CN109101982A - The recognition methods of target object and device - Google Patents
The recognition methods of target object and device Download PDFInfo
- Publication number
- CN109101982A CN109101982A CN201810835895.9A CN201810835895A CN109101982A CN 109101982 A CN109101982 A CN 109101982A CN 201810835895 A CN201810835895 A CN 201810835895A CN 109101982 A CN109101982 A CN 109101982A
- Authority
- CN
- China
- Prior art keywords
- edge point
- similarity measure
- images
- recognized
- direction vector
- 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
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
Abstract
The invention discloses a kind of recognition methods of target object and devices.Wherein, this method comprises: obtaining the direction vector of each first edge point in the first identification region in sample image, wherein include target object in the first identification region；Obtain the direction vector of each second edge point in the second identification region in images to be recognized；According to the direction vector of all first edge points in the direction vector of each second edge point and the first identification region, the similarity measure of each second edge point is obtained；Based on the similarity measure of each second edge point, the recognition result of images to be recognized is obtained, wherein whether recognition result is for characterizing in images to be recognized comprising target object.The technical issues of present invention solves the influence to the recognition methods of target object by ambient lighting in the prior art, causes recognition accuracy low and low efficiency.
Description
Technical field
The present invention relates to field of image recognition, recognition methods and device in particular to a kind of target object.
Background technique
Object identification orientation in automatic field, the introducing of machine vision keep a few thing stable, reliable, high
It is more intelligent while effect, eliminate manual operation the drawbacks of, worker can also be made to be detached from uninteresting single work.But it is actually answering
In, machine vision often generates unstable result due to various reasons.Such as in actual production, environment is extremely difficult to manage
The state thought, a large amount of illumination variation influence image procossing also very big.Not high deposit is standardized additionally, due to identified object itself
In irregular phenomenon, causes various interference to generate, influence discrimination.In this case, often using increasingly complex
Image processing method, processing speed will be greatly reduced, it is difficult to meet the certain pairs of high stations of efficiency requirements.In the group of four-way valve
It during dress, needs to distinguish stiffening plate front and back sides, has one in its front and stamp out " L " mark come.But due to reinforcing
Plate manufacturing process itself will lead to its surface and have various scratches, interfere to the identification of mark, in addition due to station sky
Between limitation, environment light can not be blocked, the variation of ambient lighting substantially reduces discrimination.
For above-mentioned problem, currently no effective solution has been proposed.
Summary of the invention
The embodiment of the invention provides a kind of recognition methods of target object and devices, right in the prior art at least to solve
The technical issues of recognition methods of target object is influenced by ambient lighting, causes recognition accuracy low and low efficiency.
According to an aspect of an embodiment of the present invention, a kind of recognition methods of target object is provided, comprising: obtain sample
In image in the first identification region each first edge point direction vector, wherein in the first identification region include target object；
Obtain the direction vector of each second edge point in the second identification region in images to be recognized；According to the side of each second edge point
The direction vector of all first edge points into the first identification region of vector sum, obtains the similarity measure of each second edge point；
Based on the similarity measure of each second edge point, obtain the recognition result of images to be recognized, wherein recognition result for characterize to
Identify in image whether include target object.
According to another aspect of an embodiment of the present invention, a kind of identification device of target object is additionally provided, comprising: first obtains
Modulus block, for obtaining the direction vector of each first edge point in the first identification region in sample image, wherein the first identification
It include target object in region；Second obtains module, for obtaining in images to be recognized each second side in the second identification region
The direction vector of edge point；First processing module, in the direction vector and the first identification region according to each second edge point
The direction vector of all first edge points obtains the similarity measure of each second edge point；Second processing module, for based on every
The similarity measure of a second edge point, obtains the recognition result of images to be recognized, wherein recognition result is for characterizing figure to be identified
It whether include target object as in.
According to another aspect of an embodiment of the present invention, a kind of storage medium is additionally provided, storage medium includes the journey of storage
Sequence, wherein equipment where control storage medium executes the recognition methods of above-mentioned target object in program operation.
According to another aspect of an embodiment of the present invention, a kind of processor is additionally provided, processor is used to run program,
In, program executes the recognition methods of above-mentioned target object when running.
In embodiments of the present invention, obtain the direction of each first edge point in the first identification region in sample image to
Amount, while the direction vector of each second edge point in the second identification region in images to be recognized is obtained, further according to each
The direction vector of all first edge points, obtains each second edge in the direction vector of second edge point and the first identification region
The similarity measure of point, so as to obtain the recognition result of images to be recognized according to similarity measure.Compared with prior art, from
In one identification region obtain first edge point direction vector, from the second identification region obtain second edge point direction to
Amount, and according to the direction vector of all first edge points in the direction vector of each second edge point and the first identification region, it obtains
To the similarity measure of each second edge point, to avoid being blocked, the influence of chaotic and non-linear illumination variation, reach
Discrimination is improved, recognition speed is promoted, vision system is made to have more the technical effect of robustness, and then is solved in the prior art
The technical issues of recognition methods of target object is influenced by ambient lighting, causes recognition accuracy low and low efficiency.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present invention, constitutes part of this application, this hair
Bright illustrative embodiments and their description are used to explain the present invention, and are not constituted improper limitations of the present invention.In the accompanying drawings:
Fig. 1 is a kind of flow chart of the recognition methods of target object according to an embodiment of the present invention；And
Fig. 2 is a kind of schematic diagram of the identification device of target object according to an embodiment of the present invention.
Specific embodiment
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work
It encloses.
It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, "
Two " etc. be to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that using in this way
Data be interchangeable under appropriate circumstances, so as to the embodiment of the present invention described herein can in addition to illustrating herein or
Sequence other than those of description is implemented.In addition, term " includes " and " having " and their any deformation, it is intended that cover
Cover it is non-exclusive include, for example, the process, method, system, product or equipment for containing a series of steps or units are not necessarily limited to
Step or unit those of is clearly listed, but may include be not clearly listed or for these process, methods, product
Or other step or units that equipment is intrinsic.
Embodiment 1
According to embodiments of the present invention, a kind of embodiment of the recognition methods of target object is provided, it should be noted that
The step of process of attached drawing illustrates can execute in a computer system such as a set of computer executable instructions, also,
It, in some cases, can be to be different from shown in sequence execution herein although logical order is shown in flow charts
The step of out or describing.
Fig. 1 is a kind of flow chart of the recognition methods of target object according to an embodiment of the present invention, as shown in Figure 1, the party
Method includes the following steps:
Step S102 obtains the direction vector of each first edge point in the first identification region in sample image, wherein the
It include target object in one identification region.
Specifically, above-mentioned sample image can be stiffening plate image, and above-mentioned target object can be stiffening plate front
" L " mark.For sample image, target object can be marked in sample image by way of manually marking.To reinforcement
The implementation that part front and back sides distinguish be identification image in whether comprising " L " identify, therefore, can only by reinforcer just
Face image is as sample image.
In order to avoid carrying out subsequent processing to entire sample image, promotion recognition efficiency can be previously according to reinforcer just
The position that " L " is identified in face, determines the first identification region, cuts to sample image, that is, available " L " identifies institute
In region.For images to be recognized, if can identify that " L " is identified in corresponding region, it can determine that the image is to add
Strong part direct picture may further determine reinforcer front；If in corresponding region it is unidentified go out " L " mark, can be with
It determines that the image is reinforcer verso images, may further determine reinforcer reverse side.
In a kind of optional scheme, for sample image, the sense comprising " L " mark can be extracted in sample image
Then interest region obtains the direction vector of each first edge point by boundary filter.Getting each first edge
After the direction vector of point, template can be expressed as to point set pi=(ri+ci)TAnd each direction vector di=(ti+ui)T, i
=1 ..., n.
Step S104 obtains the direction vector of each second edge point in the second identification region in images to be recognized.
Specifically, above-mentioned images to be recognized can be the image of the reinforcer arrived shot by capture apparatus, can be with
It is front or reverse side.In order to promote recognition efficiency, the second identification region, the second identification can be determined according to the first identification region
Region can be identical as the first identification region, can also be adjusted according to recognition result.
In a kind of optional scheme, for images to be recognized, in order to ensure in subsequent processes similarity measure not by
The influence with confusion is blocked, image can be filtered using boundary filter identical with template, obtain second edge point
Direction vector eR, c=(vR, c, wR, c)T。
Step S106, according to all first edge points in the direction vector of each second edge point and the first identification region
Direction vector obtains the similarity measure of each second edge point.
Specifically, similarity measure is used to characterize the identical percentage of images to be recognized and sample image, for example, figure to be identified
" L " as in is identified with 50% and blocks, then similarity measure does not exceed 0.5.Similarity measure is bigger, shows that the goodness of fit is higher,
It may thereby determine that being more possible in images to be recognized includes " L " mark.
It should be noted that existing Similarity Measures are the exhausted of difference between calculating sample image and images to be recognized
To the quadratic sum (SSD) of value summation (SAD) or all differences, and another kind NCC.First two scheme only light conditions not
It just can be used in the case where changing, last one kind can only also be applicable in linear look after and change, and can not accomplish not blocked, mix
Random and non-linear illumination variation.
To solve the above-mentioned problems, the present invention in the following method, in images to be recognized any point q=(r,
c)T, the dot product of the direction vector of the direction vector and respective point of all first edge points, is finding out average value in calculation template,
The similarity measure s of the available point.It, can be according to the side after normalization since direction vector is influenced by brightness of image
To vector, available similarity measure s ' is calculated.It further, can be to phase in order to adapt to nonlinear illumination variation
Likelihood metric s ' takes absolute value, and obtains eventually for the similarity measure s " for determining recognition result.
Step S108 obtains the recognition result of images to be recognized based on the similarity measure of each second edge point, wherein
Whether recognition result is for characterizing in images to be recognized comprising target object.
It specifically, can be by by second edges all in images to be recognized in order to obtain final recognition result
The similarity measure of point is compared with threshold value, when similarity measure is more than the threshold value, can determine images to be recognized and sample graph
As consistent, identified in images to be recognized comprising " L ", that is, images to be recognized is the direct picture of reinforcer.If all similar
Measurement is less than the threshold value, can determine that images to be recognized is inconsistent with sample image, does not mark comprising " L " in images to be recognized
Know, that is, images to be recognized is the verso images of reinforcer.
Using the above embodiment of the present invention, the direction of each first edge point in the first identification region in sample image is obtained
Vector, while the direction vector of each second edge point in the second identification region in images to be recognized is obtained, further according to every
The direction vector of all first edge points, obtains each second side in the direction vector of a second edge point and the first identification region
The similarity measure of edge point, so as to obtain the recognition result of images to be recognized according to similarity measure.Compared with prior art, from
In first identification region obtain first edge point direction vector, from the second identification region obtain second edge point direction to
Amount, and according to the direction vector of all first edge points in the direction vector of each second edge point and the first identification region, it obtains
To the similarity measure of each second edge point, to avoid being blocked, the influence of chaotic and non-linear illumination variation, reach
Discrimination is improved, recognition speed is promoted, vision system is made to have more the technical effect of robustness, and then is solved in the prior art
The technical issues of recognition methods of target object is influenced by ambient lighting, causes recognition accuracy low and low efficiency.
Optionally, in the above embodiment of the present invention, in step S106, according to the direction vector of each second edge point and
The direction vector of all first edge points in one identification region, before obtaining the similarity measure of each second edge point, this method
Further include: translation transformation is carried out to the direction vector of each first edge point, obtain each first edge point it is transformed to
Amount；According to the transformed vector of the direction vector of each second edge point and each first edge point, each second side is obtained
The similarity measure of edge point.
Specifically, affine transformation, all the points running transform along images to be recognized, that is, to the can be carried out to template
The direction vector of one marginal point carries out translation transformation, is then compared transformed template with images to be recognized, each
The position of marginal point calculates similarity measure.Above-mentioned translation transformation can be realized by the multiplication of translation matrix and direction vector.
Optionally, in the above embodiment of the present invention, according to the direction vector of each second edge point and each first edge
The transformed vector of point, obtains the similarity measure of each second edge point, comprising: obtain the direction of each second edge point to
The dot product of amount and the transformed vector of each first edge point；Obtain direction vector and the first identification of each second edge point
The average value of the dot product of the transformed vector of all first edge points in region, obtains the similarity of each second edge point
Amount.
Specifically, for any point q=(r, c) in images to be recognizedT, can calculate in transformed template and own
The dot product of the direction vector of respective point in the direction vector and images to be recognized of first edge point, then average value is found out as transformation
The similarity measure s that template afterwards is put at this.Formula is as follows:
Wherein, n is the quantity of first edge point, d 'iFor the transformed vector of i-th of first edge point, eq+p′For this point
Direction vector.
Optionally, in the above embodiment of the present invention, in the direction vector for obtaining each second edge point and each first side
Before the dot product of the transformed vector of edge point, method further include: to the direction vector and each first of each second edge point
The transformed vector of marginal point is normalized, obtain each second edge point normalized vector and each first side
The normalized vector of edge point；Obtain the normalized vector of each second edge point and the normalized vector of each first edge point
Dot product.
Specifically, since the direction vector got is influenced by brightness of image, similarity measure s cannot not light completely
According to the influence of variation, in order to improve this point, direction vector can be normalized, and calculate new similarity measure
S ', formula are as follows:
Wherein, | | | | indicate vector field homoemorphism.
Optionally, in the above embodiment of the present invention, the direction vector and the first identification region of each second edge point are obtained
The average value of the dot product of the transformed vector of interior all first edge points obtains the similarity measure of each second edge point, packet
It includes: obtaining the average value of the dot product of the normalized vector of each second edge point and the normalized vector of each first edge point,
Obtain the initial similarity measure of each second edge point；The absolute value for obtaining the initial similarity measure of each second edge point, obtains
To the similarity measure of each second edge point.
Specifically, it in order to adapt to nonlinear illumination variation, takes absolute value and asks to the similarity measure s ' being calculated before
Final similarity measure s ", while again not by blocking and chaotic influenced.Formula is as follows:
Wherein, | | indicate absolute value.
It should be noted that similarity measure s " is the numerical value less than 1, for indicating sample image and images to be recognized
Identical percentage, s "=1 indicates that template fits like a glove with image, and opposite its closer to 0 indicates that the two is more inconsistent.
Optionally, in the above embodiment of the present invention, step S108 is obtained based on the similarity measure of each second edge point
The recognition result of images to be recognized, comprising: judge in the second identification region whether in similarity measure to be greater than or equal to preset threshold
Second edge point；If the second edge point in the second identification region there are similarity measure more than or equal to preset threshold,
Whether the similarity measure for judging second edge point is local maximum in predeterminable area, wherein second edge point is preset areas
The central point in domain, the similarity measure that local maximum is used to characterize second edge point are greater than other second edge points in predeterminable area
Similarity measure；If the similarity measure of second edge point is the local maximum in predeterminable area, it is determined that recognition result is
It include target object in images to be recognized；If there is no similarity measures to be greater than or equal to preset threshold in the second identification region
The similarity measure of second edge point or second edge point is not the local maximum in predeterminable area, it is determined that recognition result is
It does not include target object in images to be recognized.
Specifically, above-mentioned preset threshold can be according to experiment, pre-set for determining images to be recognized and sample
The whether consistent threshold value s " of this imagemin。
In a kind of optional scheme, the similarity measure of each second edge point in the second identification region can be successively calculated
S ", then by each similarity measure s " with threshold value s "minIt is compared, when some second edge point position in images to be recognized
On, obtain s " > s "min, while similarity measure s " is also local maximum, then can determine and recognize " L " mark, namely really
Determine to identify in images to be recognized comprising " L ", is the direct picture of reinforcer；If all second edge points in images to be recognized
On position, s " > s " is not obtainedmin, or be not local maximum, then it can determine that unidentified arrive " L " is identified, namely determine
It is not identified comprising " L " in images to be recognized, is the verso images of reinforcer.
Optionally, it in the above embodiment of the present invention, in step S104, obtains every in the second identification region in images to be recognized
After the direction vector of a second edge point, this method further include: according to any one second edge point in the second identification region
Direction vector and the first identification region in all first edge points direction vector, obtain the phase of any one second edge point
Likelihood metric；Judge whether the similarity measure of any one second edge point is greater than or equal to preset threshold；If any one
The similarity measure of two marginal points be greater than or equal to preset threshold, then judge any one second edge point similarity measure whether be
Local maximum in predeterminable area；If the similarity measure of any one second edge point is less than preset threshold or any one
The similarity measure of a second edge point is not the local maximum in predeterminable area, then using next second edge point as any
One second edge point, and return to execution and own according in the direction vector of any one second edge point and the first identification region
The direction vector of first edge point, the step of obtaining the similarity measure of any one second edge point, until next second side
Edge point is the last one second edge point in the second identification region；If the similarity measure of any one second edge point is pre-
If the local maximum in region, it is determined that recognition result is in images to be recognized comprising target object；If the last one
It is not in predeterminable area that the similarity measure of two marginal points, which is less than preset threshold or the similarity measure of the last one second edge point,
Local maximum, it is determined that recognition result is in images to be recognized not comprising target object.
It specifically,, can be from first second edge in order to improve recognition speed since the calculation amount of similarity measure is huge
Point starts, and successively calculates the similarity measure to j-th of second edge pointIf at this point, calculating
The s arrivedj> s "min, while similarity measure sjIt is also local maximum, then can stops all second edge points after calculating
Similarity measure, and determine recognize " L " mark, namely determine images to be recognized in comprising " L " identify, be the front of reinforcer
Image；If the similarity measure is unsatisfactory for sj> s "min, or be not local maximum, then it can calculate next second edge
The similarity measure of point, if the similarity measure for the last one the second edge point being calculated also is unsatisfactory for sj> s "min, or
It is not local maximum, then can determines that unidentified does not include that " L " is identified into " L " mark, namely determining images to be recognized, be
The verso images of reinforcer.
Optionally, in the above embodiment of the present invention, images to be recognized is in the image pyramid based on original image building
Any one tomographic image, wherein the resolution ratio of every tomographic image is different in image pyramid, corresponding second identification region of every tomographic image
Difference, the corresponding preset threshold of every tomographic image are different.
Specifically, above-mentioned original image can be the image shot to reinforcer, in order to efficiently search for target
Object can create image pyramid for original image.Second edge pixel is obtained in every tomographic image of image pyramid
Direction vector.Above-mentioned identification process can be carried out in the every tomographic image of image pyramid, it therefore, can be every by image pyramid
Tomographic image is handled as images to be recognized.Since the resolution ratio of different levels image is different, in order to ensure discrimination, from every
The second identification region extracted in tomographic image is different, and the preset threshold for obtaining recognition result is also different.
Optionally, it in the above embodiment of the present invention, is obtained in step S108 based on the similarity measure of each second edge point
To after the recognition result of images to be recognized, this method further include: any one tomographic image recognition result be images to be recognized
In comprising being obtained down using next tomographic image as images to be recognized according to target object present position in the case where target object
One tomographic image corresponding second is and to be returned to execution by region and obtained in images to be recognized each second side in the second identification region
The direction vector of edge point, obtains the recognition result of next tomographic image, until images to be recognized is the last layer in image pyramid
Image；The recognition result of the last layer image be any one tomographic image in include target object in the case where, it is determined that it is original
It include target object in image；It is the feelings for not including target object in any one tomographic image in the recognition result of any one tomographic image
Under condition, it is determined that do not include target object in original image.
In a kind of optional scheme, since image pyramid is middle-level higher, resolution ratio is lower, therefore, can be in height
Threshold value s " in layerminIt is arranged lower, the of next layer can be determined according to the position for searching " L " expression in high-rise image
Two identification regions, then calculate similarity measure in the area.It circuits sequentially down, until search is indicated less than " L ", or arrives
Up to the bottom of image pyramid.So as to efficiently search accurate " L " mark.
Optionally, in the above embodiment of the present invention, step S102 is obtained in sample image each the in the first identification region
The direction vector of one marginal point, comprising: the first identification region is filtered using boundary filter, obtains each first
The direction vector of marginal point, wherein during being filtered to the first identification region, at Threshold segmentation
Reason；Step S104 obtains the direction vector of each second edge point in the second identification region in images to be recognized, comprising: utilize
Boundary filter is filtered the second identification region, obtains the direction vector of each second edge point.
Specifically, it is affected since Threshold segmentation is illuminated by the light, know to first extracted from sample image
When other region carries out edge filter processing, without Threshold segmentation, the direction vector of each first edge point is only obtained.Equally
Ground, when carrying out edge filter processing to the second identification region for extracting from images to be recognized, also without Threshold segmentation,
Only obtain the direction vector of each second edge point.
Through the above scheme, a kind of quick, stable image processing method is provided, there are scratches reinforcing plate surface
(confusion), mark is unintelligible, in the case where ambient lighting variation greatly, quick, stable " L " mark searched in stiffening plate,
To identify the front and back sides of stiffening plate, discrimination is improved, recognition speed is accelerated, makes vision system with more robustness.
Embodiment 2
According to embodiments of the present invention, a kind of embodiment of the identification device of target object is provided.
Fig. 2 is a kind of schematic diagram of the identification device of target object according to an embodiment of the present invention, as shown in Fig. 2, the dress
It sets and includes:
First obtains module 22, for obtain the direction of each first edge point in the first identification region in sample image to
Amount, wherein include target object in the first identification region.
Specifically, above-mentioned sample image can be stiffening plate image, and above-mentioned target object can be stiffening plate front
" L " mark.For sample image, target object can be marked in sample image by way of manually marking.To reinforcement
The implementation that part front and back sides distinguish be identification image in whether comprising " L " identify, therefore, can only by reinforcer just
Face image is as sample image.
In order to avoid carrying out subsequent processing to entire sample image, promotion recognition efficiency can be previously according to reinforcer just
The position that " L " is identified in face, determines the first identification region, cuts to sample image, that is, available " L " identifies institute
In region.For images to be recognized, if can identify that " L " is identified in corresponding region, it can determine that the image is to add
Strong part direct picture may further determine reinforcer front；If in corresponding region it is unidentified go out " L " mark, can be with
It determines that the image is reinforcer verso images, may further determine reinforcer reverse side.
In a kind of optional scheme, for sample image, the sense comprising " L " mark can be extracted in sample image
Then interest region obtains the direction vector of each first edge point by boundary filter.Getting each first edge
After the direction vector of point, template can be expressed as to point set pi=(ri+ci)TAnd each direction vector di=(ti+ui)T, i
=1 ..., n.
Second obtains module 24, for obtaining the direction of each second edge point in the second identification region in images to be recognized
Vector.
Specifically, above-mentioned images to be recognized can be the image of the reinforcer arrived shot by capture apparatus, can be with
It is front or reverse side.In order to promote recognition efficiency, the second identification region, the second identification can be determined according to the first identification region
Region can be identical as the first identification region, can also be adjusted according to recognition result.
In a kind of optional scheme, for images to be recognized, in order to ensure in subsequent processes similarity measure not by
The influence with confusion is blocked, image can be filtered using boundary filter identical with template, obtain second edge point
Direction vector eR, c=(vR, c, wR, c)T。
First processing module 26, for according in the direction vector of each second edge point and the first identification region all
The direction vector of one marginal point obtains the similarity measure of each second edge point.
Specifically, similarity measure is used to characterize the identical percentage of images to be recognized and sample image, for example, figure to be identified
" L " as in is identified with 50% and blocks, then similarity measure does not exceed 0.5.Similarity measure is bigger, shows that the goodness of fit is higher,
It may thereby determine that being more possible in images to be recognized includes " L " mark.
It should be noted that existing Similarity Measures are the exhausted of difference between calculating sample image and images to be recognized
To the quadratic sum (SSD) of value summation (SAD) or all differences, and another kind NCC.First two scheme only light conditions not
It just can be used in the case where changing, last one kind can only also be applicable in linear look after and change, and can not accomplish not blocked, mix
Random and non-linear illumination variation.
To solve the above-mentioned problems, the present invention in the following method, in images to be recognized any point q=(r,
c)T, the dot product of the direction vector of the direction vector and respective point of all first edge points, is finding out average value in calculation template,
The similarity measure s of the available point.It, can be according to the side after normalization since direction vector is influenced by brightness of image
To vector, available similarity measure s ' is calculated.It further, can be to phase in order to adapt to nonlinear illumination variation
Likelihood metric s ' takes absolute value, and obtains eventually for the similarity measure s " for determining recognition result.
Second processing module 28 obtains the identification of images to be recognized for the similarity measure based on each second edge point
As a result, wherein whether recognition result is for characterizing in images to be recognized comprising target object.
It specifically, can be by by second edges all in images to be recognized in order to obtain final recognition result
The similarity measure of point is compared with threshold value, when similarity measure is more than the threshold value, can determine images to be recognized and sample graph
As consistent, identified in images to be recognized comprising " L ", that is, images to be recognized is the direct picture of reinforcer.If all similar
Measurement is less than the threshold value, can determine that images to be recognized is inconsistent with sample image, does not mark comprising " L " in images to be recognized
Know, that is, images to be recognized is the verso images of reinforcer.
Using the above embodiment of the present invention, the direction of each first edge point in the first identification region in sample image is obtained
Vector, while the direction vector of each second edge point in the second identification region in images to be recognized is obtained, further according to every
The direction vector of all first edge points, obtains each second side in the direction vector of a second edge point and the first identification region
The similarity measure of edge point, so as to obtain the recognition result of images to be recognized according to similarity measure.Compared with prior art, from
In first identification region obtain first edge point direction vector, from the second identification region obtain second edge point direction to
Amount, and according to the direction vector of all first edge points in the direction vector of each second edge point and the first identification region, it obtains
To the similarity measure of each second edge point, to avoid being blocked, the influence of chaotic and non-linear illumination variation, reach
Discrimination is improved, recognition speed is promoted, vision system is made to have more the technical effect of robustness, and then is solved in the prior art
The technical issues of recognition methods of target object is influenced by ambient lighting, causes recognition accuracy low and low efficiency.
Embodiment 3
According to embodiments of the present invention, a kind of embodiment of storage medium is provided, storage medium includes the program of storage,
In, in program operation, equipment where control storage medium executes the recognition methods of the target object in above-described embodiment 1.
Embodiment 4
According to embodiments of the present invention, a kind of embodiment of processor is provided, processor is for running program, wherein journey
The recognition methods of the target object in above-described embodiment 1 is executed when sort run.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
In the above embodiment of the invention, it all emphasizes particularly on different fields to the description of each embodiment, does not have in some embodiment
The part of detailed description, reference can be made to the related descriptions of other embodiments.
In several embodiments provided herein, it should be understood that disclosed technology contents can pass through others
Mode is realized.Wherein, the apparatus embodiments described above are merely exemplary, such as the division of the unit, Ke Yiwei
A kind of logical function partition, there may be another division manner in actual implementation, for example, multiple units or components can combine or
Person is desirably integrated into another system, or some features can be ignored or not executed.Another point, shown or discussed is mutual
Between coupling, direct-coupling or communication connection can be through some interfaces, the INDIRECT COUPLING or communication link of unit or module
It connects, can be electrical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
On unit.It can some or all of the units may be selected to achieve the purpose of the solution of this embodiment according to the actual needs.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can for personal computer, server or network equipment etc.) execute each embodiment the method for the present invention whole or
Part steps.And storage medium above-mentioned includes: that USB flash disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
Reservoir (RAM, Random Access Memory), mobile hard disk, magnetic or disk etc. be various to can store program code
Medium.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (13)
1. a kind of recognition methods of target object characterized by comprising
Obtain the direction vector of each first edge point in template, wherein the template is the extracted from sample image
One identification region includes target object in the template；
Obtain the direction vector of each second edge point in the second identification region in images to be recognized；
According to the direction vector of the direction vector of each first edge point and corresponding second edge point, obtain described wait know
The similarity measure of other image, wherein the similarity measure is used to characterize the similarity of the images to be recognized Yu the template；
Based on the similarity measure of the images to be recognized, the recognition result of the images to be recognized is obtained, wherein the identification knot
Whether fruit is for characterizing in the images to be recognized comprising the target object.
2. the method according to claim 1, wherein according to the direction vector of each first edge point and
The direction vector of corresponding second edge point, before obtaining the similarity measure of the images to be recognized, the method also includes:
Translation transformation is carried out to the direction vector of each first edge point, after obtaining the transformation of each first edge point
Vector；
According to the direction vector of the corresponding second edge point of transformed vector sum of each first edge point, obtain described
Similarity measure.
3. according to the method described in claim 2, it is characterized in that, according to the transformed vector of each first edge point
With the direction vector of corresponding second edge point, the similarity measure is obtained, comprising:
The dot product for obtaining the direction vector of the corresponding second edge point of transformed vector sum of each first edge point, obtains
To the corresponding dot product of each first edge point；
The average value for obtaining the corresponding dot product of all first edge points in the template, obtains the similarity measure.
4. according to the method described in claim 3, it is characterized in that, the normalization for obtaining the corresponding second edge point to
The dot product of amount, before obtaining the corresponding dot product of each first edge point, the method also includes:
The transformed vector of each first edge point is normalized, and to the corresponding second edge point
Direction vector be normalized, obtain each first edge point normalized vector and corresponding second side
The normalized vector of edge point；
Obtain the point of the normalized vector of each first edge point and the normalized vector of the corresponding second edge point
Product, obtains the corresponding dot product of each first edge point.
5. according to the method described in claim 4, it is characterized in that, obtaining the corresponding point of all first edge points in the template
Long-pending average value obtains the similarity measure, comprising:
It is corresponding absolutely to obtain each first edge point for the absolute value for obtaining the corresponding dot product of each first edge point
Value；
The average value for obtaining the corresponding absolute value of each first edge point, obtains the similarity measure.
6. the method according to claim 1, wherein the similarity measure based on the images to be recognized, obtains institute
State the recognition result of images to be recognized, comprising:
Whether judge in second identification region in second edge point of the similarity measure more than or equal to preset threshold；
If sentenced in second identification region there are the second edge point that similarity measure is greater than or equal to the preset threshold
Whether the similarity measure of the second edge point of breaking is local maximum in predeterminable area, wherein the second edge point is
The central point of the predeterminable area, the similarity measure that the local maximum is used to characterize the second edge point are greater than described pre-
If the similarity measure of other second edge points in region；
If the similarity measure of the second edge point is the local maximum in the predeterminable area, it is determined that the identification knot
Fruit is in the images to be recognized comprising the target object；
If there is no the second edge points that similarity measure is greater than or equal to the preset threshold in second identification region, or
The similarity measure of the second edge point is not the local maximum in the predeterminable area, it is determined that the recognition result is institute
It states in images to be recognized not comprising the target object.
7. the method according to claim 1, wherein each in the second identification region in obtaining images to be recognized
After the direction vector of second edge point, the method also includes:
According to institute in the direction vector of any one second edge point in second identification region and first identification region
There is the direction vector of first edge point, obtains the similarity measure of any one second edge point；
Judge whether the similarity measure of any one second edge point is greater than or equal to preset threshold；
If the similarity measure of any one second edge point is greater than or equal to the preset threshold, judge described any
Whether the similarity measure of one second edge point is local maximum in predeterminable area；
If the similarity measure of any one second edge point is less than the preset threshold or any one described second side
The similarity measure of edge point is not the local maximum in the predeterminable area, then using next second edge point as described any
One second edge point, and execution is returned according in the direction vector and first identification region of any one second edge point
The direction vector of all first edge points, the step of obtaining the similarity measure of any one second edge point, until described
Next second edge point is the last one second edge point in second identification region；
If the similarity measure of any one second edge point is the local maximum in the predeterminable area, it is determined that institute
Stating recognition result is in the images to be recognized comprising the target object；
If the similarity measure of the last one second edge point is less than the preset threshold or the last one described second side
The similarity measure of edge point is not the local maximum in the predeterminable area, it is determined that the recognition result is the figure to be identified
It does not include the target object as in.
8. the method according to claim 1, wherein the images to be recognized is the figure based on original image building
As any one tomographic image in pyramid, wherein the resolution ratio of every tomographic image is different in described image pyramid, every tomographic image
Corresponding second identification region is different, and the corresponding preset threshold of the every tomographic image is different.
9. according to the method described in claim 8, it is characterized in that, in the similarity measure based on each second edge point,
After obtaining the recognition result of the images to be recognized, the method also includes:
The recognition result of any one tomographic image be the images to be recognized in include the target object in the case where, will
It is corresponding to obtain next tomographic image according to the target object present position as the images to be recognized for next tomographic image
Second be by region, and return execute the direction for obtaining each second edge point in the second identification region in images to be recognized to
Amount, obtains the recognition result of next tomographic image, until the images to be recognized is the last layer in described image pyramid
Image；
The recognition result of the last layer image be any one tomographic image in include the target object in the case where,
It then determines in the original image comprising the target object；
The case where the recognition result of any one tomographic image is not include the target object in any one tomographic image
Under, it is determined that do not include the target object in the original image.
10. the method according to claim 1, wherein
Obtain the direction vector of each first edge point in the first identification region in sample image, comprising: utilize boundary filter
First identification region is filtered, obtains the direction vector of each first edge point, wherein to described
During first identification region is filtered, without Threshold segmentation processing；
Obtain the direction vector of each second edge point in the second identification region in images to be recognized, comprising: utilize the edge
Filter is filtered second identification region, obtains the direction vector of each second edge point.
11. a kind of identification device of target object characterized by comprising
First obtains module, for obtaining the direction vector of each first edge point in the first identification region in sample image,
In, it include target object in first identification region；
Second obtains module, for obtaining the direction vector of each second edge point in the second identification region in images to be recognized；
First processing module, for owning according in the direction vector of each second edge point and first identification region
The direction vector of first edge point obtains the similarity measure of each second edge point；
Second processing module obtains the knowledge of the images to be recognized for the similarity measure based on each second edge point
Other result, wherein whether the recognition result is for characterizing in the images to be recognized comprising the target object.
12. a kind of storage medium, which is characterized in that the storage medium includes the program of storage, wherein run in described program
When control the storage medium where equipment perform claim require any one of 1 to 9 described in target object recognition methods.
13. a kind of processor, which is characterized in that the processor is for running program, wherein right of execution when described program is run
Benefit require any one of 1 to 10 described in target object recognition methods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810835895.9A CN109101982B (en) | 2018-07-26 | 2018-07-26 | Target object identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810835895.9A CN109101982B (en) | 2018-07-26 | 2018-07-26 | Target object identification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109101982A true CN109101982A (en) | 2018-12-28 |
| CN109101982B CN109101982B (en) | 2022-02-25 |
Family
ID=64847715
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810835895.9A Active CN109101982B (en) | 2018-07-26 | 2018-07-26 | Target object identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109101982B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090304285A1 (en) * | 2006-07-17 | 2009-12-10 | Panasonic Corporation | Image processing device and image processing method |
| CN105261012A (en) * | 2015-09-25 | 2016-01-20 | 上海瑞伯德智能系统科技有限公司 | Template matching method based on Sobel vectors |
| CN105760842A (en) * | 2016-02-26 | 2016-07-13 | 北京大学 | Station caption identification method based on combination of edge and texture features |
| CN105930858A (en) * | 2016-04-06 | 2016-09-07 | 吴晓军 | Fast high-precision geometric template matching method enabling rotation and scaling functions |
| CN106485701A (en) * | 2016-09-26 | 2017-03-08 | 成都交大光芒科技股份有限公司 | Based on the whether anti-loaded detection method of the railway overhead contact system catenary seat of image |
| CN107671896A (en) * | 2017-05-19 | 2018-02-09 | 重庆誉鸣科技有限公司 | Fast vision localization method and system based on SCARA robots |
-
2018
- 2018-07-26 CN CN201810835895.9A patent/CN109101982B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090304285A1 (en) * | 2006-07-17 | 2009-12-10 | Panasonic Corporation | Image processing device and image processing method |
| CN105261012A (en) * | 2015-09-25 | 2016-01-20 | 上海瑞伯德智能系统科技有限公司 | Template matching method based on Sobel vectors |
| CN105760842A (en) * | 2016-02-26 | 2016-07-13 | 北京大学 | Station caption identification method based on combination of edge and texture features |
| CN105930858A (en) * | 2016-04-06 | 2016-09-07 | 吴晓军 | Fast high-precision geometric template matching method enabling rotation and scaling functions |
| CN106485701A (en) * | 2016-09-26 | 2017-03-08 | 成都交大光芒科技股份有限公司 | Based on the whether anti-loaded detection method of the railway overhead contact system catenary seat of image |
| CN107671896A (en) * | 2017-05-19 | 2018-02-09 | 重庆誉鸣科技有限公司 | Fast vision localization method and system based on SCARA robots |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109101982B (en) | 2022-02-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111784685B (en) | Power transmission line defect image identification method based on cloud edge cooperative detection | |
| CN108596277B (en) | Vehicle identity recognition method and device and storage medium | |
| CN108898047B (en) | Pedestrian detection method and system based on blocking and shielding perception | |
| CN110738101A (en) | Behavior recognition method and device and computer readable storage medium | |
| CN109359666A (en) | A kind of model recognizing method and processing terminal based on multiple features fusion neural network | |
| CN106650615B (en) | A kind of image processing method and terminal | |
| CN105005565B (en) | Live soles spoor decorative pattern image search method | |
| CN108205676B (en) | The method and apparatus for extracting pictograph region | |
| CN109919002B (en) | Yellow stop line identification method and device, computer equipment and storage medium | |
| CN106373128B (en) | Method and system for accurately positioning lips | |
| CN112418360B (en) | Convolutional neural network training method, pedestrian attribute identification method and related equipment | |
| WO2019197021A1 (en) | Device and method for instance-level segmentation of an image | |
| CN110543838A (en) | Vehicle information detection method and device | |
| CN105426899A (en) | Vehicle identification method and device and client side | |
| CN108921162A (en) | Licence plate recognition method and Related product based on deep learning | |
| Duan et al. | Image classification of fashion-MNIST data set based on VGG network | |
| CN109389105A (en) | A kind of iris detection and viewpoint classification method based on multitask | |
| CN109977941A (en) | Licence plate recognition method and device | |
| CN115049954B (en) | Target identification method, device, electronic equipment and medium | |
| CN111444816A (en) | Multi-scale dense pedestrian detection method based on fast RCNN | |
| CN111382638B (en) | Image detection method, device, equipment and storage medium | |
| CN115272691A (en) | Training method, recognition method and equipment for steel bar binding state detection model | |
| CN113436162B (en) | Method and device for identifying weld defects on surface of hydraulic oil pipeline of underwater robot | |
| CN109101982A (en) | The recognition methods of target object and device | |
| CN113128308B (en) | Pedestrian detection method, device, equipment and medium in port scene |
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 |