CN105894015A - Gate state analyzing method and system - Google Patents
Gate state analyzing method and system Download PDFInfo
- Publication number
- CN105894015A CN105894015A CN201610187111.7A CN201610187111A CN105894015A CN 105894015 A CN105894015 A CN 105894015A CN 201610187111 A CN201610187111 A CN 201610187111A CN 105894015 A CN105894015 A CN 105894015A
- Authority
- CN
- China
- Prior art keywords
- state
- banister
- difference
- window
- target
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a gate state analyzing method and system. The method includes the steps of automatically obtaining input images, of a monitored gate, acquired by external equipment, determining a target window where the target gate is located from the input images, determining the character area where switch state indicating characters are located from the target window, differentiating the character area, and outputting the switch state indicating signals matching with the differentiating result. When the switch state of a gate is monitored by the technical scheme disclosed in the embodiments of the invention, system automatic monitoring is realized, the monitoring process is safe and reliable, and no on-site checking by an operation work is required. Labor can be reduced.
Description
Technical field
The present invention relates to monitoring device technical field, is specifically related to a kind of banister state analysis method and system.
Background technology
In transformer station, banister is transformer substation switch instrument, and it has two states, is respectively a point shape
State and the state of conjunction, the Daily Round Check of banister state is one of very important project in the artificial O&M of transformer station,
Monitoring and the early warning in time of banister state have considerable meaning in transformer station's O&M.And in reality
O&M during, generally use the mode of manual inspection that the state of banister is patrolled and examined, manual inspection
Mode be usually present not in time, uneasy congruence hidden danger and defect.
Summary of the invention
In view of this, the embodiment of the present invention provides a kind of banister state analysis method and system, to realize solving
Certainly the mode of manual inspection exist patrol and examine not in time, uneasy congruence hidden danger and defect.
For achieving the above object, the following technical scheme of embodiment of the present invention offer:
A kind of banister state analysis method, including:
Obtain input picture；
By the target window determining target banister place in described input picture；
Character zone by the on off state pointing character place determining target banister in described target window；
Described character zone is carried out difference, output and difference result match for characterizing described banister
The on off state indication signal of on off state.
Preferably, in above-mentioned banister state analysis method, described determined target track by described input picture
The target window at lock place, including:
Described input picture is scanned by the window using pre-set dimension, the son that will be got by scanning
Input picture region is as candidate window；
By described candidate window is carried out color histogram contrast with default template image one by one, it is judged that right
Than result whether more than threshold value, if it does not, extract the gal cypress characteristics of image of described candidate window and carry out propping up
Hold the judgement of vector machine model, will determine that result is that positive window is as candidate target window, described support
Vector machine model be by extract the gal cypress characteristics of image of each training sample be supported vector machine model instruct
The model for carrying out target window differentiation got；
Select supporting vector machine model judgement time highest scoring candidate target window as target window.
Preferably, in above-mentioned banister state analysis method, described determined target track by described target window
The character zone at the on off state pointing character place of lock, including:
The bright dark feature utilizing target window builds matrix model；
Integrogram is used to calculate the average of each element blocks in matrix model；
Determine each element blocks in described matrix model and the maximum of the difference of other element blocks being adjacent
Value, forms maximum set；
By element blocks corresponding for one or more maximums of the maximum in described maximum set at target window
In Kou, area defined is as character zone.
Preferably, in above-mentioned banister state analysis method, described described character zone is carried out difference, defeated
Go out the condition indicative signal of the on off state for characterizing described banister matched with difference result, including:
Described character zone is carried out the difference of the first pre-set color passage, it is thus achieved that the first difference result；
Described character zone is carried out the difference of the second pre-set color passage, it is thus achieved that the second difference result；
Described first difference result and the second difference result are sued for peace；
The on off state exporting the on off state for characterizing banister matched with summed result judges letter
Number.
Preferably, in above-mentioned banister state analysis method, described poor to described first difference result and second
Divide result summation；Export the on off state of the on off state for characterizing banister matched with summed result
Judge signal, including:
Judge whether described first difference result and the second difference result sum are positioned at preset range, if
It is to export and compare described first difference result with described and switch shape that the second difference result sum matches
State judges signal, if it does not, use preset characters model to judge described character zone, obtains also
Output is for characterizing the on off state indication signal of the on off state of banister, and described preset characters model is logical
Cross the model for judging the judgement of target window on off state according to character zone that training obtains.
A kind of banister state analysis system, including:
Image acquisition units, is used for obtaining input picture；
Target location determines unit, for by the target window determining target banister place in described input picture
Mouthful；
Character zone determines unit, for by determining in described target window that the on off state of target banister refers to
Show the character zone at character place；
State analysis unit, for described character zone carries out difference, output matches with difference result
The on off state indication signal of on off state for characterizing described banister.
Preferably, in above-mentioned banister state analysis system, described target location determines unit, including:
Scanning element, for using the window of pre-set dimension to be scanned described input picture, will pass through
The sub-input picture region that scanning gets is as candidate window；
Unit chosen by candidate target window, for by by described candidate window one by one with default template image
Carry out color histogram contrast, it is judged that whether comparing result is more than threshold value, if it does not, extract described candidate
The gal cypress characteristics of image of window also carries out the judgement of supporting vector machine model, will determine that result is positive window
As candidate target window, described supporting vector machine model is the Jia Baitu by extracting each training sample
The model for carrying out target window differentiation obtained is trained as feature is supported vector machine model；
Target window chooses unit, the candidate of highest scoring during for selecting the judgement of supporting vector machine model
Target window is as target window.
Preferably, in above-mentioned banister state analysis system, described character zone determines unit, including:
Matrix model construction unit, for utilizing the bright dark feature of target window to build matrix model；
Integrogram computing unit, for using integrogram to calculate the average of each element blocks in matrix model；
Difference identifying unit, for determine each element blocks in described matrix model and be adjacent other
The maximum of the difference of element blocks, forms maximum set；
Character zone positioning unit, for determining maximum 1 in described maximum set in target window
Individual or the region at multiple element blocks place, using this region as character zone.
Preferably, in above-mentioned banister state analysis system, described state analysis unit, including:
First difference unit, for described character zone carries out the difference of the first pre-set color passage, obtains
Obtain the first difference result；
Second difference unit, for described character zone carries out the difference of the second pre-set color passage, obtains
Obtain the second difference result；
On off state identifying unit, for described first difference result and the second difference result are sued for peace, and
The on off state exporting the on off state for characterizing banister matched with summed result judges signal.
Preferably, in above-mentioned banister state analysis system, described on off state identifying unit, including:
Dif ference judgment unit, is used for judging described first difference result and the second difference result sum whether position
In preset range, if it is, output the first triggering signal, otherwise output the second triggering signal；
First identifying unit, for after getting described first triggering signal, it is judged that described first difference
Result and the size of the second difference result, output compares described first difference result and the second difference with described
The on off state that result sum matches judges signal；
Second identifying unit, for, after getting described second triggering signal, using preset characters model
Described character zone is judged, obtains and export the on off state of on off state for characterizing banister
Indication signal, described preset characters model is to judge target by what training obtained for foundation character zone
The model that window on off state judges.
Based on technique scheme, the banister state analysis method of embodiment of the present invention offer and system, lead to
Cross the input picture automatically obtaining the monitored banister that external equipment collects, by described input picture
Determine the target window at target banister place, then by the switch determining target banister in described target window
The character zone at case pointer symbol place, finally carries out difference, output and difference to described character zone
The on off state indication signal that result matches.As can be seen here, technology disclosed in the above embodiments of the present application
When the on off state of banister is monitored by scheme, feasible system monitors automatically, and monitors process peace
Complete reliable, it is not necessary to operation maintenance personnel scene is checked, has liberated labour force.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to reality
Execute the required accompanying drawing used in example or description of the prior art to be briefly described, it should be apparent that below,
Accompanying drawing in description is only embodiments of the invention, for those of ordinary skill in the art, not
On the premise of paying creative work, it is also possible to obtain other accompanying drawing according to the accompanying drawing provided.
Fig. 1 is a kind of banister state analysis method flow chart disclosed in the embodiment of the present application；
Fig. 2 is a kind of banister state analysis method flow chart disclosed in another embodiment of the application；
Fig. 3 is a kind of banister state analysis method flow chart disclosed in another embodiment of the application；
Fig. 4 is a kind of banister state analysis method flow chart disclosed in another embodiment of the application；
Fig. 5 is a kind of banister state analysis method flow chart disclosed in another embodiment of the application；
Fig. 6 is the structural representation of a kind of banister state analysis system disclosed in the embodiment of the present application；
Fig. 7 is the structural representation of a kind of banister state analysis system disclosed in another embodiment of the application；
Fig. 8 is the structural representation of a kind of on off state identifying unit disclosed in the embodiment of the present application.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out
Clearly and completely describe, it is clear that described embodiment is only a part of embodiment of the present invention, and
It is not all, of embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art are not doing
Go out the every other embodiment obtained under creative work premise, broadly fall into the scope of protection of the invention.
Applicant is found by research, demand based on banister state-detection, needs first to position plug-in strip target,
After plug-in strip target determines, in plug-in strip target zone, find the character at corresponding case pointer symbol place
Region, then, then by judge that described character zone judges that the character in described character zone characterizes time
" dividing " state of banister still " closes " state.
Therefore, the whole technical scheme of the present invention is broadly divided into three phases: one, determine that banister is defeated
Enter the position on image；Two, in the region of banister, status word is found " to divide " " conjunction " region；Three,
Character in the character zone determining second stage is identified.Concrete, banister disclosed in the present application
State analysis method, sees Fig. 1, may include that
Step S101: obtain the input picture that external equipment collects；
Wherein, described input picture can be to be entered monitoring region by the image capture device of predetermined position
The image that row image acquisition obtains；
Step S102: by the target window determining target banister place in described input picture；
In this step, the input picture collected due to described step S101 can not only exist required
, in addition to the target of required monitoring, also can there are other inactive area in the target (banister) of monitoring,
Accordingly, it would be desirable to by the target window searching, determining banister place in described input picture；
Step S103: by the on off state pointing character place determining target banister in described target window
Character zone；
In this step, the image after described target window determines, included in described target window
Information, in addition to the character information for the current state characterizing banister, also includes the surface texture of banister
Etc. information, therefore, if to determine the current state of described banister, in this step, in addition it is also necessary to described
The region at the on off state pointing character place of banister positions；
Step S104: described character zone is carried out difference, output and difference result match for table
Levy the on off state indication signal of the on off state of described banister.
After the character zone residing for the on off state pointing character determining banister, need described character
The implication represented by pointing character in region resolves, owing to the on off state of usual described banister refers to
Showing that color and the symbol of character are distinct, such as, red " conjunction " represents the current state of described banister
For closing state, green " dividing " represents that current banister state is point state, refers to different on off states
After showing that the character zone residing for symbol carries out difference, the difference result that obtains also differs, and therefore, can pass through
This difference result judges the on off state of described banister.
In sum, when using the on off state of method detection banister disclosed in the above embodiments of the present application,
Automatically the input picture of the monitored banister that external equipment collects is obtained, by true in described input picture
Set the goal the target window at banister place, then by the switch shape determining target banister in described target window
The character zone at state pointing character place, finally carries out difference to described character zone, and output is tied with difference
The on off state indication signal that fruit matches.As can be seen here, technical side disclosed in the above embodiments of the present application
When the on off state of banister is monitored by case, feasible system monitors automatically, and monitors process safety
Reliably, it is not necessary to operation maintenance personnel scene is checked, has liberated labour force.
From the overall process of such scheme, during above-mentioned detection, it is mainly concerned with the detection word of object
Symbol detection, the common method for object detection mainly has two classes, a class to be by substantial amounts of sample training
Generating model, the model utilizing training to obtain carries out the differentiation of object；Another kind of is to use image template
The method joined, the method be by using a target image as template, in image to be detected progressively
Join find in image to be detected defeated as candidate target closest to most like block with target image
Go out.
When determining target window, the mode that sliding window can be used to carry out traveling through carries out object detection, makes
Remove the regional area obtaining in image successively with the window of a slidably fixed size, use image
Feature characterizes this regional area, and goes to discriminate whether the target for be looked for grader.The cunning of traversal
Dynamic window can be very time-consuming, because a secondary image to be detected may produce thousands of window areas,
Based on this, the application can carry out pretreatment to image, and first comparison window image carries out figure with target image
As the comparison of distribution of color, the subregion that those are the biggest with target difference is filtered out in advance, without
Use grader to go to distinguish, so can filter out the backdrop window of 80% in advance, be greatly promoted speed.
When determining target window, template matching technique is used to realize target detection, but template matching mode
Needing reference map as template, and the picture obtained at present is because of the change such as ambient lighting inevitably and base
Quasi-figure creates deviation, and when this deviation is bigger, target will be incapable of recognizing that.Therefore, template
The change impacts such as formula formula and ambient lighting are relatively big, cause this kind of mode unreliable.
Being directed to the problems referred to above, the application above-mentioned steps S102 can first use sliding window to obtain region, profit
Carry out coarse sizing with color histogram information, extract region gal cypress characteristics of image the most again, use support to
Amount machine grader accurately differentiates, concrete, and described step S102 may include that
Step S1021: use the window of pre-set dimension to be scanned described input picture, will be by scanning
The sub-input picture region got is as candidate window；
In this step, owing to the position of banister is relatively fixed, and banister corresponding to each preset point
Size is relatively fixed, and method disclosed in the above embodiments of the present application can obtain by the way of man-machine interaction
The approximate dimensions of target banister, thus, determines that the problem of target window just transfers to and finds in the input image
The problem of the target banister object of known dimensions；
For input picture, use the dimension information of the target banister got by above-mentioned man-machine interaction mode,
Determine sliding window size, use the sliding window of fixed dimension that described input picture is scanned, obtain
Take each image region of obtaining of scanning to be as candidate window, the size of the most described candidate window
By the size of the target banister that man-machine interactively mode gets；
Step S1022: by described candidate window is carried out color histogram pair with default template image one by one
Ratio, it is judged that whether comparing result is more than threshold value, if it does not, perform step S1023；
In this step, the most described candidate window image first can be with a template image (mesh in Sample Storehouse
Mark banister image is as template) carry out the comparison of color histogram, judge described candidate's window according to comparing result
Whether mouth is probably target window, particularly as follows: obtain rectangular histogram comparing result, by comparing result more than threshold
The candidate window of value is considered as non-targeted window, this window rejected in advance is not entered next stage more accurate
Judgement, comparing result is considered as doubtful non-targeted window less than the candidate window of threshold value, enters next step
Judge more accurately；Owing in described input picture, target banister compares other area difference in input picture
Different more obvious, most uncorrelated region generally can be weeded out by the process of coarse sizing；
Step S1023: extract the gal cypress characteristics of image of described candidate window and carry out supporting vector machine model
Judge, it is judged that whether result is just, if it is, perform step S1024；
This step belongs to the fine judgement to target window, is mainly used in by getting in step S1022
Determining real target window in suspected target window, detailed process is by extracting suspected target window
The gal cypress characteristics of image of candidate window, is supported according to the gal cypress characteristics of image of each candidate window respectively
The judgement of vector machine model, will determine that candidate window that result is positive suspected target window is as candidate's mesh
Mark window；
The algorithm that the present invention uses need one can determine whether whether candidate window is the model of target window, should
Model is to be obtained by the training of sample.Wherein, the process of training can be that off-line is carried out, and passes through
Learn the object features of a large amount of training sample and realize.The substantial amounts of samples pictures of training need of model, sample
Originally can divide positive sample and negative sample, positive sample is all angles, the picture of the target banister of each illumination,
Negative sample is other regional location and the picture of other various non-targeted banisters in original image, by all of
Positive and negative samples normalization is to the size preset, and the gal cypress characteristics of image extracting each sample is supported vector
Machine model training, can generate for judging the model that banister differentiates；This model carries out sentencing of target window
Not；
In technical scheme disclosed in the present application, can be by using multiple data mining algorithm to training sample
Model is trained by this, obtains the model that the application is finally required, wherein, the data of required employing
Mining algorithm can be support vector machine (Support Vector Machine, SVM), decision tree, nerve
Network algorithm etc..
Step S1024: will determine that result is that positive window is as candidate target window；
In this step, the gal cypress characteristics of image of each candidate window is supported sentencing of vector machine model
Have no progeny, obtain judged result, will be deemed as positive window as candidate target window；
Step S1025: select the candidate target window conduct of highest scoring during the judgement of supporting vector machine model
Target window；
It is to choose mesh in candidate target window that this step is used for by the judged result of multiple supporting vector machine models
Mark window, the candidate target window that the judged result score of supporting vector machine model is the highest, it is the most likely
Being target window, described candidate target window becomes big with score and constantly updates, and chooses the time of highest scoring
Select target window as target window, certainly when using said method to judge, need to obtain all of time
Select the judged result of target window, the candidate target window corresponding by choosing maximum in all of judged result
Mouth is as target window, it is seen then that the method to be supported vector machine model to all of candidate target
Judging, operand is big, and it is possible that multiple invalid judgement, such as, if to first candidate
Target window is supported result that vector machine judges when being maximum, other candidate target window follow-up
The judgement of mouth is invalid judgement (on choosing result without impact), therefore can reduce the judgement of whole program
Efficiency, is directed to this, a predeterminable judge index, when certain candidate target window is supported vector
When the judged result that machine model judges is more than this judge index, the most automatically assert sentencing of this candidate target window
Disconnected result is the maximum in the judged result of all candidate window, does not carry out other candidate target windows follow-up
The judgement of the supporting vector machine model of mouth, thus reduce system operations amount, improve operation efficiency.
Step S103 is mainly used in being determined character zone by described target window, in the target residing for banister
In window, status word is found " to divide " " conjunction " region.By the analysis of banister can be found, although become
In power station, banister size is the most different, and case pointer symbol distribution is also not quite similar, but have one identical
Feature, character is in dial plate white portion (light tone district), and the surrounding of dial plate is black (dark-coloured district),
Therefore, it can utilize this feature to build model (character zone gray value deducts field gray value for maximum),
In conjunction with the use of image integration figure, distribution of color in dial plate region can be quickly found out and meet the district of banister feature
Territory is as character zone.Concrete, see Fig. 3, in said method disclosed in the embodiment of the present application, described
Step S103, specifically may include that
Step S1031: utilizing the bright dark feature of target window to build matrix model, this matrix model is similar
The matrix model of endian format, its multiple element blocks being distributed by ranks form；
Step S1032: use integrogram to calculate the average of each element blocks in matrix model；
Step S1033: determine each element blocks in described matrix model and other element blocks of being adjacent it
The maximum of difference, forms maximum set；
Step S1034: by element corresponding for one or more maximums of the maximum in described maximum set
Block in target window area defined as character zone；
Wherein, during one or more element blocks maximum in choosing described maximum set, predeterminable
One maximum discrimination standard, just available more than the maximum of this discrimination standard in the most described maximum set
In determining character zone, in maximum set, the element blocks corresponding more than the maximum of described discrimination standard
Area defined is as character zone.
Described step S104, for judging banister according to the on off state designator in described character zone
Current state, utilizes the color characteristic of character zone breaker in middle status indicator to sentence the instruction of described on off state
Symbol is " conjunction " or " dividing ", so that it is determined that banister state.
Owing on banister, the color of two on off state designators of different conditions is different, therefore can be by carrying
Take the colouring information of character zone image, and judge in current character region through certain calculating process
Character time " dividing " or " conjunction ".Generally, described on off state designator generally uses red mark
" close " character, green mark " divide " character, therefore can by character zone is carried out red channel with
The difference of green channel, if difference result is green, green channel classification red channel to be significantly greater than divides
Amount, the on off state of the most described banister is for dividing；If red, red channel component is the greenest
Chrominance channel component, the on off state of the most described banister is for closing.It is directed to this, sees Fig. 4, described step
S104 specifically may include that
Step S1041: described character zone is carried out the difference of the first pre-set color passage, it is thus achieved that first is poor
Divide result；
Step S1042: described character zone is carried out the difference of the second pre-set color passage, it is thus achieved that second is poor
Divide result；
Wherein, described first pre-set color passage refers to two on off state designators with described banister
In the Color Channel that matches of the color of a designator, such as it can be the face that designator " closes "
Chrominance channel, if it is red that described designator " closes ", the most described first pre-set color passage is redness
Passage；Arranging of described second pre-set color passage regular arranges rule with described first pre-set color passage
Then being similar to, such as it can be and the green green color channel that " dividing ", designator matched；
Step S1043: described first difference result and the second difference result are sued for peace；
Step S1044: export the switch shape of the on off state for characterizing banister matched with summed result
State judges signal；
In described step S1043 and S1044, after getting the first difference result and the second difference result,
Described difference result is sued for peace, i.e. can determine that the on off state of described banister according to the value of summed result
The color of designator, and then deciliter state of described banister can be judged according to this color, such as, when described
First pre-set color passage is green channel, the second pre-set color passage when being red channel, now, as
The most described first difference result and the second difference result sum big 0 time, then it is believed that described character zone is whole
Body is partial to redness, and the most described banister is in " conjunction " state, if both sums are less than 0, then
It is believed that described character zone entirety is partial to green, the most described banister is in " dividing " state.
The color characteristic of image is produced dry by the external factor such as applicant passes through further study show that, illumination
Disturb, if to be positioned at marginal value (the most green and red for described first difference result and the second difference result sum
During the marginal value 0 of color) time, the on off state judged result now obtained may be inaccurate, is directed to
This, the application is above-mentioned judged center may be incorporated into threshold value (positive and negative a pair threshold value) reduce outside because of
The element interference to judged result, particularly as follows: make described first difference result and the second difference result sum
For difference result, it is judged that whether described difference result is in preset range, if it is, think now
The judged result of banister state is effective, otherwise, it is judged that result is invalid, existing also with the first pre-set color passage
As a example by being red channel for green channel, the second pre-set color passage, if difference result is more than described pre-
If the positive peak of scope or during less than the negative minimum of described preset range, then it is assumed that sentencing of banister state
Disconnected result is effective, and otherwise judged result is invalid.When described difference result is positioned within described preset range,
Owing to the on off state value of described banister the most generally only has " dividing " and " conjunction " or " OFF " and " ON ",
Therefore, default preset characters model now can be used to assist the judgement of banister state, now use
Preset characters model similar with the establishment principle of above-mentioned supporting vector machine model, trained by great amount of samples
Generation can distinguish be " dividing " character and " conjunction " character, " OFF " character and " ON " character or other
The model of character, utilizes model to go to identify the character on banister so that it is determined that state.Concrete, see Fig. 5,
Step S104 disclosed in the above embodiments of the present application, may include that
Step S1041: described character zone is carried out the difference of the first pre-set color passage, it is thus achieved that first is poor
Divide result；
Step S1042: described character zone is carried out the difference of the second pre-set color passage, it is thus achieved that second is poor
Divide result；
Step S1043: described first difference result and the second difference result are sued for peace；
Step S1045: judge the value of described summed result whether in preset range, if it is, perform step
Rapid S1044, otherwise, performs step S1046；
Step S1044: export the switch shape of the on off state for characterizing banister matched with summed result
State judges signal；
Step S1046: use preset characters model that described character zone is judged, obtain and export use
On off state indication signal in the on off state characterizing banister.
Certainly, when training pattern, in addition to using support vector machine to be trained, it would however also be possible to employ
Decision tree, neutral net scheduling algorithm carry out model training, i.e. described supporting vector machine model can use
Decision-tree model or log on model etc. are replaced, described preset characters model can use support to
Amount machine, decision tree, the training of log on scheduling algorithm generate.
It is understood that corresponding with said method, disclosed herein as well is a kind of banister state analysis
System, this system can mutually use for reference with said method, see Fig. 6, and this system may include that
Image acquisition units 100, is used for obtaining input picture；
Target location determines unit 200, for by the target determining target banister place in described input picture
Window；
Character zone determines unit 300, for by the on off state determining target banister in described target window
The character zone at pointing character place；
State analysis unit 400, for described character zone carries out difference, exports and difference result phase
The on off state indication signal of the on off state for characterizing described banister joined.
Described image acquisition units 100, after getting the input picture that external equipment sends, uses described
Target location determine unit 200 by described input picture determines the target window at target banister place, when
After described target window determines, unit 300 is by true in described target window to use described character zone to determine
Set the goal the character zone on off state pointing character place of banister, after described character zone determines,
Use described state analysis unit 400 that described character zone carries out difference again, export and difference result phase
The on off state indication signal of the on off state for characterizing described banister of coupling.
Corresponding with said method, see Fig. 7, described target location determines unit 200, specifically can wrap
Include:
Scanning element 210, for using the window of pre-set dimension that described input picture is scanned, will be logical
The sub-input picture region that overscanning gets is as candidate window；
Unit 220 chosen by candidate target window, for by by described candidate window one by one with default Prototype drawing
As carrying out color histogram contrast, it is judged that whether comparing result is more than threshold value, if it does not, extract described time
Select the gal cypress characteristics of image of window and carry out the judgement of supporting vector machine model, will determine that result is positive window
Mouth is as candidate target window, and described supporting vector machine model is the Jia Bai by extracting each training sample
Characteristics of image is supported vector machine model and trains the model for carrying out target window differentiation obtained；
Target window chooses unit 230, the time of highest scoring during for selecting the judgement of supporting vector machine model
Select target window as target window.
Concrete, described target window chooses unit 230 when determining described target window, can be by owning
Candidate target window supporting vector machine model judged result in, choose the candidate target window of highest scoring
Mouth, as target window, monitors the judged result obtained in real time, when judged result is more than presetting judge index
Time, determine that candidate target window corresponding to this judged result is as target window.
Corresponding with said method, see Fig. 7, described character zone determines unit 300, may include that
Matrix model construction unit 310, for utilizing the bright dark feature of target window to build matrix model；
Integrogram computing unit 320, for using integrogram to calculate the average of each element blocks in matrix model；
Difference identifying unit 330, for determine each element blocks in described matrix model and be adjacent its
The maximum of the difference of its element blocks, forms maximum set；
Character zone positioning unit 340, for determining the maximum in described maximum set in target window
The region at one or more element blocks places, using this region as character zone.
Corresponding with said method, see Fig. 7, described state analysis unit 400, may include that
First difference unit 410, for described character zone being carried out the difference of the first pre-set color passage,
Obtain the first difference result；
Second difference unit 420, for described character zone being carried out the difference of the second pre-set color passage,
Obtain the second difference result；
On off state identifying unit 430, for described first difference result and the second difference result are sued for peace,
And the on off state exporting the on off state for characterizing banister matched with summed result judges signal.
Corresponding with said method, see Fig. 8, described on off state identifying unit 430 specifically may include that
Dif ference judgment unit 431, for judging described first difference result difference with the second difference result
Whether absolute value is more than preset value, if it is, output the first triggering signal, otherwise output the second triggering letter
Number；
First identifying unit 432, for after getting described first triggering signal, it is judged that described first poor
Point result and the size of the second difference result, export and described described first difference result and second poor of comparing
Point on off state that result sum matches judges signal；
Second identifying unit 433, for, after getting described second triggering signal, using preset characters mould
Described character zone is judged by type, obtains and export the switch shape of on off state for characterizing banister
State indication signal, described preset characters model is to judge mesh by what training obtained for foundation character zone
The model that mark window on off state judges.
In this specification, each embodiment uses the mode gone forward one by one to describe, and each embodiment stresses
Being the difference with other embodiments, between each embodiment, identical similar portion sees mutually.
For system disclosed in embodiment, owing to it corresponds to the method disclosed in Example, so describing
Fairly simple, relevant part sees method part and illustrates.
Described above to the disclosed embodiments, makes professional and technical personnel in the field be capable of or uses
The present invention.Multiple amendment to these embodiments will be aobvious and easy for those skilled in the art
See, generic principles defined herein can without departing from the spirit or scope of the present invention,
Realize in other embodiments.Therefore, the present invention is not intended to be limited to the embodiments shown herein,
And it is to fit to the widest scope consistent with principles disclosed herein and features of novelty.
Claims (10)
1. a banister state analysis method, it is characterised in that including:
Obtain input picture；
By the target window determining target banister place in described input picture；
Character zone by the on off state pointing character place determining target banister in described target window；
Described character zone is carried out difference, output and difference result match for characterizing described banister
The on off state indication signal of on off state.
Banister state analysis method the most according to claim 1, it is characterised in that described by described
Input picture determines the target window at target banister place, including:
Described input picture is scanned by the window using pre-set dimension, the son that will be got by scanning
Input picture region is as candidate window；
By described candidate window is carried out color histogram contrast with default template image one by one, it is judged that right
Than result whether more than threshold value, if it does not, extract the gal cypress characteristics of image of described candidate window and carry out propping up
Hold the judgement of vector machine model, will determine that result is that positive window is as candidate target window, described support
Vector machine model be by extract the gal cypress characteristics of image of each training sample be supported vector machine model instruct
The model for carrying out target window differentiation got；
Select supporting vector machine model judgement time highest scoring candidate target window as target window.
Banister state analysis method the most according to claim 1, it is characterised in that described by described
Target window determines the character zone at the on off state pointing character place of target banister, including:
The bright dark feature utilizing target window builds matrix model；
Integrogram is used to calculate the average of each element blocks in matrix model；
Determine each element blocks in described matrix model and the maximum of the difference of other element blocks being adjacent
Value, forms maximum set；
By element blocks corresponding for one or more maximums of the maximum in described maximum set at target window
In Kou, area defined is as character zone.
Banister state analysis method the most according to claim 1, it is characterised in that described to described
Character zone carries out difference, the on off state for characterizing described banister that output matches with difference result
Condition indicative signal, including:
Described character zone is carried out the difference of the first pre-set color passage, it is thus achieved that the first difference result；
Described character zone is carried out the difference of the second pre-set color passage, it is thus achieved that the second difference result；
Described first difference result and the second difference result are sued for peace；
The on off state exporting the on off state for characterizing banister matched with summed result judges letter
Number.
Banister state analysis method the most according to claim 4, it is characterised in that described to described
First difference result and the summation of the second difference result；Output and summed result match for characterizing banister
On off state on off state judge signal, including:
Judge whether described first difference result and the second difference result sum are positioned at preset range, if
It is to export and compare described first difference result with described and switch shape that the second difference result sum matches
State judges signal, if it does not, use preset characters model to judge described character zone, obtains also
Output is for characterizing the on off state indication signal of the on off state of banister, and described preset characters model is logical
Cross the model for judging the judgement of target window on off state according to character zone that training obtains.
6. a banister state analysis system, it is characterised in that including:
Image acquisition units, is used for obtaining input picture；
Target location determines unit, for by the target window determining target banister place in described input picture
Mouthful；
Character zone determines unit, for by determining in described target window that the on off state of target banister refers to
Show the character zone at character place；
State analysis unit, for described character zone carries out difference, output matches with difference result
The on off state indication signal of on off state for characterizing described banister.
Banister state analysis system the most according to claim 6, it is characterised in that described target position
Put and determine unit, including:
Scanning element, for using the window of pre-set dimension to be scanned described input picture, will pass through
The sub-input picture region that scanning gets is as candidate window；
Unit chosen by candidate target window, for by by described candidate window one by one with default template image
Carry out color histogram contrast, it is judged that whether comparing result is more than threshold value, if it does not, extract described candidate
The gal cypress characteristics of image of window also carries out the judgement of supporting vector machine model, will determine that result is positive window
As candidate target window, described supporting vector machine model is the Jia Baitu by extracting each training sample
The model for carrying out target window differentiation obtained is trained as feature is supported vector machine model；
Target window chooses unit, the candidate of highest scoring during for selecting the judgement of supporting vector machine model
Target window is as target window.
Banister state analysis system the most according to claim 6, it is characterised in that described character area
Territory determines unit, including:
Matrix model construction unit, for utilizing the bright dark feature of target window to build matrix model；
Integrogram computing unit, for using integrogram to calculate the average of each element blocks in matrix model；
Difference identifying unit, for determine each element blocks in described matrix model and be adjacent other
The maximum of the difference of element blocks, forms maximum set；
Character zone positioning unit, for determining maximum 1 in described maximum set in target window
Individual or the region at multiple element blocks place, using this region as character zone.
Banister state analysis system the most according to claim 6, it is characterised in that described state is divided
Analysis unit, including:
First difference unit, for described character zone carries out the difference of the first pre-set color passage, obtains
Obtain the first difference result；
Second difference unit, for described character zone carries out the difference of the second pre-set color passage, obtains
Obtain the second difference result；
On off state identifying unit, for described first difference result and the second difference result are sued for peace, and
The on off state exporting the on off state for characterizing banister matched with summed result judges signal.
Banister state analysis system the most according to claim 9, it is characterised in that described switch
State determination unit, including:
Dif ference judgment unit, is used for judging described first difference result and the second difference result sum whether position
In preset range, if it is, output the first triggering signal, otherwise output the second triggering signal；
First identifying unit, for after getting described first triggering signal, it is judged that described first difference
Result and the size of the second difference result, output compares described first difference result and the second difference with described
The on off state that result sum matches judges signal；
Second identifying unit, for, after getting described second triggering signal, using preset characters model
Described character zone is judged, obtains and export the on off state of on off state for characterizing banister
Indication signal, described preset characters model is to judge target by what training obtained for foundation character zone
The model that window on off state judges.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610187111.7A CN105894015B (en) | 2016-03-28 | 2016-03-28 | A kind of banister state analysis method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610187111.7A CN105894015B (en) | 2016-03-28 | 2016-03-28 | A kind of banister state analysis method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105894015A true CN105894015A (en) | 2016-08-24 |
| CN105894015B CN105894015B (en) | 2019-08-20 |
Family
ID=57013940
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610187111.7A Active CN105894015B (en) | 2016-03-28 | 2016-03-28 | A kind of banister state analysis method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105894015B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019062818A1 (en) * | 2017-09-30 | 2019-04-04 | 施耐德电气工业公司 | Method and device for identifying state of electrical apparatus |
| CN109711257A (en) * | 2018-11-27 | 2019-05-03 | 成都宜泊信息科技有限公司 | A kind of banister condition detection method and system based on image recognition |
| CN110443229A (en) * | 2019-08-22 | 2019-11-12 | 国网四川省电力公司信息通信公司 | A kind of equipment display content identification method based on artificial intelligence |
| CN112418226A (en) * | 2020-10-23 | 2021-02-26 | 济南信通达电气科技有限公司 | Method and device for identifying opening and closing states of fisheyes |
| CN112634352A (en) * | 2020-10-30 | 2021-04-09 | 国网山东省电力公司济宁供电公司 | Transformer substation fisheye type opening and closing state identification method and system |
| CN114928720A (en) * | 2022-05-13 | 2022-08-19 | 重庆云凯科技有限公司 | System and method for detecting state of parking barrier gate rod |
| US11961292B2 (en) | 2021-01-21 | 2024-04-16 | Abb S.P.A. | Computer-implemented method for assisting a user in interacting with an electronic relay for electric power distribution grids |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101470910A (en) * | 2007-12-27 | 2009-07-01 | 程雄 | Automatic recognition management system for vehicle thoroughfare |
| CN102799173A (en) * | 2012-08-03 | 2012-11-28 | 严浙恒 | Intelligent fault diagnosis device of parking lot and road charging and channel managing system |
| CN103745596A (en) * | 2014-01-16 | 2014-04-23 | 深圳市旺龙智能科技有限公司 | Abnormal vehicle passing detection system and abnormal vehicle passing detection method |
-
2016
- 2016-03-28 CN CN201610187111.7A patent/CN105894015B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101470910A (en) * | 2007-12-27 | 2009-07-01 | 程雄 | Automatic recognition management system for vehicle thoroughfare |
| CN102799173A (en) * | 2012-08-03 | 2012-11-28 | 严浙恒 | Intelligent fault diagnosis device of parking lot and road charging and channel managing system |
| CN103745596A (en) * | 2014-01-16 | 2014-04-23 | 深圳市旺龙智能科技有限公司 | Abnormal vehicle passing detection system and abnormal vehicle passing detection method |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019062818A1 (en) * | 2017-09-30 | 2019-04-04 | 施耐德电气工业公司 | Method and device for identifying state of electrical apparatus |
| US11302099B2 (en) | 2017-09-30 | 2022-04-12 | Schneider Electric Industries Sas | Method and device for recognizing states of electrical devices |
| CN109711257A (en) * | 2018-11-27 | 2019-05-03 | 成都宜泊信息科技有限公司 | A kind of banister condition detection method and system based on image recognition |
| CN110443229A (en) * | 2019-08-22 | 2019-11-12 | 国网四川省电力公司信息通信公司 | A kind of equipment display content identification method based on artificial intelligence |
| CN112418226A (en) * | 2020-10-23 | 2021-02-26 | 济南信通达电气科技有限公司 | Method and device for identifying opening and closing states of fisheyes |
| CN112418226B (en) * | 2020-10-23 | 2022-11-25 | 济南信通达电气科技有限公司 | Method and device for identifying opening and closing states of fisheyes |
| CN112634352A (en) * | 2020-10-30 | 2021-04-09 | 国网山东省电力公司济宁供电公司 | Transformer substation fisheye type opening and closing state identification method and system |
| CN112634352B (en) * | 2020-10-30 | 2023-02-17 | 国网山东省电力公司济宁供电公司 | Transformer substation fisheye type opening and closing state identification method and system |
| US11961292B2 (en) | 2021-01-21 | 2024-04-16 | Abb S.P.A. | Computer-implemented method for assisting a user in interacting with an electronic relay for electric power distribution grids |
| CN114928720A (en) * | 2022-05-13 | 2022-08-19 | 重庆云凯科技有限公司 | System and method for detecting state of parking barrier gate rod |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105894015B (en) | 2019-08-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105894015A (en) | Gate state analyzing method and system | |
| CN105788142B (en) | A kind of fire detection system and detection method based on Computer Vision | |
| CN105868722A (en) | Identification method and system of abnormal power equipment images | |
| CN109308447A (en) | The method of equipment operating parameter and operating status is automatically extracted in remote monitoriong of electric power | |
| CN103632158B (en) | Forest fire prevention monitor method and forest fire prevention monitor system | |
| CN104331710B (en) | On off state identifying system | |
| CN104809463B (en) | A kind of high-precision fire disaster flame detection method for converting dictionary learning based on intensive scale invariant feature | |
| CN108537154A (en) | Transmission line of electricity Bird's Nest recognition methods based on HOG features and machine learning | |
| CN103761529A (en) | Open fire detection method and system based on multicolor models and rectangular features | |
| CN110675588B (en) | Forest fire detection device and method | |
| CN105404867B (en) | A kind of substation isolating-switch state identification method of view-based access control model | |
| CN109977897A (en) | A kind of ship's particulars based on deep learning recognition methods, application method and system again | |
| CN109118548A (en) | A kind of comprehensive intelligent water quality recognition methods | |
| CN106971182A (en) | Embedded electric power relay pressing plate is thrown and moves back state Intelligent Identify device and implementation method | |
| CN106339657B (en) | Crop straw burning monitoring method based on monitor video, device | |
| CN102208018A (en) | Method for recognizing fire disaster of power transmission line based on video variance analysis | |
| CN105844245A (en) | Fake face detecting method and system for realizing same | |
| CN107067412A (en) | A kind of video flame smog detection method of Multi-information acquisition | |
| CN111724290B (en) | Environment-friendly equipment identification method and system based on depth layering fuzzy algorithm | |
| CN112163572A (en) | Method and device for identifying object | |
| CN104766338A (en) | Method for detecting significance of complex X-ray pseudo-color image | |
| Zhu et al. | Detection for triangle traffic sign based on neural network | |
| CN108563997A (en) | It is a kind of establish Face datection model, recognition of face method and apparatus | |
| CN113033548A (en) | Image recognition method for unit running state judgment and monitoring point alignment | |
| KR102602439B1 (en) | Method for detecting rip current using CCTV image based on artificial intelligence and apparatus thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |