CN102663360B - Method for automatic identifying steel slab coding and steel slab tracking system - Google Patents
Method for automatic identifying steel slab coding and steel slab tracking system Download PDFInfo
- Publication number
- CN102663360B CN102663360B CN 201210091150 CN201210091150A CN102663360B CN 102663360 B CN102663360 B CN 102663360B CN 201210091150 CN201210091150 CN 201210091150 CN 201210091150 A CN201210091150 A CN 201210091150A CN 102663360 B CN102663360 B CN 102663360B
- Authority
- CN
- China
- Prior art keywords
- character
- image
- monocase
- coding
- slab
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Character Discrimination (AREA)
Abstract
The invention discloses a method for automatic identifying steel slab coding and a steel slab tracking system. The method is characterized in that: from the starting of a first frame original image, automatic identification of slab coding is successively carried out on each frame of original image by the following steps: step 1, carrying out image preprocessing on the original image; step 2, carrying out binarization processing on the image after the preprocessing and carrying out slab coding detection and coding position positioning on an obtained binary image, wherein the slab coding detection and coding position positioning employ a projection processing method; and step 3, splitting the original image according to a boundary coordinate that is obtained by the coding position positioning to obtain a plurality of single character images and carrying out character identification on the each single character image, thereby completing slab coding identification on the original image ofthe current frame; and then returning to the step 1. According to the invention, the method for automatic identifying steel slab coding and the steel slab tracking system have characteristics of highautomation degree and high coding identification efficiency.
Description
Technical field
The present invention relates to a kind of iron and steel slab coding automatic identifying method and iron and steel tracking system of plate blanks.
Background technology
Development along with industrial automation, in steel industry, the requirement to the logistics management automatization level has progressively improved in enterprise, iron and steel slab information is that iron and steel slab coding is the important information of iron and steel enterprise's logistics management and production run, usually adopt certain letter or number and assembly coding thereof to represent, automatic acquisition iron and steel slab information is significant to the iron and steel slab tracking in the logistics management and enterprise automation and e-manufacturing.The iron and steel slab tracking refers to slab cutting information and the information of weighing of the slab management information picture automatic reception continuous casting level two in the steel-making side material module in the MES system.
At present, most of steel plant, the identification of iron and steel slab coding also rests on the basis of artificial cognition, lacks the measure that iron and steel slab coding is identified automatically.Simultaneously, existing character recognition technologies is mainly used in the simple occasion of optical imagery background, and the slab production line of iron and steel sheet and plate exists circumstance complication, photoenvironment to change, the impact of the abominable factor such as the character imaging background complexity of iron and steel slab side, existing character recognition technologies can't be advantageously applied to iron and steel slab code identification.In addition, the inefficiencies of artificial cognition has a strong impact on the gentle Modern Operations System of Automated water perfect of the iron and steel slab tracking in the logistics management.
At present, the iron and steel slab tracking is mainly realized the tracking to front and back two-step slab in the MES system.This semi-automatic iron and steel tracking system of plate blanks can not be accurately and real-time follow-up be sent to the slab situation of hot rolling mill from the converter shed, to the information change that produces in this transfer process, the MES system is not revised in real time, so that the information in the MES system and actual iron and steel slab information are variant, operation logistics consistent with information in kind before and after can not really realizing do not have really to realize that the iron and steel slab of a level system follows the tracks of automatically.
Summary of the invention
Technical matters to be solved by this invention provides a kind of iron and steel slab coding automatic identifying method and iron and steel tracking system of plate blanks, this iron and steel slab coding automatic identifying method and iron and steel tracking system of plate blanks have the automaticity height, the characteristics that code identification efficient is high.
The technical solution of invention is as follows:
A kind of iron and steel slab coding automatic identifying method obtains continuous multiframe original image at the transmission scene of iron and steel slab, since the first frame original image, successively each frame original image is identified the slab coding according to the following steps automatically:
Step 1: original image is carried out the image pre-service; Described pre-service comprises that gray processing is processed, mean filter is processed and difference processing; Described difference processing refers to filtered image and Background are made difference processing, obtains the difference gray-scale map; When described Background refers to not have slab in original image is judged by system, with filtered Image Saving or be updated to background picture;
Step 2: pretreated image is carried out binary conversion treatment, again the binary map of gained is carried out the slab code detection, generate testing result, in the original image of pre-treatment, contain slab coding if recognize according to testing result, then binary map is carried out slab coding site location, the projection process method is adopted in described slab code detection and coding site location;
Step 3: original image is carried out cutting according to the boundary coordinate that aforesaid coding site location obtains, obtain a plurality of monocase images, described monocase image refers to only contain a character-coded image; Each monocase image is carried out character recognition, thereby finish the slab code identification in the current frame original image;
Returning step 1 pair next frame original image processes.
In the step 2, the threshold value of described binary conversion treatment is avg+N, and the N value is the arbitrary value in 20～70; Avg is the mean value of gray-scale map entire image grey scale pixel value after the difference.
Slab code detection process in the step 2 is: described binary map is carried out vertical projection, if there are intersection point in resulting vertical projection curve and preset level line, then judge and contain the slab coding in the present image, otherwise do not have the slab coding in the explanation present image; The preset level line the vertical projection Curves the ordinate of coordinate system be 15;
Described vertical projection refers to calculate the quantity of the white pixel point of each row in the binary map as the Y-axis coordinate figure of vertical projection curve, with the X-axis coordinate figure of the pixels across point sequence number in the binary map as the vertical projection curve, forms the vertical projection curve.
The process of the slab coding site location in the step 2 is:
(1) binary map is carried out horizontal projection, obtain the horizontal projection curve; Two intersection points of horizontal projection curve and preset level line are the up-and-down boundary point;
(2) vertical projection curve and preset level line intersect, and form many antinodes, and every a pair of adjacent intersection point i.e. the border, the left and right sides of a monocase of correspondence;
(3) based on the border, the left and right sides of described up-and-down boundary point and each monocase, binary map is carried out cutting, obtain a plurality of monocase images;
Described horizontal projection refers to calculate the quantity of the white pixel point of every delegation in the binary map as the Y-axis axial coordinate value of horizontal projection curve, with the sequence number of the vertical pixel in the binary map X-axis axial coordinate value as the horizontal projection curve, forms the horizontal projection curve.
The preset level line horizontal projection curve and vertical projection Curves the ordinate of coordinate system be 15.
Monocase image character identification in the step 3 may further comprise the steps:
Step a: the monocase image is carried out gray processing, obtain the monocase gray-scale map;
Step b: described monocase gray-scale map is carried out filtering process, obtain filtered monocase gray-scale map;
Step c: described filtered monocase gray-scale map is carried out binary conversion treatment, obtain the monocase binary map;
Steps d: described monocase binary map is carried out size normalization process, obtain normalization monocase binary map;
Step e: described normalization monocase binary map is pursued the pixel feature extraction, obtain eigenvectors matrix; Describedly refer to by the pixel feature extraction: normalization monocase binary map is scanned line by line, to the white pixel point in the normalization monocase binary map, getting eigenwert is 1, to the black pixel point in the normalization monocase binary map, getting eigenwert is 0, obtain at last an eigenvectors matrix, the dimension of eigenvectors matrix equals the total number of pixel in the normalization monocase binary map;
Step f: pursue the pixel template matches:
Described eigenvectors matrix and a plurality of Character mother plate are mated, obtain a plurality of similarities, similarity refers to that the identical element number accounts for the ratio of total element number; With the matching result of the corresponding character of the Character mother plate of similarity maximum as current monocase image; The number of elements that comprises in number of elements in each Character mother plate and the eigenvectors matrix is identical.
Among the described step f, the matching way of described eigenvectors matrix and Character mother plate is: successively each element in the comparative feature vector matrix and the corresponding element in the character masterplate, judge whether identical, and the number of statistics identical element, the number of identical element is accounted for the ratio of total element number as the similarity of current monocase image with this current Character mother plate.
Among the described step f, the manufacturing process of described Character mother plate is as follows: at first obtain the slab coded image, be encoded to one group to comprise all single slabs, choose the monocase image of one group of slab coding, through the processing of step a to step e, obtain one group and slab coded image characteristic of correspondence vector matrix, this stack features is stored as file, be one group of Character mother plate corresponding with the slab coded image.
The binary conversion treatment of carrying out described monocase gray-scale map adopts large Tianjin method;
Described size normalization is processed and is referred to binary map is converted to 50 * 60 Pixel Dimensions sized images;
Described Character mother plate is many groups;
The character that the character group that described Character mother plate is corresponding comprises is at least a in digital 0-9, English capitalization, English lower case, the Chinese character.
In the code identification process, whenever identify a character, then the degree of confidence that this character is corresponding increases by 1, and the degree of confidence initial value is 0; After last coding of any iron and steel slab finished code identification, if the corresponding a plurality of identification characters of some codings are then got the highest identification character of degree of confidence as final recognition result.
A kind of slab automatic recognition and tracking system based on iron and steel slab coded image comprises ccd video camera, light end transmitter, light end receiver, display, is used for realizing described iron and steel slab coding automatically industrial computer and the on-the-spot PLC of identification with capture card; The output terminal of ccd video camera is connected with the input end of light end transmitter, and the output terminal of light end transmitter is connected by optical fiber with the input end of light end receiver; Capture card is connected with industrial computer by the PCI slot; Display is connected with industrial computer, and on-the-spot PLC is connected with the communication port of industrial computer.
Described slab automatic recognition and tracking system based on iron and steel slab coded image also comprises video distributor, monitoring screen and DVR, and light end receiver links to each other with capture card by video distributor; Monitoring screen is connected input end and all is connected with the output terminal of video distributor with DVR.
Beneficial effect:
Iron and steel slab coding automatic identifying method of the present invention and iron and steel tracking system of plate blanks, substitute and manually carry out iron and steel slab code identification, effectively avoid because the iron and steel slab coding mistake identification that workman's subjective reason causes, strengthened the automatization level of Inner Logistics Management, improve the production efficiency of enterprise, and reduced operating personnel's labour intensity.Iron and steel slab after identification coding is passed to the MES system, automatically realize the tracking of iron and steel slab information between converter shed and hot rolling mill, thereby reach all fronts of iron and steel slab information are followed the tracks of, improve the logistics management level between two factories.The machine Interaction Interface Design of this system is simple, is convenient to the site operation personnel and uses.This system recognition rate is high, and speed is fast, and working stability is reliable, has stronger practical.
Character identifying method in the iron and steel slab coded image of the present invention, feature for iron and steel slab coding, adopt a kind of new method to carry out feature extraction, improved the comprehensive of image information features extraction, it is simple to have feature extraction, computing is convenient, and can effectively realize the identification of iron and steel slab coding.
The present invention adopts effective Preprocessing Technique, and extracts appropriate iron and steel slab coded character feature, in conjunction with iron and steel slab coded character inspection technology, realizes the effective identification to iron and steel slab coding in the iron and steel slab image that obtains in the complex industrial environment.
Because iron and steel slab coding is normally formed by the industrial coating spraying, and industrial coating is granular, cause iron and steel slab coding connectedness and standardization not good, thereby cause existing feature extracting method can't effectively extract the character feature of iron and steel slab coding.And the characteristic extraction procedure of this method is simple conversion process, is about to black (gray-scale value is 0) pixel and white (gray-scale value is 255) pixel in the binary map, represents with eigenwert 0 and 1 respectively, thereby obtains proper vector.Thereby method of the present invention can comprehensively be extracted image information features, and characteristic extraction procedure is simple, and computing is convenient, realizes easily.
The present invention also has following advantage:
Discrimination is high: for iron and steel slab coded image target detection and cutting and monocase image recognition, adopted respectively suitable preprocess method to process, improved the success ratio of target detection and character recognition.
Wherein: the pre-service of target detection and cutting comprises that gray processing, mean filter and modified manually arrange binary conversion treatment; The pre-service of monocase image recognition comprises gray processing, mini-value filtering and large Tianjin method binary conversion treatment.
Simultaneously, for character recognition, select appropriate feature extraction and template matching method, namely by the pixel feature extraction with by the pixel template matches, improved the comprehensive of image information features extraction, and the validity of template matching method.
In addition, by verification, guaranteed the accuracy of output recognition result.
Because the slab coding may be damaged, cause relevant similar character the situation of mistake identification to occur, the similar character of slab coding has: 1 and 7,6 and 8,0 and 8,3 and 5,2 and 7.The processing of two aspects is arranged for similar character:
For similar character, such as 1 and 7,6 and 8.In the situation that Character mother plate is constant in template base, difference for these symbol local feature performances of easily becoming literate by mistake, adopt the half template (first half, Lower Half, left side or right-hand part) coupling identification, when calculating similarity, the standard form coupling that the character to be identified of mistake identification also will be similar to it easily, take similar character 6 and 8 as example, character 6 is identified as 8 sometimes, by analyzing as can be known, 6 and 8 first half global feature otherness is maximum, therefore after the template matches identification first time, with recognition result be 8 character get its first half again with template base in character 6 and 8 the first half mate identification, the similarity of twice coupling is added up, and getting its maximal value is last recognition result.Reduced the false recognition rate to similar character.
Recognition result and the iron and steel slab coded message of obtaining from on-the-spot PLC10 (on-the-spot PLC10 is original equipment) are compared, and herein, if do not find identical slab coding, and recognition result is (such as 16060
03024020) encode (such as 16060 with certain slab
83024020) the character difference numeral of underscore (namely with) is only arranged, this a pair of kinds of characters, just be similar character pair, then this moment, this iron and steel slab coding (such as 1606083024020) is set to recognition result, and (then final recognition result is: 1606083024020).
Speed is fast: each software module of system's parallel calling is carried out multiple task management and parallel processing, has effectively improved processing power and the speed of system.
Each modular algorithm principle is simple, and program realizes convenient, and has adopted ripe Effective arithmetic, has guaranteed the execution efficient of algorithm.Especially in traditional character identifying method, Pattern matching aspect relatively consuming time, system only just finishes by simply conversion and comparison, has effectively improved the travelling speed of system.Simultaneously, system software structure is reasonable in design, and is optimized, and guarantees effective, the fast express agency of system.
Reliable and stable: system adopts parallel processing mode, has increased system reliability.Each modular algorithm principle is simple, and program realizes convenient, and has adopted ripe Effective arithmetic, has guaranteed the reliability of algorithm.System software structure is reasonable in design, and is optimized, and has reduced the probability of system exception.Man Machine Interface is provided, can have artificially processed timely the relevant abnormalities situation.That local data base adopts is ripe, oracle database and ADO technology are applied to System data management reliably.Native system adopts Visual C++6.0 language development in Windows Sever 2003 systems, has adopted ripe, reliable development platform and running environment.
Description of drawings
Fig. 1 is the hardware architecture block diagram based on the iron and steel slab automatic tracking system of iron and steel slab coded image identification;
Fig. 2 is the principle of work block diagram based on the iron and steel slab automatic tracking system of iron and steel slab coded image identification;
Fig. 3 is the software flow pattern based on the iron and steel slab automatic tracking system of iron and steel slab coded image identification;
Fig. 4 is the iron and steel slab image that contains 3 characters;
Fig. 5 is the iron and steel slab image that contains 5 characters;
Fig. 6 is the iron and steel slab image that contains 7 characters;
Fig. 7 is the iron and steel slab image that contains 10 characters;
Fig. 8 is the iron and steel slab image that first width of cloth contains complete 13 characters;
Fig. 9 is the iron and steel slab image that second width of cloth contains complete 13 characters;
Figure 10 for the normalization of 3 monocase images Fig. 4 being carried out cutting and obtain after the monocase binary map;
Figure 11 for the normalization of 5 monocase images Fig. 5 being carried out cutting and obtain after the monocase binary map
Figure 12 for the normalization of 7 monocase images Fig. 6 being carried out cutting and obtain after the monocase binary map
Figure 13 for the normalization of 10 monocase images Fig. 7 being carried out cutting and obtain after the monocase binary map
Figure 14 is the normalization effect synoptic diagram of whole 13 monocase images;
Figure 15 is the horizontal projection curve;
Figure 16 is the vertical projection curve;
The single character picture synoptic diagram of Figure 17 after for corresponding with Figure 16 cutting apart.
Embodiment
Below with reference to the drawings and specific embodiments the present invention is described in further details:
What the present invention relates to may further comprise the steps the monocase image-recognizing method:
1. the image that contains single slab coded character that cutting from iron and steel slab coded image is obtained is numbered in order, and carries out gray processing and process;
2. monocase (containing single iron and steel slab coded character) gray-scale map being carried out filtering processes;
3. image adopts the self-adaption binaryzation cutting techniques after filtering being processed, and carries out binary conversion treatment;
4. to the monocase binary map, carry out size normalization and process;
5. to monocase binary map after the normalization, pursue the pixel feature extraction.The key content of this patent has 2 points: 1, the appropriate combination of gray processing, filtering and binaryzation technology; (existing main array mode: gray processing, mean filter and manually set the global threshold method; ) 2, by the pixel feature extraction.
The gray feature of selecting character picture by the pixel feature extraction belongs to a kind of character statistics and transform characteristics as character feature, adopts greyscale transformation and binary conversion treatment to obtain the gray feature vector matrix of single character picture.
With the something in common of prior art be: by the characteristic information of mapping algorithm extraction character, mapping algorithm is ripe algorithm;
Difference is: mapping algorithm reasonably combined, and the choosing of character feature, in the document that has searched, do not find the character feature identical with this patent; Adopt greyscale transformation and binary conversion treatment to obtain the gray feature vector matrix of single character picture.
6. to the monocase binary map, pursue the pixel template matches;
7. repeating step 1 arrives step 6, until identify all the single characters in complete the slab coded image;
8. to all the single characters in the whole iron and steel slab coded image of identifying rear acquisition, the numbering during according to cutting makes up in order, obtains complete iron and steel slab coding.The complete iron and steel slab coding that identification is obtained can carry out checking treatment according to the coded format of self-defining.The effect of checking treatment: there are the situations such as damaged in iron and steel slab coding in image, and when causing recognition result to occur violating the situation of coding rule, by verification, can find problems, and the output information warning, reduces the probability of output error recognition result.The standard of verification succeeds refers to that whether the coded string of identifying acquisition satisfies whole coding rules, if satisfy whole coding rules, then judges verification succeeds; Otherwise verification failure.
If verification succeeds proves and identifies successfully, output iron and steel slab code identification result; If the verification failure proves this recognition failures, and output information warning " recognition failures! ";
Further:
In described step 1, the transformation for mula that gray processing is processed is as follows:
f(i，j)＝0.299R(i，j)+0.587G(i，j)+0.114B(i，j))；
Wherein, i, j are positive integer, the coordinate of pixel in (i, j) presentation video; The gray-scale value of pixel behind the f representation conversion; R, G, B are the three-components of cromogram RGB model, represent respectively the three-component intensity level of red, green, blue.
In described step 2, the window of filtering algorithm employing 3 * 3 carries out mini-value filtering;
Mini-value filtering: principle is the minimum value that certain any gray-scale value is set to all pixel gray-scale values in this vertex neighborhood window in the image.Specific implementation: sort by the grey scale pixel value to the neighborhood window, form the ordered data of dull decline (or rising), obtain the wherein minimum value of pixel gray-scale value, and the gray-scale value of current pixel point is set to this value.This patent adopts 3 * 3 neighborhood window.
In described step 3, described self-adaption binaryzation cutting techniques refers to large Tianjin (Ostu) method, i.e. maximum variance between clusters.This is a kind of based on discriminant analysis, the method for dynamic calculation optimal threshold, and adaptivity is stronger, and method is simple and practical, and the slab coded image of single character is had good binarization segmentation effect;
In described step 4, described size normalization is processed, and refers to that it is 50 * 60 Pixel Dimensions that the outer rim of character is carried out in proportion that linearity zooms in or out;
In described step 5, described by the pixel feature extraction, refer to normalization monocase binary map is carried out line by line scanning, it is 1 that white pixel in image point (pixel value is 255) is got its eigenwert, it is 0 that black picture element in the image (pixel value is 0) is got its eigenwert, obtain at last an eigenvectors matrix, its dimension equals the total number of pixel in the image.
In described step 6, described by the pixel template matches, feature with the single slab coded character that extracts, mate with 0～9 each Character mother plate, matching way is each element in the simple successively comparative feature vector matrix and the corresponding element in the character masterplate just, judge whether identical, and the number of statistics identical element, the number of identical element is accounted for the ratio of total element number as the similarity of current monocase image with this current Character mother plate, the size of comparative sample and each Character mother plate similarity is matching result with the corresponding character of the Character mother plate of both similarity maximums.The manufacturing process of above-mentioned character masterplate is as follows: at first obtain iron and steel slab coded image, to comprise all single iron and steel slabs codings (namely 0～9) as one group, choose one group and be respectively 0～9 single iron and steel slab coded character image, through the processing of step 1 to step 5, obtain one group of 0～9 iron and steel slab coded character image characteristic of correspondence vector matrix, this stack features is stored as file, is one group of 0～9 character masterplate corresponding to iron and steel slab coded image;
In the described step 7, described complete iron and steel slab coded image contains 13 characters.
Embodiment 1:
In the present embodiment based on the slab automatic recognition and tracking system of iron and steel slab coded image by ccd video camera 1, light end transmitter 2, light end receiver 3, video distributor 4, monitoring screen 5, DVR 6, capture card 7, industrial computer 8 and display 9, and on-the-spot PLC10 forms, as shown in Figure 1, ccd video camera 1 is responsible for taking the slab video.Fiber optic is the equipment that light signal and electric signal are changed mutually, and light end transmitter 2 is to receive electric signal, converts light signal to; Light end receiver 3 is receiving optical signals, converts electric signal to; Effect is to convert video information to optical information, realizes transmission of video by optical fiber, and Optical Fiber Transmission has the advantages such as low, the anti-interference and capacity of loss is large, thus, reduce the loss of vision signal, strengthen antijamming capability and transmission real-time, guarantee in the transmission course quality of vision signal.On-the-spot PLC10 and industrial computer 8 communication connections are used for to industrial computer 8 transmission iron and steel slab codings.
Video distributor 4 is the equipment that a video signal source average mark is made into multi-channel video signal.Be about to ccd video camera 1 outputting video signal and all give monitoring screen 5, DVR 6 and capture card 7.
Monitoring screen 5, just the real-time shooting picture that shows ccd video camera 1 belongs to auxiliary facility, and is irrelevant with the system identification function.
Capture card 7 is used for from the video acquisition of image data of ccd video camera 1 shooting, processes for system's industrial computer 8.Industrial computer 8 is operation platforms of system software, and display 9 is interfaces of man-machine interaction.The identification of the detection by the iron and steel slab image that obtains being carried out slab and slab coding shows recognition result, store and transmits, and finishes automatic identification and the tracking of slab, and the theory diagram of whole system as shown in Figure 2.Whole software systems comprise automatic identification and the human-computer interaction interface of iron and steel slab coding, and software flow figure as shown in Figure 3.
Existing iron and steel slab is about to enter ccd video camera 1 shooting area, and for the process of this iron and steel slab by ccd video camera 1 shooting area, the operational process of describing system in detail is as follows:
The concrete treatment scheme of the first frame original image (Fig. 4):
1. original image is carried out gray processing and process, obtain gray-scale map;
2. adopting 3 * 3 windows to carry out mean filter to gray-scale map processes, obtains filtered gray-scale map;
3. gray-scale map after the filtering and Background are made difference processing, obtain the difference gray-scale map, described Background refers to when not having slab in system's judgement original image, with filtered Image Saving or be updated to background picture;
4. the mean value of gray-scale map entire image grey scale pixel value after the statistics difference, be 60, to be worth with 50 summations and [consider that (the iron sheet oxidation forms for the oxidation spot that exists in the variation of on-the-spot illumination condition and the image, present with hickie in binary map) disturb, simultaneously, the ratio that target character occupies in the slab coded image is less, cause gray average low, split image effectively, thus consider with average and a fixed value and as threshold value, and the variation of on-the-spot illumination condition, making target character gray-scale value and slab gray-scale value difference scope is 20-70, value is too small, because the variation of illumination can cause erroneous judgement; Value is excessive, reduces the sensitivity that detects, and can't in time detect character, considers, and chooses this value], get 110, it as threshold value, is carried out binary conversion treatment to gray-scale map, obtain binary map;
5. binary map is carried out projection process, there are intersection point in vertical projection curve and preset level line (ordinate is 15) [image pixel is of a size of: 900 * 75 pixels], judge to contain the slab coding in the present image; According to horizontal projection curve and vertical projection curve, take the image upper right corner as true origin, determine the upper left corner, the lower left corner, the upper right corner and the lower right corner coordinate of single character;
Employing positions and cutting coding region based on the method for projection properties, and concrete steps are as follows:
Step 1: the image after the binaryzation is carried out horizontal projection, obtain horizontal projection curve as shown in figure 15.The broadband part that an obvious projection is arranged in the drop shadow curve is the horizontal projection of coding region in the image to be identified.Two intersection points of horizontal projection curve and preset level line (ordinate is 15) are the up-and-down boundary point of coding region, such as the edge1 among Figure 15, edge2.
Step 2: the image after the binaryzation is carried out vertical projection, obtain vertical projection curve as shown in figure 16.There are 3 antinodes successively in vertical projection curve and preset level line (ordinate is 15), i.e. a, a ', b, b ', c, c ', wherein a, a ' are the left and right boundary point of first character, and b, b ' are the left and right boundary point of second character, and c, c ' are the left and right boundary point of the 3rd character.
By step 1 and step 2, can obtain the zone location coordinate of each single character.As shown in figure 17 for cutting apart the single character synoptic diagram of acquisition.
6. according to coordinate information, cutting obtains several images that contains single character from original image;
7. several images that contain single character are carried out gray processing and process, obtain the gray level image that several contain single character, be called for short the monocase image;
8. adopting successively 3 * 3 windows to carry out mini-value filtering to several monocase gray-scale maps processes, obtains filtered gray-scale map;
9. filtered gray-scale map is adopted large Tianjin method successively, carry out self-adaption binaryzation and process, obtain binary map;
10. binary map is carried out size normalization successively and process, obtain the binary map of 50 * 60 Pixel Dimensions sizes;
11. to monocase binary map after the normalization, as shown in figure 10, pursue successively the pixel feature extraction, namely to image scanning line by line, it is 1 that white pixel in image point (pixel value is 255) is got its eigenwert, it is 0 that black picture element in the image (pixel value is 0) is got its eigenwert, obtains at last an eigenvectors matrix, and its dimension equals the total number (300) of pixel in the image;
12. to the monocase binary map, pursue successively the pixel template matches, obtain matching result for being followed successively by 1,5,0, similarity is followed successively by S
11=94, S
21=90, S
31=95, and make corresponding degree of confidence, [degree of confidence mainly is as a differentiation when statistics multiframe original image.As add up multiframe original image first character image and be identified as 1 number of times, realize by cumulative degree of confidence, thereby judge the recognition result of first character image.And for example, be respectively 3 and 2 if the first character image is identified as 1 and 7 number of times, judge that then recognition result is 1.] be K
11=1, K
21=1, K
31=1.Detailed process is as follows:
Proper vector to the single iron and steel slab coded character that extracts, mate with the proper vector of 0～9 each template, matching way is the corresponding element element in each element and the character masterplate in the simple successively matching characteristic vector matrix just, judge whether identical, and the number of statistics identical element, final calculating identical element accounts for the ratio of total element number (300), similarity as sample and this Character mother plate, the size of comparative sample and each Character mother plate similarity, the corresponding character of the Character mother plate of both similarity maximums is matching result.The manufacturing process of above-mentioned character masterplate is as follows: at first obtain iron and steel slab coded image, to comprise all single slabs codings (namely 0～9) as one group, choose one group and be respectively 0～9 single iron and steel slab coded character image, through the processing of step 1 to step 5, obtain one group of 0～9 iron and steel slab coded character image characteristic of correspondence vector matrix, this stack features is stored as file, is one group of 0～9 character masterplate corresponding to iron and steel slab coded image.
Front 11 steps of all the other each frame original images process identical with the processing of the first frame original image, and only cutting obtains iron and steel slab coded character number difference.Matching result information is as shown in the table in detail:
The template matches treatment scheme of the second frame original image:
To the monocase binary map, to pursue successively the pixel feature extraction and by the pixel template matches, obtain matching result for being followed successively by 1,5,0,5,8, similarity is followed successively by S
12=96, S
22=89, S
32=93, S
41=85, S
51=83, and make that corresponding degree of confidence is K
12=1, K
22=1, K
32=1, K
41=1, K
51=1.
The concrete treatment scheme of the template matches of the 3rd frame original image:
To the monocase binary map, to pursue successively the pixel feature extraction and by the pixel template matches, obtain matching result for being followed successively by 1,5,0,5,8,6,8, similarity is followed successively by S
13=92, S
23=92, S
33=94, S
42=88, S
52=85, S61=89, S71=87, and make that corresponding degree of confidence is K13=1, K23=1, K33=1, K42=1, K52=1, K61=1, K71=1.
The concrete treatment scheme of the template matches of the 4th frame original image:
To the monocase binary map, to pursue successively the pixel feature extraction and by the pixel template matches, obtain matching result for being followed successively by 1,5,0,5,8,6,8,0,1,2, similarity is followed successively by S14=94, S24=90, S34=96, S43=89, S53=86, S62=91, S72=89, S81=95, S91=93, S101=89, and make that corresponding degree of confidence is K14=1, K24=1, K34=1, K43=1, K53=1, K62=1, K72=1, K81=1, K91=1, K101=1.
The concrete treatment scheme of the template matches of the 5th frame original image:
To the monocase binary map, pursue successively the pixel feature extraction and by the pixel template matches, obtain matching result for being followed successively by 1,5,0,5,8,6,8,0,1,2,0,1,0, similarity is followed successively by S15=92, S25=88, S35=94, S44=86, S54=85, S63=94, S73=88, S82=92, S92=91, S102=90, S111=95, S121=94, S131=93, and make that corresponding degree of confidence is K15=1, K25=1, K35=1, K44=1, K54=1, K63=1, K73=1, K82=1, K92=1, K102=1, K111=1, K121=1, K131=1.
The concrete treatment scheme of the template matches of the 6th frame original image:
To the monocase binary map, pursue successively the pixel feature extraction and by the pixel template matches, obtain matching result for being followed successively by 1,5,0,5,8,6,8,0,1,2,0,1,0, similarity is followed successively by S16=94, S26=90, S36=96, S45=86, S55=88, S64=91, S74=86, S83=95, S93=89, S103=89, S112=92, S122=91, S132=95, and make that corresponding degree of confidence is K16=1, K26=1, K36=1, K45=1, K55=1, K64=1, K74=1, K83=1, K93=1, K103=1, K112=1, K122=1, K132=1.
After finishing the identifying processing to the multiframe original image, unification is added up the recognition result of multiframe original image, at first compare the first character in each frame original image recognition result, the result all is all 1 mutually, the accumulative total degree of confidence is K1=6 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 1.
Compare successively each character in each frame original image recognition result, the results obtained are as follows:
Second character of iron and steel slab coding: the result all is all 5 mutually, and the accumulative total degree of confidence is K
2=6 (need not statistics accumulative total similarity), then the recognition result of iron and steel slab coding first character is 5;
Iron and steel slab the 3rd character of encoding: the result all is all 0 mutually, and the accumulative total degree of confidence is K3=6 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 0;
Iron and steel slab the 4th character of encoding: the result all is all 5 mutually, and the accumulative total degree of confidence is K4=5 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 5;
Iron and steel slab the 5th character of encoding: the result all is all 8 mutually, and the accumulative total degree of confidence is K5=5 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 8;
Iron and steel slab the 6th character of encoding: the result all is all 6 mutually, and the accumulative total degree of confidence is K6=4 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 6;
Iron and steel slab the 7th character of encoding: the result all is all 8 mutually, and the accumulative total degree of confidence is K7=4 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 8;
Iron and steel slab the 8th character of encoding: the result all is all 0 mutually, and the accumulative total degree of confidence is K8=3 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 8;
Iron and steel slab the 9th character of encoding: the result all is all 1 mutually, and the accumulative total degree of confidence is K9=3 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 1;
Iron and steel slab the tenth character of encoding: the result all is all 2 mutually, and the accumulative total degree of confidence is K10=3 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 2;
Iron and steel slab the 11 character of encoding: the result all is all 0 mutually, and the accumulative total degree of confidence is K11=2 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 0;
Iron and steel slab the 12 character of encoding: the result all is all 1 mutually, and the accumulative total degree of confidence is K12=2 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 1;
Iron and steel slab the 13 character of encoding: the result all is all 0 mutually, and the accumulative total degree of confidence is K13=2 (need not statistics accumulative total similarity), and then the recognition result of iron and steel slab coding first character is 0;
The monocase recognition result is combined into character string, namely 1505868012010, be the preliminary recognition result of iron and steel slab coding.Preliminary other result and custom coding rule are carried out verification, and the specific coding rule is as follows:
Rule 1: coding contains 13 characters;
Rule 2: first character is the mantissa in current time from left to right;
Rule 3: second character only may be 5 or 6 from left to right;
Rule 4: the 3rd to the 7th character span is 00001～99999 from left to right, such as 03380 in the example codes;
Rule 5: the 8th character span is 0～2 from left to right;
Rule 6: the 9th character span is 0～2 from left to right;
Rule 7: the tenth character span is 1～4 from left to right;
Rule 8: the 11 character perseverance is 0 from left to right;
Rule 9: the 12 character span is 1～9 from left to right, namely can not be 0;
Rule 10: the 13 character perseverance is 0 from left to right.
Prove that through verification preliminary recognition result satisfies above all custom codings rule.
Preliminary recognition result namely 1505868012010 is compared with the iron and steel slab coded message of obtaining from on-the-spot PLC10, and partial information is as follows in the iron and steel slab coded message that current on-the-spot PLC10 obtains:
(1)1606023024020；
(2)1606023023020；
(3)1505868012010；
(4)1505867011040；
Consistent with iron and steel slab coding in (3) by the comparison discovery, prove that this recognition result is accurate, then on system interface, export recognition result: 1505868012010.With this iron and steel slab coding, and the iron and steel slab begins to enter and leave the time of ccd video camera (1), as recognition result information, real-time storage is in local data base, and be Fig. 9 with the image that contains complete iron and steel slab coding that moving object detection and segmentation software module obtain, store in the local hard drive; With recognition result information, be sent in real time the operation department of hot rolling mill by the PORT COM on the industrial computer.
Claims (7)
1. an iron and steel slab coding automatic identifying method is characterized in that, obtains continuous multiframe original image at the transmission scene of iron and steel slab, since the first frame original image, successively each frame original image is identified the slab coding according to the following steps automatically:
Step 1: original image is carried out the image pre-service; Described pre-service comprises that gray processing is processed, mean filter is processed and difference processing; Described difference processing refers to filtered image and Background are made difference processing, obtains the difference gray-scale map;
Step 2: pretreated image is carried out binary conversion treatment, again the binary map of gained is carried out the slab code detection, generate testing result, in the original image of pre-treatment, contain slab coding if recognize according to testing result, then binary map is carried out slab coding site location, the projection process method is adopted in described slab code detection and coding site location;
Step 3: original image is carried out cutting according to the boundary coordinate that aforesaid coding site location obtains, obtain a plurality of monocase images, described monocase image refers to only contain a character-coded image; Each monocase image is carried out character recognition, thereby finish the slab code identification in the current frame original image;
Returning step 1 pair next frame original image processes;
Monocase image character identification in the step 3 may further comprise the steps:
Step a: the monocase image is carried out gray processing, obtain the monocase gray-scale map;
Step b: described monocase gray-scale map is carried out filtering process, obtain filtered monocase gray-scale map;
Step c: described filtered monocase gray-scale map is carried out binary conversion treatment, obtain the monocase binary map;
Steps d: described monocase binary map is carried out size normalization process, obtain normalization monocase binary map;
Step e: described normalization monocase binary map is pursued the pixel feature extraction, obtain eigenvectors matrix; Describedly refer to by the pixel feature extraction: normalization monocase binary map is scanned line by line, to the white pixel point in the normalization monocase binary map, getting eigenwert is 1, to the black pixel point in the normalization monocase binary map, getting eigenwert is 0, obtain at last an eigenvectors matrix, the dimension of eigenvectors matrix equals the total number of pixel in the normalization monocase binary map;
Step f: pursue the pixel template matches:
Described eigenvectors matrix and alphabet template are mated, obtain a plurality of similarities, similarity refers to that identical element number accounts for the ratio of total element number; With the matching result of the corresponding character of the Character mother plate of similarity maximum as current monocase image; The number of elements that comprises in number of elements in each Character mother plate and the eigenvectors matrix is identical.
2. iron and steel slab coding automatic identifying method according to claim 1 is characterized in that in the described step 2, the threshold value of described binary conversion treatment is avg+N, and the N value is the arbitrary value in 20～70; Avg is the mean value of gray-scale map entire image grey scale pixel value after the difference.
3. iron and steel slab according to claim 1 coding automatic identifying method, it is characterized in that, slab code detection process in the step 2 is: described binary map is carried out vertical projection, if there are intersection point in resulting vertical projection curve and preset level line, then judge and contain the slab coding in the present image, otherwise do not have the slab coding in the explanation present image; The preset level line the vertical projection Curves the ordinate of coordinate system be 15;
Described vertical projection refers to calculate the quantity of the white pixel point of each row in the binary map as the Y-axis coordinate figure of vertical projection curve, with the X-axis coordinate figure of the pixels across point sequence number in the binary map as the vertical projection curve, forms the vertical projection curve.
4. iron and steel slab coding automatic identifying method according to claim 1 is characterized in that, the process of the slab coding site location in the step 2 is:
(1) binary map is carried out horizontal projection, obtain the horizontal projection curve; Two intersection points of horizontal projection curve and preset level line are the up-and-down boundary point;
(2) vertical projection curve and preset level line intersect, and form many antinodes, and every a pair of adjacent intersection point i.e. the border, the left and right sides of a monocase of correspondence;
(3) based on the border, the left and right sides of described up-and-down boundary point and each monocase, binary map is carried out cutting, obtain a plurality of monocase images;
Described horizontal projection refers to calculate the quantity of the white pixel point of every delegation in the binary map as the Y-axis coordinate figure of horizontal projection curve, with the sequence number of the vertical pixel in the binary map X-axis coordinate figure as the horizontal projection curve, forms the horizontal projection curve;
Described preset level line, horizontal projection curve and vertical projection Curves the ordinate of coordinate system be 15.
5. iron and steel slab according to claim 1 coding automatic identifying method, it is characterized in that, among the described step f, the matching way of described eigenvectors matrix and single Character mother plate is: successively each element in the comparative feature vector matrix and the corresponding element in the Character mother plate, judge whether identical, and the number of statistics identical element, the number of identical element is accounted for the ratio of total element number as the similarity of current monocase image with this current Character mother plate;
Among the described step f, the manufacturing process of described Character mother plate is as follows: at first obtain the slab coded image, be encoded to one group to comprise all single slabs, choose the monocase image of one group of slab coding, through the processing of step a to step e, obtain one group and slab coded image characteristic of correspondence vector matrix, this stack features vector matrix is stored as file, be one group of Character mother plate corresponding with the slab coded image.
6. iron and steel slab coding automatic identifying method according to claim 1 is characterized in that described monocase gray-scale map carries out binary conversion treatment and adopts large Tianjin method;
Described size normalization is processed and is referred to binary map is converted to 50 * 60 Pixel Dimensions sized images;
Described Character mother plate is many groups;
The character that the character group that described Character mother plate is corresponding comprises is at least a in digital 0-9, English capitalization, English lower case, the Chinese character.
7. each described iron and steel slab coding automatic identifying method is characterized in that in the code identification process, whenever identify a character, then the degree of confidence that this character is corresponding increases by 1 according to claim 1-6, and the degree of confidence initial value is 0; After last coding of any iron and steel slab finished code identification, if the corresponding a plurality of identification characters of some codings are then got the highest identification character of degree of confidence as final recognition result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210091150 CN102663360B (en) | 2012-03-30 | 2012-03-30 | Method for automatic identifying steel slab coding and steel slab tracking system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201210091150 CN102663360B (en) | 2012-03-30 | 2012-03-30 | Method for automatic identifying steel slab coding and steel slab tracking system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102663360A CN102663360A (en) | 2012-09-12 |
| CN102663360B true CN102663360B (en) | 2013-10-23 |
Family
ID=46772844
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201210091150 Expired - Fee Related CN102663360B (en) | 2012-03-30 | 2012-03-30 | Method for automatic identifying steel slab coding and steel slab tracking system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102663360B (en) |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102929190A (en) * | 2012-10-17 | 2013-02-13 | 宁波钢铁有限公司 | System and method for calibrating number of cold and hot rolled finished product steel coil |
| CN104408452B (en) * | 2014-11-17 | 2019-01-15 | 深圳市比一比网络科技有限公司 | A kind of Latin character correcting inclination method and system based on rotation projection width |
| CN104751194A (en) * | 2015-04-27 | 2015-07-01 | 陈包容 | Processing method and processing device for financial expense reimbursement |
| CN106127205A (en) * | 2016-06-22 | 2016-11-16 | 山东鲁能智能技术有限公司 | A kind of recognition methods of the digital instrument image being applicable to indoor track machine people |
| CN106372218A (en) * | 2016-09-07 | 2017-02-01 | 上海东震冶金工程技术有限公司 | Casting blank encoding and recognizing method for use in ferrous metallurgy |
| CN107392205B (en) * | 2017-06-09 | 2020-07-07 | 广州视源电子科技股份有限公司 | Code value table generation method and device of remote controller |
| CN107451507A (en) * | 2017-08-03 | 2017-12-08 | 青岛海信电器股份有限公司 | A kind of two-dimensional code identification method being used in dynamic image and device |
| CN107452144A (en) * | 2017-08-17 | 2017-12-08 | 成都工业学院 | Automatic charging method and device |
| CN107688813A (en) * | 2017-09-24 | 2018-02-13 | 中国航空工业集团公司洛阳电光设备研究所 | A kind of character identifying method |
| CN110069794B (en) * | 2018-01-23 | 2023-08-15 | 宝山钢铁股份有限公司 | Symbol-based small square billet support tracking method |
| CN108055650B (en) * | 2018-01-31 | 2024-06-07 | 四川金互通科技股份有限公司 | Data processing apparatus and method |
| CN108427954B (en) * | 2018-03-19 | 2021-08-27 | 上海壹墨图文设计制作有限公司 | Label information acquisition and recognition system |
| CN109543770A (en) * | 2018-11-30 | 2019-03-29 | 合肥泰禾光电科技股份有限公司 | Dot character recognition methods and device |
| CN110361625B (en) * | 2019-07-23 | 2022-01-28 | 中南大学 | Method for diagnosing open-circuit fault of inverter and electronic equipment |
| CN110688997B (en) * | 2019-09-24 | 2023-04-18 | 北京猎户星空科技有限公司 | Image processing method and device |
| CN111368833B (en) * | 2020-03-06 | 2021-06-04 | 北京科技大学 | Full-automatic steel loading method for detecting slab number based on machine vision |
| CN118230327A (en) * | 2020-04-10 | 2024-06-21 | 支付宝实验室(新加坡)有限公司 | Machine-readable code identification method, device, electronic equipment and storage medium |
| CN112328234B (en) * | 2020-11-02 | 2023-12-08 | 广州博冠信息科技有限公司 | Image processing method and device |
| CN114004858B (en) * | 2021-11-19 | 2024-07-12 | 清华大学 | Method and device for identifying surface codes of aerial cables based on machine vision |
| CN114266908A (en) * | 2021-12-17 | 2022-04-01 | 福建顺景机械工业有限公司 | Character screening method based on image pixel value characteristics |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101615244A (en) * | 2008-06-26 | 2009-12-30 | 上海梅山钢铁股份有限公司 | Handwritten plate blank numbers automatic identifying method and recognition device |
| CN102170558A (en) * | 2010-12-30 | 2011-08-31 | 财团法人车辆研究测试中心 | Obstacle detection alarm system and method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005159662A (en) * | 2003-11-25 | 2005-06-16 | Ricoh Co Ltd | Image processing apparatus, image processing program and image processing method |
-
2012
- 2012-03-30 CN CN 201210091150 patent/CN102663360B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101615244A (en) * | 2008-06-26 | 2009-12-30 | 上海梅山钢铁股份有限公司 | Handwritten plate blank numbers automatic identifying method and recognition device |
| CN102170558A (en) * | 2010-12-30 | 2011-08-31 | 财团法人车辆研究测试中心 | Obstacle detection alarm system and method |
Non-Patent Citations (1)
| Title |
|---|
| JP特开2005-159662A 2005.06.16 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102663360A (en) | 2012-09-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102663360B (en) | Method for automatic identifying steel slab coding and steel slab tracking system | |
| US10311322B2 (en) | Character information recognition method based on image processing | |
| CN111539330B (en) | Transformer substation digital display instrument identification method based on double-SVM multi-classifier | |
| CN106682665B (en) | Seven-segment type digital display instrument number identification method based on computer vision | |
| CN113837991B (en) | Image processing method, device, equipment and storage medium | |
| CN109829458B (en) | Method for automatically generating log file for recording system operation behavior in real time | |
| CN102663380A (en) | Method for identifying character in steel slab coding image | |
| CN116721107B (en) | Intelligent monitoring system for cable production quality | |
| CN110096945B (en) | Indoor monitoring video key frame real-time extraction method based on machine learning | |
| CN113947563A (en) | Cable process quality dynamic defect detection method based on deep learning | |
| CN115880566A (en) | Intelligent marking system based on visual analysis | |
| CN113538585A (en) | High-precision multi-target intelligent identification, positioning and tracking method and system based on unmanned aerial vehicle | |
| CN116051539A (en) | Diagnosis method for heating fault of power transformation equipment | |
| CN117351499B (en) | Split-combined indication state identification method, system, computer equipment and medium | |
| WO2022222036A1 (en) | Method and apparatus for determining parking space | |
| CN116524410A (en) | Deep learning fusion scene target detection method based on Gaussian mixture model | |
| CN111832337A (en) | License plate recognition method and device | |
| CN111652055B (en) | Intelligent switch instrument identification method based on two-stage positioning | |
| CN202523078U (en) | Slab automatic identifying and tracking system based on coded image of steel slab | |
| CN112036393A (en) | Identification method based on shale gas field production single-pointer meter reading | |
| CN106845504B (en) | Method for identifying tile of yarn car | |
| CN112597916B (en) | Face image snapshot quality analysis method and system | |
| CN118038026B (en) | Substation secondary wiring credible quality inspection method based on rotation target detection | |
| CN209879574U (en) | Moving target detection system based on intelligent video analysis | |
| CN116630251A (en) | Deep learning-based anode copper plate defect detection method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20131023 Termination date: 20210330 |