CN110263736A - A kind of component identification method, apparatus, storage medium and system - Google Patents
A kind of component identification method, apparatus, storage medium and system Download PDFInfo
- Publication number
- CN110263736A CN110263736A CN201910553765.0A CN201910553765A CN110263736A CN 110263736 A CN110263736 A CN 110263736A CN 201910553765 A CN201910553765 A CN 201910553765A CN 110263736 A CN110263736 A CN 110263736A
- Authority
- CN
- China
- Prior art keywords
- component
- identified
- image
- images
- abnormal point
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/42—Document-oriented image-based pattern recognition based on the type of document
- G06V30/422—Technical drawings; Geographical maps
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
Embodiments herein provides a kind of component identification method, apparatus, storage medium and system, and method includes: to obtain the slice images comprising abnormal point mark and component to be identified；It is marked according to the abnormal point, eliminates the abnormal point in the slice images；Determine to characterize the component image to be identified of the component to be identified from the slice images for eliminating abnormal point；By the default component images match in the component image to be identified and component base, to identify the component to be identified.By determining the abnormal point in slice images and eliminating abnormal point, abnormal point can be avoided to treat the influence of identification means image as far as possible, component image to be identified is influenced so as to reduce by abnormal point and the profile to component image is led to problems such as to cause to block or connect, and the identification mistake further caused.Thus, it is possible to improve the accuracy rate to component identification in image.
Description
Technical field
This application involves field of image processings, such as are related to a kind of component identification method, apparatus, storage medium and system.
Background technique
In existing component identification method, when identifying to the component in image, the accuracy rate of identification is to be improved.
Summary of the invention
In view of this, the application's is designed to provide a kind of component identification for improving the recognition accuracy of component in image
Method, apparatus, storage medium and system.
To achieve the goals above, embodiments herein is accomplished in that
In a first aspect, embodiments herein provides a kind of component identification method, comprising: obtain comprising abnormal point mark and
The slice images of component to be identified；It is marked according to the abnormal point, eliminates the abnormal point in the slice images；It is abnormal from eliminating
It determines to characterize the component image to be identified of the component to be identified in the slice images of point；By the component image to be identified with
Default component images match in component base, to identify the component to be identified.
By determining the abnormal point in slice images and eliminating abnormal point, abnormal point can be avoided to treat knowledge as far as possible
The influence of other component image influences component image to be identified by abnormal point and leads to the profile to component image so as to reduce
The problems such as blocking or connecting is caused, and the identification mistake further caused.Thus, it is possible to improve to component identification in image
Accuracy rate.
With reference to first aspect, in the first possible implementation of the first aspect, described according to the abnormal point mark
Note, eliminates the abnormal point in the slice images, comprising: by clustering algorithm, determine from the slice images described different
The pixel value range of the pixel value range and background area often put；The pixel value range of the abnormal point is adjusted to the background
The pixel value range in region.
By using clustering algorithm, the presence for rapidly and accurately determining abnormal point can be marked according to abnormal point, and
Determine the pixel value range of abnormal point.Thus, it is possible to pass through the picture that the pixel value range of abnormal point is adjusted to background area
Element value range, to efficiently eliminate abnormal point.
With reference to first aspect, in the second possible implementation of the first aspect, described from point for eliminating abnormal point
Determine to characterize the component image to be identified of the component to be identified in picture, comprising: using Selective search (choosing
The search of selecting property) algorithm, extracted region is carried out to the slice images for eliminating abnormal point, to extract the component image to be identified.
Extracted region is carried out to the slice images after eliminating abnormal point by using Selective search algorithm, it can be with
Efficiently and accurately extract component image to be identified.
With reference to first aspect, in a third possible implementation of the first aspect, described by the component to be identified
Default component images match in image and component base, to identify the component to be identified, comprising: obtain the component to be identified
(image moment is existed by Hu (Visual pattern recognition by moment invariants) to the Hu square of image
It is proposed within 1962 that there is translation, rotation and scale invariability) feature vector；Calculate the component image to be identified with it is described
The distance between the Hu Character eigenvector of component image value is each preset in component base；Determine that wherein the smallest distance value is corresponding
Default component image is target member image, with the component to be identified according to the target member image recognition.
Feature by using Hu Character eigenvector as component image to be identified, and by calculating component image to be identified
Hu Character eigenvector and default component image the distance between Hu Character eigenvector value, with determine component image to be identified with
Default component diagram seems no for similar component.And Hu Character eigenvector has the characteristic of image rotation, scaling and feature invariant, it can
To identify the component image of all angles and/or size as precisely as possible, to guarantee the accuracy rate of component identification.
With reference to first aspect or first aspect the first into the third any possible implementation, in first aspect
The 4th kind of possible implementation in, it is described obtain comprising abnormal point mark and component to be identified slice images before,
The method also includes: obtain the images to be recognized comprising abnormal point mark and component mark；From the images to be recognized really
Make image corresponding with component mark, wherein it is the default component image that the component, which marks corresponding image,；Root
Fragment rule is determined according to the default component image, and fragment is carried out to the images to be recognized according to the fragment rule, with
Obtain the slice images.
By carrying out fragment to the images to be recognized comprising abnormal point mark and component mark, multiple fragment figures can be made
As being processed in parallel simultaneously, thus the operational efficiency of lifting member recognition methods on the whole.
The 4th kind of possible implementation with reference to first aspect, in the 5th kind of possible implementation of first aspect
In, it is described according to the default component image determine fragment rule, and according to the fragment rule to the images to be recognized into
Row fragment, comprising: determine the size of the default component image, the size of the default component image include horizontal size and
Vertical size；Fragment rule is determined according to the horizontal size and the vertical size；According to the fragment rule, sliding is utilized
Window algorithm carries out fragment to the images to be recognized, to obtain the slice images.
Fragment rule is determined by the horizontal size and vertical size according to component image to be identified, is done a sum orally in conjunction with sliding window
Method, the size of intersection when can determine the size and division of sliding window, it is possible thereby to avoid carrying out to image as far as possible
When fragment, the cutting of component image is being led to when identifying to component nothing due to component is imperfect in two panels slice images
The case where method identifies, therefore the accuracy of component identification can be further increased.
The third possible implementation with reference to first aspect, in the 6th kind of possible implementation of first aspect
In, after in determination, wherein the corresponding default component image of the smallest distance value is target member image, comprising: judgement is described most
Whether small distance value is more than preset threshold；If so, determining that the component to be identified is not that the target member image is corresponding
Component；If it is not, determining that the component to be identified is the corresponding component of the target member image.
By judging whether the lowest distance value exceeds threshold after the lowest distance value for determining Hu Character eigenvector
Value, so as to exclude component corresponding to component to be identified and non-default component diagram picture as far as possible, is able to ascend component identification
Accuracy rate.
With reference to first aspect, in a seventh possible implementation of the first aspect, in the slice images
After the completion of component identification, further includes: summarize to the component identification result of the slice images；Determine every component identification letter
Cease position of the corresponding component in the images to be recognized；According to position of the component in the images to be recognized, filter out
Duplicate component identification information.
By when determining multiple components, to belonging to same position same kind of and in images to be recognized
Component is excluded, and can include the component that is, in multiple slice images, thus by the component to avoid the case where identifying is repeated
Identification is repeatedly determined as multiple component situations, can also further increase the accuracy of component identification in this way.
Second aspect, embodiments herein provide a kind of component identification device, comprising: slice images obtain module, use
In slice images of the acquisition comprising abnormal point mark and component to be identified；Abnormal point cancellation module, for according to the abnormal point
Mark, eliminates the abnormal point in the slice images；Region extraction module, for being determined from the slice images for eliminating abnormal point
The component image to be identified of the component to be identified is characterized out；Component identification module, for will the component image to be identified and
Default component images match in component base, to identify the component to be identified.
The third aspect, embodiments herein provide a kind of storage medium, and the storage medium is stored with one or more
A program, one or more of programs can be executed by one or more processor, to realize such as first aspect or first
The first of aspect is into the 7th kind of possible implementation the step of any described component identification method.
Fourth aspect, embodiments herein provide a kind of electronic equipment, including memory and processor, the memory
For storing the information including program instruction, the processor is used to control the execution of program instruction, and described program instruction is located
The first of realization first aspect or first aspect any institute into the 7th kind of possible implementation when reason device is loaded and executed
The step of component identification method stated.
5th aspect, embodiments herein provide a kind of component identification system, comprising: processing unit and multiple calculating are single
Member includes that abnormal point marks the images to be recognized marked with component for obtaining；Fragment is carried out to the images to be recognized, to obtain
Slice images are taken, and the slice images are sent to multiple computing units；Each computing unit, for obtaining slice images；
It is marked according to the abnormal point, eliminates the abnormal point in the slice images；It is determined from the slice images for eliminating abnormal point
Characterize the component image to be identified of the component to be identified；And by the default structure in the component image to be identified and component base
Part images match to identify the component to be identified, and will identify that the identification information after component to be identified is sent to the processing
Unit；The processing unit is also used to integrate the identification information, to obtain structure corresponding with the images to be recognized
Part recognition result.
To enable the above objects, features, and advantages of the application to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows a kind of schematic diagram of component identification system provided by the embodiments of the present application.
Fig. 2 shows the structural schematic diagrams of a kind of electronic equipment provided by the embodiments of the present application.
Fig. 3 shows a kind of timing diagram of component identification method provided by the embodiments of the present application.
Fig. 4 shows a kind of images to be recognized provided by the embodiments of the present application.
Fig. 5 shows the schematic diagram of image slices provided by the embodiments of the present application.
Fig. 6 shows a kind of structural block diagram of component identification device provided by the embodiments of the present application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application is described.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing.Meanwhile the application's
In description, term " first ", " second " etc. are only used for distinguishing description, are not understood to indicate or imply relative importance.
Currently, when identifying to the component in image, the accuracy rate of identification is by various in existing component identification method
Factor influences, and generally, lacks the high component identification method of recognition accuracy.Based on this, applicant of the present invention provides one
Kind component identification method, to improve the accuracy rate of component identification.
Referring to Fig. 1, embodiments herein provides a kind of component identification system 10, including processing unit 11 and multiple meters
Calculate unit 12.Processing unit 11 is connect with multiple computing units 12, is worked in coordination to execute component provided by the embodiments of the present application
Recognition methods.
In the present embodiment, component identification can be used as multiple electronic equipments 20 that independently can be handled and be calculated
Processing unit 11 and multiple computing units 12 in system 10, to form component identification system 10；It is also possible to an electronics
Processing part and multiple calculating sections in equipment 20 is as the processing unit 11 and multiple calculating list in component identification system 10
Member 12, to form component identification system 10, herein not as restriction.
In the present embodiment, electronic equipment 20 can be terminal, such as smart phone, tablet computer, PC, individual
Digital assistants etc.；Electronic equipment 20 or server, such as network server, Cloud Server, server cluster, data clothes
Business device etc., is not construed as limiting herein.
Illustratively, electronic equipment 20 may include: the communication module 22 being connect by network with the external world, for executing journey
One or more processors 24, bus 23 and the various forms of memories 21 of sequence instruction, for example, disk, ROM (Read-Only
Memory, read-only memory) or RAM (Random Access Memory, random access memory), or any combination thereof.
Illustratively, program is stored in memory 21.Processor 24 can call from memory 21 and run these journeys
Sequence, so that component identification method can be executed by running program.Processor 24 by the execution to component identification method,
It may be implemented to carry out component identification to images to be recognized, to identify the component in images to be recognized.
Component identification system 10 can execute component identification method as shown in Figure 3 provided by the embodiments of the present application.
In the present embodiment, component identification method may include: step S11, step S12, step S13, step S14, step
Rapid S21, step S22, step S23 and step S24.Wherein, step S11, step S12, step S13 and step S14 can be by structures
Processing unit 11 in part identifying system 10 executes；And step S21, step S22, step S23 and step S24 can be single by calculating
Member 12 executes.Certainly, this is one way in which, in some other possible implementation, can not also distinguish processing
Unit and computing unit are executed each step of component identification method by component identification system 10, therefore, are not construed as herein pair
The restriction of the application.
In order to more clearly introduce the component identification method of the present embodiment, before processing unit 11 executes step S11, first
Images to be recognized is introduced.
Referring to Fig. 4, including component to be identified and abnormal point in images to be recognized, wherein to be identified in the present embodiment
The component image that component representation needs to identify according to the actual demand of user, abnormal point indicate include in images to be recognized one
It will affect the identified pixel of component, such as some textual annotations in image, chaotic lines, pixel, Yi Jitu
The textual annotation being superimposed as in component, wall body stripe etc..
In the present embodiment, user can be labeled images to be recognized, for example, some in mark images to be recognized
Component, alternatively, mark abnormal point etc..These mark can make component identification system 10 to the component in images to be recognized into
When row identification, more targetedly, so that recognition result more meets the demand of user.Illustratively, user chooses wait know
One of other image part is labeled, and using component identification method provided by the embodiments of the present application, identifies figure to be identified
Belong to of a sort component with the component of mark as in.It is of course also possible to use the component image marked in advance as component
Mark, such as the component image library of the external component image comprising mark, are treated according to the component image in component image library
Component in identification image is identified.And abnormal point mark can detect abnormal point by some algorithms or model,
To generate abnormal point mark, or according to the label of user, abnormal point mark is generated, certainly, abnormal point is also possible to preset
, therefore, it is not construed as the restriction to the application herein.With user images to be recognized will be labeled in the present embodiment
For situation, component identification method is introduced.
By taking Fig. 4 as an example, it is assumed that partial component and abnormal point in user annotation images to be recognized.So, component identification
System 10 can generate component mark corresponding with the operation of user and abnormal point mark based on the labeling operation of user.It needs
Illustrate, component mark and abnormal point mark herein are equivalent to one and component identification system 10 needed to execute corresponding actions
Label.
Processing unit 11 can execute step S11.
Step S11: the images to be recognized comprising abnormal point mark and component mark is obtained.
In the present embodiment, the available images to be recognized comprising abnormal point mark and component mark of processing unit 11.
After processing unit 11 obtains the images to be recognized comprising abnormal point mark and component mark, step can be executed
S12。
Step S12: image corresponding with component mark is determined from the images to be recognized, wherein the structure
It is the default component image that part, which marks corresponding image,.
In the present embodiment, processing unit 11 can be marked according to component, be labeled in images to be recognized to component corresponding
Region determine that corresponding component image, the image can indicate the image for the component that user is marked.Determine component
After marking corresponding image, processing unit 11 can as indicate mark component default component image and there are components
In library.
After determining default component image, component identification system 10 is (either processing unit 11, is also possible to calculate
Unit 12) component image can also be preset to this handle, obtain corresponding Hu Character eigenvector, and by the Hu moment characteristics to
Amount is associated with the default component image.
It should be noted that aforesaid operations are carried out to the component of each mark when the component of user annotation is multiple, and
It determines corresponding default component image and is stored in component base.
After determining default component image, processing unit can execute step S13.
Step S13: according to the default component image determine fragment rule, and according to the fragment rule to described wait know
Other image carries out fragment, to obtain the slice images.
In the present embodiment, when default component image is one, the available default component image of processing unit 11
Size, including horizontal size and vertical size.And when default component image is multiple, processing unit 11 can be determined often
The size of a default component image, and therefrom determine the maximum default component image of size in the horizontal direction, it obtains horizontal
Size, and, it therefrom determines the maximum default component image of size in the vertical direction, obtains vertical size.
After determining horizontal size and vertical size, it can determine according in horizontal size and vertical size for treating
Identify that image carries out the fragment rule of fragment.Illustratively, fragment rule can be with are as follows: determines level side when horizontal size is fragment
The size of the lap of upward two neighboring slice images determines when vertical size is fragment two neighboring point on vertical direction
The size of the lap of picture.Certainly, the horizontal size and vertical size of lap can also be more than or less than default
The corresponding horizontal size of component image and vertical size, herein not as restriction.
Adjacent two panels slice images overlapping portion when maximum horizontal size and vertical size are as fragment using in pre-set image
The size divided can make processing unit since the size of lap is to preset the full-size of component image in component base
11 when carrying out fragment to images to be recognized, and regardless of the mode of cutting, the size in images to be recognized is not more than default structure
The component to be identified of the corresponding horizontal size of part image and vertical size, it is at least complete in the presence of one in the slice images being syncopated as
Whole component image to be identified, thus can be effectively prevented from be split because of image after cause the unrecognized situation of component.
After determining fragment rule, computing unit 11 can be according to fragment rule, using sliding window algorithm to be identified
Image carries out fragment, to obtain slice images.
As shown in figure 5, illustrative, x-axis indicates that the horizontal direction of images to be recognized, y-axis indicate the perpendicular of images to be recognized
Histogram is to WsegIndicate the width of sliding window in the horizontal direction, HsegIndicate the height of sliding window in the vertical direction,
Wcomp_maxIt indicates to preset the full-size in component image horizontal direction, H in component basecomp_maxIt indicates to preset structure in component base
Full-size on part image vertical direction, certainly, Wseg> Wcomp_max, Hseg> Hcomp_max, the intersection point of x-axis and y-axis is set as
Origin (0,0).
In the present embodiment, the width W of sliding window in the horizontal directionsegHeight H in the vertical directionsegIt can be with
To preset size, such as WsegSize is 10000 pixel units, HsegSize is 8000 pixel units, and certainly, size can
According to the maximum width W for presetting component imagecomp_maxWith maximum height Hcomp_maxAnd it sets.Fragment rule can control cunning
Dynamic window from the origin (0,0) in the upper left corner of images to be recognized be starting point, in the horizontal direction from left to right and along the vertical direction from
Sequence under carries out cutting to images to be recognized, obtains slice images.And in the slice images obtained include the fragment figure
As the location information in images to be recognized, for example, the coordinate representation on each vertex of slice images can be used, it can also use and divide
The width and height of the coordinate combination slice images on the part vertex of picture indicate, herein not as restriction.
It should be noted that sliding window to images to be recognized carry out cutting when, sliding window cut out in water
The intersection size in the horizontal direction of adjacent two slice images can be with W square upwardscomp_maxIt is in the same size；And
On vertical direction the intersection of two adjacent slice images the size of vertical direction can be with Hcomp_maxIt is in the same size.?
When being sliced into the right end of images to be recognized, the horizontal size of images to be recognized remainder is less than sliding window in level side
To width when, can continue with sliding window carry out cutting, obtain images to be recognized remainder horizontal size point
Picture, can also be by being to carry out cutting referring to interim with the right end of images to be recognized, to obtain the consistent fragment figure of size
Picture, herein not as restriction.
Similar, similar measure can also be taken on the vertical direction of images to be recognized, every time in horizontal direction
After images to be recognized cutting, just sliding window is moved vertically downwards so that sliding window cutting in vertical direction
The height and H of intersection possessed by upper adjacent two panels slice imagescomp_maxUnanimously.It is being sliced into images to be recognized most
When bottom end, the vertical size of images to be recognized remainder is less than sliding window in the height of vertical direction, can continue benefit
Carry out cutting with sliding window, obtain the slice images of the vertical size of images to be recognized remainder, can also by with to
The bottom end for identifying image is to carry out cutting referring to interim, to obtain the consistent slice images of size, herein not as restriction.
Slice images are obtained in this way, the component to be identified that can not only be effectively prevented from images to be recognized
Due to being split cannot identified situation, guarantee identification accuracy rate；It also, can also include every fragment in slice images
Relative position of the image in images to be recognized, so that it is determined that going out the position of the component identified.
After determining slice images, the slice images of cutting can be allocated by processing unit 11, by slice images
It is sent to multiple computing units 12.And each computing unit 12 is due to only needing to handle small number of slice images, and it is multiple
Computing unit is handled simultaneously, can save many processing times, to greatly improve the efficiency of component identification method execution.
In the present embodiment, after slice images are sent to multiple computing units 12 by processing unit 11, each computing unit
12 can execute step S21.
Step S21: the slice images comprising abnormal point mark and component to be identified are obtained.
In the present embodiment, the available slice images of computing unit 12, and in slice images may include abnormal point mark
Note and component to be identified, it is also possible to all not include comprising one such or both.Both the present embodiment will be to including
Situation is illustrated, to introduce computing unit 12 all-sidely to the treatment process of slice images.Certainly, only include in slice images
When one of abnormal point mark and component to be identified, computing unit 12 can correspond to execution and abnormal point mark or structure to be identified
Part corresponding operation is not construed as the restriction to the application herein.After obtaining slice images, computing unit 12 can execute step
S22。
Step S22: it is marked according to the abnormal point, eliminates the abnormal point in the slice images.
In the present embodiment, there are when abnormal point in slice images, computing unit 12 can be marked according to abnormal point, be disappeared
Except the abnormal point in slice images.
Illustratively, computing unit 12 can be by clustering algorithm, according to the region where abnormal point mark, to abnormal point
It is detected.For example, the pixel value of each abnormal point in slice images, shape can be detected by using isolated forest algorithm
At the set comprising abnormal point Yu its pixel value, so that it is determined that the pixel value range of abnormal point out.Likewise, cluster can be passed through
Algorithm determines the pixel value range of background area.It should be noted that background area can be the part area comprising abnormal point
Domain, or the background area of entire slice images, herein not as restriction.It is the part comprising abnormal point in background area
When region, the pixel value range of background area can be determined, the abnormal point in background area is adjusted to the picture of background area
Element value range, to eliminate abnormal point.When in the region where the profile of component to be identified there are abnormal point, and component to be identified
Pixel value range and slice images entirety background area pixel value range it is inconsistent when, can in this way, with
Reach better abnormal point eradicating efficacy.
In addition, the elimination to the abnormal point in slice images, can be it is selective, for example, marking institute according to abnormal point
Region in abnormal point eliminated, and the abnormal point for not being labeled can not be eliminated, herein not as limit
It is fixed.
Abnormal point Processing for removing is carried out to slice images, abnormal pixel, chaotic can be reduced or eliminated as far as possible
The abnormal points such as lines, text are negatively affected caused by the component to be identified in slice images, to improve the standard of component identification
True property.
In the processing that slice images are carried out with abnormal point elimination, after obtaining the slice images for eliminating abnormal point, computing unit
Step S23 can be executed.
Step S23: it determines to characterize the component to be identified of the component to be identified from the slice images for eliminating abnormal point
Image.
In the present embodiment, slice images eliminate abnormal point after, computing unit can be extracted from slice images to
Region where identification means, to obtain component image to be identified.
Illustratively, the region of component to be identified is extracted, Selectivesearch algorithm can be used, to fragment figure
Extraction as carrying out component to be identified.Component image to be identified is extracted by using Selectivesearch algorithm, it can be efficient
Accurately extract the region of the component to be identified in slice images.
In the present embodiment, the component image to be identified determined, can also be comprising component image to be identified in fragment figure
Determine component to be identified to be identified so as to combine position of the slice images in images to be recognized in position as in
Position in image.
After determining component image to be identified, computing unit 12 can execute step S24.
Step S24: by the default component images match in the component image to be identified and component base, with identify it is described to
Identification means.
In the present embodiment, computing unit 12 can treat identification means image and be handled, and determine images to be recognized
Hu Character eigenvector.And after determining the Hu Character eigenvector of images to be recognized, it can each be preset with component base
The Hu Character eigenvector of component image carries out the calculating of Hu Character eigenvector distance value, calculates and each presets component image
The distance value of Hu Character eigenvector.The corresponding default component image of lowest distance value is determined as target member image, and wait know
Other component may be consistent with component represented by target member image.It should be noted that being extracted in slice images more
When a component image to be identified, each component image to be identified can determine that Hu Character eigenvector distance value is the smallest therewith
One default component image is as its corresponding target member image.
By using Hu Character eigenvector as the default component image measured in component image and component base to be identified
Mode with degree since Hu Character eigenvector is not influenced by variations such as the rotation of image, scalings, thus can be calculated steadily
The matching degree of component image to be identified and default component image, to guarantee the accuracy of component identification.
After determining the corresponding target member image of component image to be identified, computing unit 12 can be to processing unit 11
Identification information is sent, may include location information, to be identified component diagram of the slice images in images to be recognized in identification information
As the Hu moment characteristics of location information, target member image and target member image and component image to be identified in slice images
The information etc. of vector distance value.
And processing unit 11 can receive the identification information that each computing unit 12 is sent, to execute step S14.
Step S14: identification means are treated according to identification information and are screened.
In the present embodiment, processing unit 11 treats the screening of identification means progress, may include: to treat identification means
The screening of recognition result confirmation；And duplicate removal screening is carried out to the recognition result of confirmation.
Wherein, the screening for treating the recognition result confirmation of identification means, can be with are as follows:
Whether the Hu Character eigenvector distance value for judging component image to be identified and target member image is more than preset threshold.
Illustratively, preset threshold can be set to 1.5, certainly, only illustrates, should not be regarded as limiting, preset threshold can root herein
It is set according to actual needs.
Judge that Hu Character eigenvector distance value has been more than preset threshold, illustrates component image to be identified and target member figure
The similarity of picture has gap, not can determine that component to be identified is the corresponding component of target member image.And in Hu Character eigenvector
When distance value is no more than preset threshold, it can determine that component to be identified is the corresponding component of target member image.
Screening mode in this way can exclude to be not belonging to because of component to be identified any default in component base as far as possible
The case where component image corresponding component, to further increase the accuracy of component identification.
After the screening for treating identification means progress recognition result confirmation, processing unit 11 can be to the recognition result of confirmation
Carry out duplicate removal screening.
In the present embodiment, the component that component identification result can be determined as same class component by processing unit 11 sieves
Choosing.Illustratively, position of the corresponding component of component identification information in images to be recognized can be determined；And according to component to
It identifies the position in image, filters out duplicate component identification information, so that it may which the component being overlapped to position screens, with complete
At duplicate removal.
It, can be to avoid a component in images to be recognized because in multi-disc fragment by carrying out duplicate removal screening to recognition result
Chaotic problem is identified caused by after being identified simultaneously in image, so as to the accuracy of further lifting member identification.
Embodiments herein also provides a kind of storage medium, and storage medium is stored with one or more program, and one
Or multiple programs can be executed by one or more processor, to realize component identification method provided by the embodiments of the present application
Step.
Referring to Fig. 6, embodiments herein also provides a kind of component identification device 30, comprising: slice images obtain mould
Block 33, for obtaining the slice images comprising abnormal point mark and component to be identified；Abnormal point cancellation module 34, for according to institute
Abnormal point mark is stated, the abnormal point in the slice images is eliminated；Region extraction module 35, for the fragment from elimination abnormal point
Determine to characterize the component image to be identified of the component to be identified in image；Component identification module 36, for by described wait know
Default component images match in other component image and component base, to identify the component to be identified.
In the present embodiment, abnormal point cancellation module 34 is also used to through clustering algorithm, is determined from the slice images
The pixel value range of the pixel value range of the abnormal point and background area out；By the pixel value range of the abnormal point adjust to
The pixel value range of the background area.
In the present embodiment, region extraction module 35 are also used to using Selectivesearch algorithm, abnormal to eliminating
The slice images of point carry out extracted region, to extract the component image to be identified.
In the present embodiment, component identification module 36, be also used to obtain the Hu moment characteristics of the component image to be identified to
Amount；Calculate the Hu square spy that component image is each preset in the Hu Character eigenvector and the component base of the component image to be identified
Levy the distance between vector value；Determine that the corresponding default component image of wherein the smallest distance value is target member image, with root
According to component to be identified described in the target member image recognition.
In the present embodiment, component identification device 30 further include: images to be recognized obtains module 31, for obtaining comprising different
The often images to be recognized of point mark and component mark；Figure corresponding with component mark is determined from the images to be recognized
Picture, wherein it is the default component image that the component, which marks corresponding image,；Fragment is determined according to the default component image
Rule, and fragment is carried out to the images to be recognized according to the fragment rule, to obtain the slice images.
In the present embodiment, images to be recognized fragment module 32 is also used to determine the size of the default component image,
The size of the default component image includes horizontal size and vertical size；It is true according to the horizontal size and the vertical size
Determine fragment rule；According to the fragment rule, fragment is carried out to the images to be recognized using sliding window algorithm, to obtain
State slice images.
In the present embodiment, component identification device 30 further include: component screening module 37, for judge it is described it is the smallest away from
It whether is more than preset threshold from value；If so, determining that the component to be identified is not the corresponding component of the target member image；If
It is no, determine that the component to be identified is the corresponding similar component of the target member image.
In the present embodiment, component screening module 37 is also used to converge the component identification result of the slice images
Always；Determine position of the corresponding component of every component identification information in the images to be recognized；According to component described wait know
Position in other image filters out duplicate component identification information.
In conclusion embodiments herein provides a kind of component identification method, apparatus, storage medium and system, pass through
It determines the abnormal point in slice images and eliminates abnormal point, abnormal point can be avoided as far as possible to treat identification means image
It influences, component image to be identified is influenced so as to reduce by abnormal point and the profile to component image is caused to cause to block or even
The problems such as connecing, and the identification mistake further caused.Thus, it is possible to improve the accuracy rate to component identification in image.
More than, the only specific embodiment of the application, but the protection scope of the application is not limited thereto, and it is any to be familiar with
Those skilled in the art within the technical scope of the present application, can easily think of the change or the replacement, and should all cover
Within the protection scope of the application.Therefore, the protection scope of the application should be subject to the protection scope in claims.
Claims (11)
1. a kind of component identification method characterized by comprising
Obtain the slice images comprising abnormal point mark and component to be identified；
It is marked according to the abnormal point, eliminates the abnormal point in the slice images；
Determine to characterize the component image to be identified of the component to be identified from the slice images for eliminating abnormal point；
By the default component images match in the component image to be identified and component base, to identify the component to be identified.
2. component identification method according to claim 1, which is characterized in that it is described to be marked according to the abnormal point, it eliminates
Abnormal point in the slice images, comprising:
By clustering algorithm, the pixel value range of the abnormal point and the pixel of background area are determined from the slice images
It is worth range；
The pixel value range of the abnormal point is adjusted to the pixel value range of the background area.
3. component identification method according to claim 1, which is characterized in that described from the slice images for eliminating abnormal point
Determine to characterize the component image to be identified of the component to be identified, comprising:
Using Selectivesearch algorithm, extracted region is carried out to the slice images for eliminating abnormal point, with extract it is described to
Identification means image.
4. component identification method according to claim 1, which is characterized in that described by the component image to be identified and structure
Default component images match in part library, to identify the component to be identified, comprising:
Obtain the Hu Character eigenvector of the component image to be identified；
Calculate the Hu square that component image is each preset in the Hu Character eigenvector and the component base of the component image to be identified
The distance between feature vector value；
Determine that the corresponding default component image of wherein the smallest distance value is target member image, according to the target member figure
As identifying the component to be identified.
5. component identification method described in any one of -4 according to claim 1, which is characterized in that obtain described comprising different
Often before the slice images of point mark and component to be identified, the method also includes:
Obtain the images to be recognized comprising abnormal point mark and component mark；
Image corresponding with component mark is determined from the images to be recognized, wherein the component mark is corresponding
Image is the default component image；
Fragment rule is determined according to the default component image, and the images to be recognized is divided according to the fragment rule
Piece, to obtain the slice images.
6. component identification method according to claim 5, which is characterized in that described to be determined according to the default component image
Fragment rule, and fragment is carried out to the images to be recognized according to the fragment rule, comprising:
Determine the size of the default component image, the size of the default component image includes horizontal size and vertical ruler
It is very little；
Fragment rule is determined according to the horizontal size and the vertical size；
According to the fragment rule, fragment is carried out to the images to be recognized using sliding window algorithm, to obtain the fragment
Image.
7. component identification method according to claim 4, which is characterized in that in determination, wherein the smallest distance value is corresponding
Default component image is after target member image, comprising:
Judge whether the smallest distance value is more than preset threshold；
If so, determining that the component to be identified is not the corresponding component of the target member image；
If it is not, determining that the component to be identified is the corresponding component of the target member image.
8. component identification method according to claim 1, which is characterized in that the component identification in the slice images
After the completion, further includes:
The component identification result of the slice images is summarized；
Determine position of the corresponding component of every component identification information in the images to be recognized；
According to position of the component in the images to be recognized, duplicate component identification information is filtered out.
9. a kind of component identification device characterized by comprising
Slice images obtain module, for obtaining the slice images comprising abnormal point mark and component to be identified；
Abnormal point cancellation module eliminates the abnormal point in the slice images for marking according to the abnormal point；
Region extraction module, for determining to characterize the to be identified of the component to be identified from the slice images for eliminating abnormal point
Component image；
Component identification module, for by the default component images match in the component image to be identified and component base, with identification
The component to be identified.
10. a kind of storage medium, which is characterized in that the storage medium is stored with one or more program, it is one or
The multiple programs of person can be executed by one or more processor, to realize the component identification as described in any one of claim 1-8
The step of method.
11. a kind of component identification system characterized by comprising processing unit and multiple computing units,
The processing unit is used to obtain the images to be recognized comprising abnormal point mark and component mark；To the images to be recognized
Fragment is carried out, to obtain slice images, and the slice images are sent to multiple computing units；
Each computing unit, for obtaining slice images；It is marked according to the abnormal point, eliminates the exception in the slice images
Point；Determine to characterize the component image to be identified of the component to be identified from the slice images for eliminating abnormal point；And by institute
State the default component images match in component image to be identified and component base, to identify the component to be identified, and will identification to
Identification information after identification means is sent to the processing unit；
The processing unit is also used to integrate the identification information, to obtain structure corresponding with the images to be recognized
Part recognition result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910553765.0A CN110263736A (en) | 2019-06-24 | 2019-06-24 | A kind of component identification method, apparatus, storage medium and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910553765.0A CN110263736A (en) | 2019-06-24 | 2019-06-24 | A kind of component identification method, apparatus, storage medium and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110263736A true CN110263736A (en) | 2019-09-20 |
Family
ID=67921237
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910553765.0A Pending CN110263736A (en) | 2019-06-24 | 2019-06-24 | A kind of component identification method, apparatus, storage medium and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110263736A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651378A (en) * | 2021-01-08 | 2021-04-13 | 刘连英 | Method, device and medium for identifying marking information of fastener two-dimensional drawing |
| CN115292540A (en) * | 2022-07-22 | 2022-11-04 | 杭州易有料科技有限公司 | Multi-mode part information identification method, device, equipment and computer readable medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103208123A (en) * | 2013-04-19 | 2013-07-17 | 广东图图搜网络科技有限公司 | Image segmentation method and system |
| CN103582310A (en) * | 2013-09-18 | 2014-02-12 | 深圳市腾世机电有限公司 | Surface-mounting control method and system for surface-mounted element of surface-mounting machine |
| CN105956632A (en) * | 2016-05-20 | 2016-09-21 | 浙江宇视科技有限公司 | Target detection method and device |
| CN106780486A (en) * | 2017-01-16 | 2017-05-31 | 大连海事大学 | A kind of Surface Defects in Steel Plate image extraction method |
| CN107358716A (en) * | 2017-06-20 | 2017-11-17 | 深圳怡化电脑股份有限公司 | Heisui River print recognition methods, device and terminal device |
| CN109325917A (en) * | 2017-08-01 | 2019-02-12 | 深圳市中兴微电子技术有限公司 | The method and device of noise, computer readable storage medium in a kind of detection image |
| CN109325401A (en) * | 2018-08-03 | 2019-02-12 | 成都准星云学科技有限公司 | The method and system for being labeled, identifying to title field are positioned based on edge |
-
2019
- 2019-06-24 CN CN201910553765.0A patent/CN110263736A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103208123A (en) * | 2013-04-19 | 2013-07-17 | 广东图图搜网络科技有限公司 | Image segmentation method and system |
| CN103582310A (en) * | 2013-09-18 | 2014-02-12 | 深圳市腾世机电有限公司 | Surface-mounting control method and system for surface-mounted element of surface-mounting machine |
| CN105956632A (en) * | 2016-05-20 | 2016-09-21 | 浙江宇视科技有限公司 | Target detection method and device |
| CN106780486A (en) * | 2017-01-16 | 2017-05-31 | 大连海事大学 | A kind of Surface Defects in Steel Plate image extraction method |
| CN107358716A (en) * | 2017-06-20 | 2017-11-17 | 深圳怡化电脑股份有限公司 | Heisui River print recognition methods, device and terminal device |
| CN109325917A (en) * | 2017-08-01 | 2019-02-12 | 深圳市中兴微电子技术有限公司 | The method and device of noise, computer readable storage medium in a kind of detection image |
| CN109325401A (en) * | 2018-08-03 | 2019-02-12 | 成都准星云学科技有限公司 | The method and system for being labeled, identifying to title field are positioned based on edge |
Non-Patent Citations (2)
| Title |
|---|
| 孔超: "基于兴趣区域的图像检索方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 赵小川: "《MATLAB图像处理 能力提高与应用案例》", 31 January 2014 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651378A (en) * | 2021-01-08 | 2021-04-13 | 刘连英 | Method, device and medium for identifying marking information of fastener two-dimensional drawing |
| CN112651378B (en) * | 2021-01-08 | 2021-10-15 | 唐旸 | Method, device and medium for identifying marking information of fastener two-dimensional drawing |
| CN115292540A (en) * | 2022-07-22 | 2022-11-04 | 杭州易有料科技有限公司 | Multi-mode part information identification method, device, equipment and computer readable medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109816644B (en) | Bearing defect automatic detection system based on multi-angle light source image | |
| CN111428723B (en) | Character recognition method and device, electronic equipment and storage medium | |
| CN107067006B (en) | Verification code identification method and system serving for data acquisition | |
| EP3176751B1 (en) | Information processing device, information processing method, computer-readable recording medium, and inspection system | |
| CN110135514B (en) | Workpiece classification method, device, equipment and medium | |
| CN110991448A (en) | Text detection method and device for nameplate image of power equipment | |
| WO2014018482A2 (en) | Form recognition method and device | |
| US8090151B2 (en) | Face feature point detection apparatus and method of the same | |
| CN110648322A (en) | Method and system for detecting abnormal cervical cells | |
| CN108154132A (en) | Method, system and equipment for extracting characters of identity card and storage medium | |
| KR20210020065A (en) | Systems and methods for finding and classifying patterns in images with vision systems | |
| CN116168351B (en) | Inspection method and device for power equipment | |
| CN110263736A (en) | A kind of component identification method, apparatus, storage medium and system | |
| CN111639653A (en) | False detection image determining method, device, equipment and medium | |
| CN110796016A (en) | Engineering drawing identification method, electronic equipment and related product | |
| US20210374455A1 (en) | Utilizing machine learning and image filtering techniques to detect and analyze handwritten text | |
| CN111461260A (en) | Target detection method, device and equipment based on feature fusion and storage medium | |
| CN113947754A (en) | Vision-based ship machinery running state monitoring method and system and storage medium | |
| CN104361339B (en) | Slap shape Graph Extraction and recognition methods | |
| CN116524205A (en) | Sewage aeration automatic detection and identification method | |
| CN110599487A (en) | Article detection method, apparatus and storage medium | |
| CN116434071B (en) | Determination method, determination device, equipment and medium for normalized building mask | |
| CN110598575B (en) | Form layout analysis and extraction method and related device | |
| CN111985506A (en) | Chart information extraction method and device and storage medium | |
| CN111898408A (en) | Rapid face recognition method and device |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190920 |
|
| RJ01 | Rejection of invention patent application after publication |