) | 2021-06-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108491498A (en) | A kind of bayonet image object searching method based on multiple features detection | |
CN104700099B (en) | The method and apparatus for recognizing traffic sign | |
CN102426649B (en) | Simple steel seal digital automatic identification method with high accuracy rate | |
EP1600883B1 (en) | Object recognition system | |
CN103699905B (en) | Method and device for positioning license plate | |
CN106683076B (en) | The method of train wheel tread damage detection based on textural characteristics cluster | |
CN103035013B (en) | A kind of precise motion shadow detection method based on multi-feature fusion | |
JP4868530B2 (en) | Image recognition device | |
CN112686812B (en) | Bank card inclination correction detection method and device, readable storage medium and terminal | |
CN101142584A (en) | Method for facial features detection | |
CN109409208A (en) | A kind of vehicle characteristics extraction and matching process based on video | |
CN108319958A (en) | A kind of matched driving license of feature based fusion detects and recognition methods | |
CN106503638B (en) | Image procossing, vehicle color identification method and system for color identification | |
CN110866430A (en) | License plate recognition method and device | |
CN112417931B (en) | Method for detecting and classifying water surface objects based on visual saliency | |
CN107862319B (en) | Heterogeneous high-light optical image matching error eliminating method based on neighborhood voting | |
CN109146859A (en) | A kind of pavement crack detection system based on machine vision | |
CN110689003A (en) | Low-illumination imaging license plate recognition method and system, computer equipment and storage medium | |
CN110334760A (en) | A kind of optical component damage detecting method and system based on resUnet | |
CN115018785A (en) | Hoisting steel wire rope tension detection method based on visual vibration frequency identification | |
CN107808165B (en) | Infrared image matching method based on SUSAN corner detection | |
CN110766728B (en) | Combined image feature accurate matching method based on deep learning | |
CN109635679B (en) | Real-time target paper positioning and loop line identification method | |
CN115035281B (en) | Rapid infrared panoramic image stitching method | |
CN115457044B (en) | Pavement crack segmentation method based on class activation mapping |
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 | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20190930 Address after: High tech Zone Xinyu South Road Ji'nan city Shandong province 250000 first Qilu Software Park Building 6 room 610 Applicant after: Jinan Shenbo Information Technology Co., Ltd. Address before: 250000 room 2511, 1 building, Orson mansion, 1162 Xinjie street, hi tech Zone, Ji'nan, Shandong. Applicant before: Shandong magic Digital Technology Co., Ltd. |
|
TA01 | Transfer of patent application right | ||
GR01 | Patent grant | ||
GR01 | Patent grant |