CN103136544A - Image judging device - Google Patents
Image judging device Download PDFInfo
- Publication number
- CN103136544A CN103136544A CN2011104121139A CN201110412113A CN103136544A CN 103136544 A CN103136544 A CN 103136544A CN 2011104121139 A CN2011104121139 A CN 2011104121139A CN 201110412113 A CN201110412113 A CN 201110412113A CN 103136544 A CN103136544 A CN 103136544A
- Authority
- CN
- China
- Prior art keywords
- cell
- image
- ruling
- template image
- input picture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Character Input (AREA)
- Image Processing (AREA)
Abstract
The invention provides an image judging device which is high in judging precision. A characteristic quantity extraction part (11), an aligning part (13) and a judging part (14) are arranged in an image extraction device (100), wherein the characteristic quantity extraction part (11) extracts grid lines and table cells from an input image; the aligning part (13) enables the table cells and the grid lines in the template image to be aligned with the table cells and the grid lines in the input image based on intervals of the grid lines in a template image and intervals of the grid lines in an input image; the judging part (14) calculates the mutual overlapping ratio of the table cells in the input image and the table cells of the template image, wherein the table cells in the input image and the table cells of the template image are aligned by the aligning part (13). The judging part (14) calculates the concordance rate of the input image and the template image based on the calculated overlapping ratio. When the calculated concordance rate is higher than a regulated threshold value, the judgment that the input image is consistent with the template image is obtained.
Description
Technical field
The present invention relates to judge the whether image judgment means consistent with template image of input picture that account ticket etc. comprises ruling and cell.
Background technology
In the issued transaction of financial circles etc., process the account ticket of the item such as filling date, the amount of money.In recent years, in order to realize Labor-saving, the exploitation that automatically is extracted in the technology of the item of filling in the account ticket by signal conditioning package was carried out.In addition, as the last stage that is extracted in the item of filling in the account ticket, the exploitation of the technology of judgement account ticket kind is also carried out.
For example, in patent documentation 1, record and be categorized as multiple account ticket any one the account ticket recognition device in originally with filling in complete account ticket.The account ticket recognition device of patent documentation 1 record judges based on the ruling that extracts and the ruling that extracts in advance whether input picture is consistent with template image from read in advance the template image that the account ticket obtains originally by scanner from read the input picture of filling in complete account ticket and obtaining by scanner.In more detail, in the frame line that the account ticket recognition device of patent documentation 1 record makes minimum rectangle in the input picture that extracts based on ruling in input picture and the template image that extracts in advance based on ruling in template image, frame line 1 pair of 1 ground of minimum rectangle carries out corresponding, calculate the distinctiveness ratio between the frame line of the minimum rectangle of correspondence, when distinctiveness ratio hour, be judged as input picture consistent with template image.
Patent documentation 1: TOHKEMY 2004-334913 communique
Summary of the invention
When extracting ruling from template image, the user confirms that whether ruling can be correctly identified, and can revise ruling as required when ruling lacks usually.Therefore, the ruling of the template image of storing in account ticket recognition device is complete, and the frame line of the minimum rectangle of extracting based on ruling is also complete.On the other hand, owing to input picture not being revised etc., therefore ought fill in defectiveness in complete account ticket, when perhaps the performance of scanner reduced, it is imperfect that ruling might become.
When the ruling in input picture is imperfect, the shape of the minimum rectangle of extraction, the size and template image differ greatly, thus in input picture in minimum rectangle and template image the corresponding relation of minimum rectangle can change.At that rate, in patent documentation 1 in the account ticket recognition device of record, just can not correctly calculate the distinctiveness ratio between the frame line of minimum rectangle of correspondence, thereby can not correctly judge.
The present invention proposes for solving above-mentioned problem, and its purpose is to provide the judgement precision high image judgment means.
Image judgment means of the present invention is characterised in that to possess: Characteristic Extraction section, and it extracts ruling and cell from input picture; Alignment part, its based on extract in advance and the template image of storage in the interval of ruling in interval and the described input picture that extracts by described Characteristic Extraction section of ruling, make extract in advance and the template image of storage in cell and snap to grid in cell and ruling and the described input picture that extracts by described Characteristic Extraction section; And judging part, it calculates by the cell Duplication each other between cell in cell and described template image in the described input picture after described alignment part alignment, calculate the concordance rate of described input picture and described template image based on the Duplication of calculating, when the concordance rate of calculating is higher than the threshold value of stipulating, be judged as described input picture consistent with described template image.
in addition, the invention is characterized in, the number of the cell that described judging part ought extract from described input picture and the number of the cell in described template image are not simultaneously, for the unit lattice in the few side's of the number of cell image, calculate the ratio of the total area of the cell in the few side's of the number of the area of overlapping part of any 1 cell in the many sides' of the number of this cell and cell image and cell image as described Duplication, the low Duplication of boundary value of ratio regulation in the Duplication that rejecting is calculated, calculate the mean value of each Duplication, with the mean value of calculating as described concordance rate.
in addition, the invention is characterized in, a plurality of ratios at the interval of ruling in the interval of ruling and described input picture in the described template image of described alignment part calculating, with a plurality of ratios of calculating as a plurality of alternative multiplying powers, each alternative multiplying power for a plurality of alternative multiplying powers, calculating make ruling in described template image any 1 with described input picture in any 1 consistent a plurality of alternative parallel of ruling, for each alternative multiplying power and each alternative parallel, evaluation is according to the zoom in or out ruling of any one party in described template image or described input picture of described alternative multiplying power, and come the consistent degree of the ruling in the ruling of any the opposing party in ruling and described template image or described input picture after the ruling of any one party in the described template image of parallel or described input picture according to described alternative parallel, alternative multiplying power when the highest according to described consistent degree and alternative parallel zoom in or out ruling and the cell of any one party in described template image or described input picture, come to align with any the opposing party's ruling and cell in described template image or described input picture by parallel.
In addition, the invention is characterized in, possesses the identification part, when the quantity of cell in described input picture is lacked than the reference value of regulation, this identification part does not make described alignment part align and described judging part is judged, when the quantity of cell in described input picture manyed than the reference value of described regulation, this identification part was alignd described alignment part and described judging part is judged.
According to the present invention, the cell in making input picture judges according to cell Duplication each other with after cell in template image aligns, and therefore do not need to consider the corresponding relation between cell, can carry out the good judgement of precision.
In addition, according to the present invention, the ratio of the total area of the cell in the side's that the area of the repeating part of any 1 cell in the side's that the cell in the few image of the number of cell and the number of cell are many image and the number of cell are few image is calculated as the Duplication of cell, to the rejecting of each Duplication of each cell in the few side's of the number of cell image the mean value of the Duplication lower than the boundary value of stipulating be calculated as concordance rate.For example, when the ruling in input picture lacked, the number of the cell in input picture can be lacked than the number of the cell in template image, correspondingly, had occurred in input picture than the large cell of cell area in template image.The Duplication of any 1 cell in the large cell of this area and template image can be lower than the boundary value of regulation, therefore do not consider in the calculating of concordance rate.That is to say, the impact of the appearance of the cell that the minimizing of the number of the cell that the disappearance that image judgment means of the present invention can be got rid of the ruling in input picture causes and area are large judges.
In addition, according to the present invention, the interval ratio each other of ruling is made as a plurality of alternative multiplying powers, each alternative multiplying power is calculated a plurality of alternative parallels, and the alternative multiplying power the highest according to the consistent degree of ruling and alternative parallel make in input picture in ruling and cell and template image ruling and cell align.Therefore, can precision align well.
In addition, according to the present invention, utilize the identification part, the quantity of cell than the reference value of regulation when lack, is not carried out based on the alignment of described alignment part and based on the judgement of described judging part, so can be reduced calculated amount in being judged as input picture.
Description of drawings
Fig. 1 is the block diagram of the formation of presentation video extraction element 100 functionally.
Fig. 2 means by scanner etc. and reads the account ticket of not filling in word etc. and the figure of the image that obtains.
Fig. 3 means the figure that is read the image of filling in complete account ticket and obtaining by scanner etc.
Fig. 4 is the process flow diagram of the processing of presentation video extraction element 100 summarily.
Fig. 5 means the ruling that extracts and the figure of cell from template image shown in Figure 2.
Fig. 6 means the ruling that extracts and the figure of cell from input image of bill shown in Figure 3.
Fig. 7 is the figure when ruling shown in Figure 5 and cell are alignd with ruling shown in Figure 6 and cell.
The explanation of Reference numeral
10... image judging part; 11... Characteristic Extraction section; 12... identification part; 13... alignment part; 14... judging part; 20... image extraction unit; 30... storage part; 100... image acquiring apparatus.
Embodiment
Below, the image acquiring apparatus 100 as embodiments of the present invention is described.Fig. 1 is the block diagram of the formation of presentation video extraction element 100 functionally.Image acquiring apparatus 100 is realized by known control algorithm circuit in the past and memory circuit that MFP (Multi Function Printer) etc. possesses, is comprised image judging part 10, image extraction unit 20 and storage part 30 on function.Image judging part 10 comprises Characteristic Extraction section 11, identification part 12, alignment part 13 and judging part 14.
Summarily say, image acquiring apparatus 100 is for the device that extracts word that the account ticket fills in etc.Fig. 2 represents to read the account ticket of not filling in word etc. (below, be called " the account ticket originally ") and the image that obtains (below, be called " template image ") by scanner etc.The account ticket is for example the recording paper of rectangular shape as shown in Figure 2, is decorated with many ruled line, at 2 ruled line K by longitudinal extension
Y2 ruled line K with horizontal expansion
XThe zone of the rectangular shape that marks off (below, be called " cell ") C
EIn, fill in word etc.Fig. 3 represents to read in by scanner etc. the account ticket of having filled in word etc. in the cell of the blank column of account ticket in originally, has namely filled in complete account ticket and the image that obtains (below, be called " input image of bill ").
Fig. 4 is the process flow diagram of the processing of presentation video extraction element 100 summarily.When image was transfused to, Characteristic Extraction section 11 extracted ruling and the cell (step S1) as this Characteristic of Image amount from the image of input.
Next, in step S2, identification part 12 is based on the ruling that extracts and cell, whether the image that judgement is transfused to is template image or input image of bill, when the image that is transfused to is template image or input image of bill, enter step S3, when the image that is transfused to is not template image or input image of bill, enter step S8.For example, the quantity of the cell of identification part 12 in the image that is transfused to is than the reference value (for example 10) of regulation when lack, and to be judged as the image that is transfused to be not template image or input image of bill.In step S8, it is not the account ticket that image acquiring apparatus 100 is notified the image that the user is transfused to, end process.
In step S3, image acquiring apparatus 100 is urged any in user's input template images or input image of bill, when input be template image the time, enter step S4, when input is the input image of bill time, enter step S5.
In step S4, the ruling of storage part 30 storing template images and the information of cell.At this moment, ruling and cell that image acquiring apparatus 100 also can the indicating template image allow the user confirm, whether have mistake in ruling and cell.In addition, image acquiring apparatus 100 also can constitute when having mistake in ruling and cell, and the user can revise.In addition, image acquiring apparatus 100 also can constitute together with information with the ruling of template image and cell, and storage is by title or kind, the region-of-interest etc. of the template image of user's input.Here, region-of-interest refers to when the input image of bill with the template image identical type is transfused to, and becomes the cell that image acquiring apparatus 100 extracts the object of words etc.
In step S5, alignment part 13 is based on the interval of ruling in the interval of the ruling in the template image of storage in storage part 30 and the input image of bill that extracts by Characteristic Extraction section 11, makes in this template image cell and snap to grid in cell and ruling and this input image of bill.Here the alignment of indication is to instigate in template image in cell and ruling and input picture cell and ruling to align at formed objects and position.Alignment is carried out at longitudinal direction and the transverse direction of image respectively.
Particularly, alignment part 13 ratio at the interval of ruling in the interval of ruling and input image of bill in the calculation template image at first.Therefore usually, in the account ticket, the interval of ruling exists a plurality ofly, will calculate a plurality of ratios.Then, a plurality of ratios that will calculate of alignment part 13 are as alternative multiplying power.If template image based on the account ticket of identical type, utilizes any 1 in a plurality of alternative multiplying powers to zoom in or out and get final product with the input image of bill, template image with input image of bill and be aligned to formed objects.
Next, alignment part 13 is when each the alternative multiplying power to a plurality of alternative multiplying powers, when having carried out zooming in or out with alternative multiplying power, set a plurality of such as ruling in template image any 1 with input picture in any 1 consistent so alternative parallel (alternative displacement) of ruling.Then, to each alternative multiplying power and each alternative parallel, evaluation zooms in or out according to the ruling of alternative multiplying power any one party in image of bill to template image or input, and comes ruling and the template image after the ruling of any one party in parallel template image or input image of bill according to alternative parallel or input the consistent degree of the ruling between the ruling of any the opposing party in image of bill.Consistent degree is high, estimates and is set as height.That is to say, if template image and the account ticket of input image of bill based on identical type, alternative multiplying power when using consistent degree the highest and the group of alternative parallel, the ruling of template image or any one party of input in image of bill of zooming in or out, can align in the ruling after parallel and template image or size and the position of inputting any the opposing party in image of bill.Like this, alignment part 13 makes cell in template image and cell and the snap to grid in ruling and input image of bill.
In step S6, judging part 14 calculates by the cell Duplication each other in cell in cell and template image in the input image of bill after alignment part 13 alignment.Then, judging part 14 calculates the concordance rate of input image of bill and template image based on the Duplication of calculating, and when the concordance rate of calculating is higher than the threshold value of regulation (for example 95%), is judged as that to input image of bill consistent with template image, enters step S7.To all template images result of aliging and judge of storage in storage part 30, when being judged as the input image of bill and template image is all inconsistent arbitrarily, enter step S9.In step S9, it is non-classified account ticket that image acquiring apparatus 100 is notified the image that the user inputs, end process.
In step S6, in the number of cell and template image in the input image of bill number of cell not simultaneously, judging part 14 calculates Duplication and concordance rate according to following mode.Below, suppose that the number of cell in the input image of bill is lacked than the number of cell in template image.Under these circumstances, the unit lattice in 14 pairs of input image of bill of judging part, the ratio of the area of the part that any 1 cell in this cell and template image is overlapping and the total area of cell is calculated as Duplication.That is to say Duplication=(total area of the area/cell of overlapping part) * 100[%].Unit lattice in 14 pairs of judging parts input image of bill calculate Duplication, reject the low Duplication of boundary value (for example 75%) of the ratio regulation in the Duplication of calculating, and calculate the mean value of each Duplication.The mean value that judging part 14 will be calculated like this is as concordance rate.
In step S7, in the cell of image extraction unit 20 from be judged as the input image of bill consistent with template image, be stored in the region-of-interest of storage part 30 and extract image.Then, image extraction unit 20 is utilized OCR (Optical Character Recognition) generating character etc. and it is stored in storage part 30 end process based on the image that extracts.
According to such image acquiring apparatus 100, after making the cell of input in image of bill and cell in template image aliging, therefore unanimously judge according to cell Duplication each other, do not need to consider the corresponding relation between cell, can carry out the good judgement of precision.
In addition, in step S6, in the side's that in the side's that the cell of the image that the number of cell is few and the number of cell are many image, the number of the area of the repeating part of any 1 cell and cell is few image, the ratio of the total area of cell is calculated as the Duplication of cell, rejecting is calculated as concordance rate about the mean value each Duplication, lower than the boundary value of regulation Duplication of each cell in the few side's of the number of cell image, therefore can carry out the good judgement of precision.For example, when in the input image of bill, ruling lacked, in the input image of bill, the number of cell was lacked than the number of cell in template image, correspondingly, in the input image of bill, had occurred than the large cell of cell area in template image.In the large cell of this area and template image, the Duplication of any 1 cell is lower than the boundary value of regulation, does not therefore consider in the calculating of concordance rate.That is to say, according to image acquiring apparatus 100, the impact that can get rid of the appearance of the large cell of the minimizing of number of the caused cell of disappearance of ruling in the input image of bill and area judges that its result can be carried out the good judgement of precision.
In addition, in step S5, with the interval ratio each other of ruling as a plurality of alternative multiplying powers, to each alternative multiplying power, set a plurality of alternative parallels, alternative multiplying power and the alternative parallel the highest according to the consistent degree of ruling make in the input image of bill in ruling and cell and template image ruling and cell align.Therefore, image acquiring apparatus 100 can precision align well, and its result can be carried out the good judgement of precision.
In addition, in step S2, when the quantity of cell in the image of input was lacked than the reference value of regulation, the image that is judged as input was not template image or input image of bill, in this case, does not carry out the processing of step S3～S7, S9.Therefore, in this case, do not carry out the alignment of alignment part 13 and the judgement of judging part 14, therefore can reduce the calculated amount of image acquiring apparatus 100.
Below, the concrete action of Characteristic Extraction section 11 is described.Characteristic Extraction section 11 extracts ruling and cell shown in Figure 5 from template image shown in Figure 2, perhaps extract ruling and cell shown in Figure 6 from input image of bill shown in Figure 3.
Particularly, at first Characteristic Extraction section 11 carries out slant correction to the image of input and processes.Slant correction process be in the situation that when reading the account ticket by scanner etc. the account ticket tilt a little, in order to proofread and correct this inclinations, make image to the processing about several times with the direction rotation of the opposite direction that tilts.
Next, the image after Characteristic Extraction section 11 processes slant correction is transformed to the image of GTG (for example, the GTG of 255 gray scales).Then, the top left pixel of the image after the conversion is take right as main scanning direction (directions X) successively, lower direction is sub scanning direction (Y-direction), individual element ground scans to carry out the extraction of pixel, the pixel of carrying out extracting is as black pixel, with 2 values as white pixel in addition.The pixel that extracts is to have the pixel that exceeds the above gray-scale value of the value (for example 50) of regulation than the gray-scale value of all around any one pixel.In addition, the pixel of the gray-scale value with same degree (for example, ± 10 degree) adjacent with the pixel that extracts previously also is extracted.Like this by carrying out 2 values, as Fig. 2 and Fig. 5 and Fig. 3 and shown in Figure 6, the cell of ash can be transformed to whitely, and be transformed to the ruling of ash black.On the other hand, if when merely utilizing 1 threshold value to carry out 2 value to the image of inputting, the cell of ash all can be transformed to black or grey ruling can be transformed to white.
Next, Characteristic Extraction section 11 is by known line segments extraction method, and for example Hough conversion is extracted in the parallel line segment of directions X or Y-direction, the XY coordinate figure of the two-end-point of storage line segment.Then, the value of 11 pairs of regulations of Characteristic Extraction section (for example 5) is with the parallel sets of line segments at interior interval, aligned position.Particularly, if the sets of line segments parallel with X-axis calculated the arithmetic mean of Y coordinate figure, the Y coordinate figure of each line segment that this value is comprised as sets of line segments.for example, when the XY coordinate figure of the two-end-point of line segment A is (1, 2), (10, 2), the XY coordinate figure of the two-end-point of line segment B is (20, 2.1), (25, 2.1), the XY coordinate figure of the two-end-point of line segment C is (11, 1.6), (15, 1.6) time, the arithmetic mean (2+2.1+1.6) of the Y coordinate figure of calculating line segment A～C/3=1.9, the XY coordinate figure of the two-end-point of line segment A is replaced into (1, 1.9), (10, 1.9), the XY coordinate figure of the two-end-point of line segment B is replaced into (20, 1.9), (25, 1.9), the XY coordinate figure of the two-end-point of line segment C is replaced into (11, 1.9), (15, 1.9).11 pairs of sets of line segments parallel with Y-axis of Characteristic Extraction section are also carried out same processing.
Next, following adjacent line segment each other for Characteristic Extraction section 11 value that connects to stipulate (for example 3).For example, the XY coordinate figure of the two-end-point of line segment A is (1,1.9), (10,1.9), the XY coordinate figure of the two-end-point of line segment B is (20,1.9), (25,1.9), the XY coordinate figure of the two-end-point of line segment C is (11,1.9), (15,1.9), so line segment A is adjacent with line segment C, line segment B is adjacent with line segment C in addition.Under these circumstances, the Characteristic Extraction 11 connecting line segment A of section and line segment C, and connecting line segment B and line segment C are linked to be 1 line segment.The XY coordinate figure of the two-end-point of this 1 line segment becomes (1,1.9), (25,1.9).
Next, whether Characteristic Extraction section 11 short line segment (for example, the length of line segment is less than 5) of confirming regulation intersects with vertical line segment.And when this short line segment does not intersect with vertical line segment, the length of perhaps intersect 1 or a plurality of vertical line segment is deleted this short line segment during less than the length of regulation (for example 10).Next, Characteristic Extraction section 11 confirms that whether remaining line segments intersect with vertical line segment, delete this line segment when not intersecting.Through above processing and remaining line segment is regarded as ruling.
Next, Characteristic Extraction section 11 according to the concern intersection point the intersection point of paying close attention to line segment from the order of left beginning, carries out the processing of following (a)～(d) cell is extracted when the Y coordinate figure equates from the top of image in order.
(a) be confirmed whether to exist and have the Y coordinate figure identical with paying close attention to intersection point and than paying close attention to intersection point intersection point (the 1st intersection point) on the right side.When existing, carry out the processing of (b), when not existing, shift to the next intersection point of paying close attention to.
(b) be confirmed whether to exist and have the X coordinate figure identical with paying close attention to intersection point and than paying close attention to the intersection point intersection point of side (the 2nd intersection point) on the lower.When existing, carry out the processing of (c), when not existing, shift to the next intersection point of paying close attention to.
(c) be confirmed whether to exist the intersection point (the 3rd intersection point) that has the X coordinate figure identical with the 1st intersection point and have the Y coordinate figure identical with the 2nd intersection point.When existing, carry out the processing of (d), when not existing, shift to the next intersection point of paying close attention to.
(d) the XY coordinate figure that will pay close attention to intersection point and the 1st intersection point～the 3rd intersection point is made as 1 group, stores accordingly with 1 cell, shifts to next and pays close attention to intersection point.
Next, the concrete action of identification part 12 described.When satisfying the condition of for example following (e)～(g), identification part 12 is judged as the image that is transfused to be template image or to input image of bill.
(e) number of cell is more than the quantity (for example 10) of regulation.
(f) the total area of the cell of the area of cell/all for the cell of the scope of regulation (for example 1%～5%) more than the quantity of regulation (for example 5).
(g) area of the integral image of the area of cell/be transfused to for the cell of the scope (for example, 0.2%～2%) of regulation more than the quantity of regulation (for example 5).Wherein, also can utilize " area that comprises the minimum rectangle of whole cells " to replace " all total areas of rectangle ".
Next, the concrete action for alignment part 13 describes.Alignment part 13 makes ruling shown in Figure 5 and cell and ruling and cell shown in Figure 6 be aligned to appearance shown in Figure 7.
Particularly, at first, 13 pairs of input image of bill of alignment part and template image calculate respectively the interval D Y between the line segment (ruling) parallel with X-axis
E, DY
TFor example, in the situation that the quantity of the line segment parallel with X-axis that exists in the input image of bill is 5, as the Y of each line segment coordinate figure Y
E1～5From a little side be in order 1,4,8,13, during 18}, be DY
E=3,4,5,5}.In addition, for example in the situation that the quantity of the line segment parallel with X-axis that exists in template image is 6, as the Y of each line segment coordinate figure Y
T1～6From a little side be in order 21,33,39,47,57, during 67}, be DY
T=12,6,8,10,10}.
Next, alignment part 13 is at DY
EEach key element and DY
TEach key element between calculate a plurality of ratios, calculate the ratio group take each ratio as key element.This calculates with DY
E, DY
TIn, a side that key element quantity is few is denominator, take many sides as molecule.In other words, at the zoom in or out few side's of the quantity of the ruling parallel with X-axis image of Y-direction, make the many sides' of the quantity of itself and ruling image alignment.Below, will be with DY
EKey element be denominator, with DY
TKey element be that the group that the ratio of molecule consists of is expressed as PY
T/EIn addition, when calculating ratio, the ratio of identical value might occur a plurality of, in this situation, does not repeat PY
T/EKey element.In addition, work as PY
T/EKey element too much, when calculated amount is huge, also can be from many sides of the appearance quantity of the ratio of identical value+in order only with the ratio of the number (for example 10) of regulation as PY
T/EKey element.
For example, work as DY
E=3,4,5,5}, DY
T=12,6,8,10, during 10}, the ratio group PY that is consisted of by a plurality of ratios
T/EBe PY
T/E=4,2,8/3,10/3,3,3/2,5/2,12/5,6/5,8/5}.Here, as PY
T/E4 expressions of key element for will be as DY
EKey element 3 with as DY
T12 alignment of key element, need to amplify 12/3=4[doubly].Similarly, as PY
T/E2 expressions of key element for will be as DY
EKey element 3 with as DY
T6 alignment of key element, need to amplify 6/3=2[doubly].Alignment part 13 can not be differentiated DY
TAny one key element and DY
EAny one key element whether corresponding, therefore suppose that as above all corresponding relations calculate PY
T/E
Next, alignment part 13 is with PY
T/EEach key element PY
T/E (k)(this value is alternative multiplying power) multiply by the Y as the Y coordinate figure in the few side's of the quantity of ruling image
E1～5, calculate the k ordered series of numbers EY as the ordered series of numbers on Y-axis
(k)For example, in upper example, as hypothesis PY
T/E (1)=4 o'clock, the 1st ordered series of numbers EY
(1)=4,16,32,52,72}.In addition, when using PY
T/E (2)=2 o'clock, the 2nd ordered series of numbers EY
(2)=2,8,16,26,36}.
Next, alignment part 13 makes k ordered series of numbers parallel on Y-axis, so that n key element E of k ordered series of numbers
nWith the Y coordinate figure Y in the many side's of the quantity of ruling image
T1～6M key element T
mUnanimously.For example, make the 1st ordered series of numbers parallel+31 (this value is alternative parallel) on Y-axis, so that the 1st ordered series of numbers EY
(1)=4,16,32,52, the 2nd key element of 72} 16 and Y coordinate figure Y
T1～6=21,33,39,47,57, the 4th key element 47 of 67} be consistent.Ordered series of numbers EY after parallel in this case
(k) nmBecome EY
(1) 24=35,47,63,83,103}.
Then, the ordered series of numbers EY after alignment part 13 these parallels of confirmation
(1) 24With Y coordinate figure Y
T1～ 6Other consistent point.Particularly, confirm at 47 the Y coordinate figure Y that has rejected as consistent key element
T1～6Each key element T
m-Min (DY
T) * 0.1～T
m+ Min (DY
TIn the scope of) * 0.1, whether there is EY
(1) 24Any one key element (remove as consistent key element 47), to consistent T
mEvery 1 of number be evaluated as 1 point.In addition, be provided with T
m-Min (DY
T) * 0.1～T
m+ Min (DY
TThe such scope in) * 0.1 is because also contain the complete inconsistent situation of each key element.
In upper example, EY
(1) 24The key element except 47 and Y coordinate figure Y
T1～6The key element except 47 all inconsistently (be not included in Y
T1～6Near scope in), therefore to EY
(1) 24Evaluation of estimate be 0 point.In addition, for example at EY
(2) 12Situation under, alternative parallel is+31, EY
(2) 12=33,39,47,57,67}, therefore with Y coordinate figure Y
T1～6In EY
(2) 12Consistent key element (remove as consistent key element 33) is 39,47,57,67 these 4 key elements, therefore to EY
(2) 12Evaluation of estimate be 4 points.Like this, 13 couples of all EY of alignment part
(k) nmCalculate respectively evaluation of estimate.
Next, alignment part 13 is selected the EY the highest with evaluation of estimate
(k) nmCorresponding alternative parallel and alternative multiplying power are as making template image and inputting image of bill in Y-direction align alternative parallel and the alternative multiplying power of use.In addition, at the highest EY of evaluation of estimate
(k) nmExist more than 2, thereby in 2 groups of above situations of the group of alternative parallel and alternative multiplying power existence, selecting to be regarded as each consistent key element is nearest one group.
At last, the side's that the quantity of 13 pairs of rulings of alignment part is few image (being the input image of bill in upper example), zoom in or out the ruling parallel with X-axis each other at the interval of Y-direction and the length of the ruling parallel with Y-axis according to the alternative multiplying power of selecting, and then alternatively come parallel according to what select, make each ruling at the Y-direction parallel.Then, 13 pairs of line segments parallel with Y-axis of alignment part (ruling) also carry out above processing, at the both direction of directions X and Y-direction, the input image of bill are alignd with template image.
Next, the concrete action of judging part 14 described.Judging part 14 calculates in the input image of bill that has been aligned as shown in Figure 7 the Duplication of cell in cell and template image.As described above, when the number of cell in the input image of bill than template image in the number of cell when lacking, each cell of cell in 14 pairs of judging parts input image of bill, the ratio of the area of any 1 cell lap in this cell and template image and the cell total area of calculating is as Duplication.Adopting any 1 cell is because of a plurality of cells situation overlapping with the cell in the input image of bill that may exist in template image.When a plurality of when overlapping, will be in template image in each cell the maximal value in overlapping part area as the area of the lap that uses in the calculating of Duplication.
In addition, in the situation that in the input image of bill, the number of cell equates with the number of cell in template image, judging part 14 also can calculate on the contrary this cell with upper example and input the area of the lap of any 1 cell in image of bill and template image in the ratio of the total area of cell as Duplication.
Next, judging part 14 calculates concordance rate based on Duplication.As described above, when the number of cell in the input image of bill than template image in the number of cell when lacking, judging part 14 reject in each Duplication of calculating, than the low Duplication of the boundary value of regulation (for example 75%), calculate the mean value of each Duplication as concordance rate.
In addition, mean value can be also arithmetic mean, can be also the weighted mean value of area.In addition, judging part 14 in the situation that in the input image of bill number of cell equate with the number of cell in template image, whether do not carry out Duplication than the low judgement of boundary value of stipulating, to all Duplication calculating mean values.This is because in this case, can think not have disappearance in ruling in the input image of bill.But, in the situation that in the input image of bill, the number of cell equates with the number of cell in template image, in the time of also can work as Duplication and compare the boundary value of stipulating and hang down, reject this Duplication.
When so calculating concordance rate, judging part 14 judges this concordance rate whether than the threshold value of regulation (for example 95%) height, and when being judged as when high, judgement input image of bill is consistent with template image.
Above such image acquiring apparatus 100 can be realized as the image extraction procedure of the program that this signal conditioning package is played a role as image acquiring apparatus 100 by storage in known signal conditioning package in the past.This image extraction procedure can be stored in the storage medium of known embodied on computer readable in the past.
By above-mentioned image extraction procedure being stored in storage medium, can carrying the storage medium that the program code (carrying out format program, intermediate code program, source program etc.) of having stored the various processing of carrying out for carries out image extraction element 100 is provided freely.
In addition, storage medium can be the storer that microcomputer is processed use, the such program medium of ROM (Read Only Memory) for example can be perhaps the program medium that reads by being inserted into program reading device that computing machine possesses as external memory.
The program of storing in storage medium can be the mode that the microprocessor access storage medium is carried out, it can be perhaps microprocessor read routine code from storage medium, the program code of reading is downloaded in the program storage area of microcomputer, thus the mode that this program is performed.The program of this download use is pre-stored within body apparatus.
here, the said procedure medium is the storage medium that can consist of discretely with body, it can be also tape, the tapes such as magnetic tape cassette system, floppy disk, the disks such as hard disk, CD-ROM (Compact Disc-Read Only Memory), MO (Magneto Optical disc), MD (Mini Disc), the dish system of the CDs such as DVD (Digital Versatile Disc), IC (Integrated Circuit) blocks (comprising memory card), the card such as light-card system, perhaps MASK ROM (mask ROM), EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), FLASH ROM etc. has comprised the medium of the code of save routine regularly of semiconductor memory.
In addition, consist of if computing machine is the system that can connect the communication network that comprises the Internet, storage medium can be also according to from the mode of the downloaded program code medium of save routine code fluidly.In addition, in the situation that so from the downloaded program, the program of this download use is stored in advance in body apparatus or from other storage medium and installs.In addition, above-mentioned image extraction procedure also can be realized by the mode of loading the computer data signal of said procedure code in carrier wave that realizes with electric transmission.
Claims (4)
1. image judgment means is characterized in that possessing:
Characteristic Extraction section, it extracts ruling and cell from input picture;
Alignment part, its based on extract in advance and the template image of storage in the interval of ruling in the interval of ruling and the described input picture that extracts by described Characteristic Extraction section, make extract in advance and the template image of storage in cell and snap to grid in cell and ruling and the described input picture that extracts by described Characteristic Extraction section; With
Judging part, it calculates by the cell Duplication each other between cell in cell and described template image in the described input picture after described alignment part alignment, calculate the concordance rate of described input picture and described template image based on the Duplication of calculating, when the concordance rate of calculating is higher than the threshold value of stipulating, be judged as described input picture consistent with described template image.
2. image judgment means according to claim 1, is characterized in that,
Described judging part constitutes:
When the number of the number of the cell that extracts from described input picture and the cell in described template image not simultaneously,
For each cell in the few side's of the number of cell image, the ratio of area of overlapping area and this cell of calculating is as described Duplication,
The low Duplication of boundary value of ratio in each Duplication that rejecting is calculated regulation is calculated the mean value of each Duplication, with the mean value of calculating as described concordance rate,
Wherein, described overlapping area is the area of equitant part between any 1 cell in the many sides' of the number of cell in the few side's of the number of cell image and cell image.
3. image judgment means according to claim 1 and 2, is characterized in that,
A plurality of ratios at the interval of ruling in the interval of ruling and described input picture in the described template image of described alignment part calculating, with a plurality of ratios of calculating as a plurality of alternative multiplying powers, each alternative multiplying power for a plurality of alternative multiplying powers, calculating make ruling in described template image any 1 with described input picture in any 1 consistent a plurality of alternative parallel of ruling
For each alternative multiplying power and each alternative parallel, evaluation according to described alternative multiplying power zoom in or out any one party in described template image or described input picture ruling and come the ruling of any one party in the described template image of parallel or described input picture according to described alternative parallel after ruling and described template image or described input picture in any the opposing party's ruling in the consistent degree of ruling
Alternative multiplying power when the highest according to described consistent degree and alternative parallel zoom in or out ruling and the cell of any one party in described template image or described input picture come to align with any the opposing party's ruling and cell in described template image or described input picture by parallel.
4. the described image judgment means of according to claim 1～3 any one, is characterized in that,
Possesses the identification part, when the quantity of cell in described input picture is lacked than the reference value of regulation, this identification part does not make described alignment part align and described judging part is judged, when the quantity of cell in described input picture manyed than the reference value of described regulation, this identification part was alignd described alignment part and described judging part is judged.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011104121139A CN103136544A (en) | 2011-11-30 | 2011-11-30 | Image judging device |
| JP2012249290A JP5619111B2 (en) | 2011-11-30 | 2012-11-13 | Image determination apparatus and program |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011104121139A CN103136544A (en) | 2011-11-30 | 2011-11-30 | Image judging device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103136544A true CN103136544A (en) | 2013-06-05 |
Family
ID=48496352
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011104121139A Pending CN103136544A (en) | 2011-11-30 | 2011-11-30 | Image judging device |
Country Status (2)
| Country | Link |
|---|---|
| JP (1) | JP5619111B2 (en) |
| CN (1) | CN103136544A (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6467787B2 (en) * | 2014-05-27 | 2019-02-13 | 株式会社リコー | Image processing system, imaging apparatus, image processing method, and program |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0973516A (en) * | 1995-09-04 | 1997-03-18 | Nippon Telegr & Teleph Corp <Ntt> | Slip type identification method |
| US20010031086A1 (en) * | 1996-12-25 | 2001-10-18 | Hitachi, Ltd. | Image processor, image processing apparatus, and image processing method |
| CN1460961A (en) * | 2003-06-27 | 2003-12-10 | 杭州信雅达系统工程股份有限公司 | Bill image face identification method |
| CN102129547A (en) * | 2010-01-20 | 2011-07-20 | 富士施乐株式会社 | Form identification apparatus, form processing system and form identification method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0728935A (en) * | 1993-07-07 | 1995-01-31 | Oki Electric Ind Co Ltd | Document image processor |
| JP3353215B2 (en) * | 1995-05-15 | 2002-12-03 | 日本電信電話株式会社 | Form format identification method and apparatus |
| JPH1027208A (en) * | 1996-07-10 | 1998-01-27 | N T T Data Tsushin Kk | Method and device for collating document |
-
2011
- 2011-11-30 CN CN2011104121139A patent/CN103136544A/en active Pending
-
2012
- 2012-11-13 JP JP2012249290A patent/JP5619111B2/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0973516A (en) * | 1995-09-04 | 1997-03-18 | Nippon Telegr & Teleph Corp <Ntt> | Slip type identification method |
| US20010031086A1 (en) * | 1996-12-25 | 2001-10-18 | Hitachi, Ltd. | Image processor, image processing apparatus, and image processing method |
| CN1460961A (en) * | 2003-06-27 | 2003-12-10 | 杭州信雅达系统工程股份有限公司 | Bill image face identification method |
| CN102129547A (en) * | 2010-01-20 | 2011-07-20 | 富士施乐株式会社 | Form identification apparatus, form processing system and form identification method |
Also Published As
| Publication number | Publication date |
|---|---|
| JP2013114678A (en) | 2013-06-10 |
| JP5619111B2 (en) | 2014-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110738602B (en) | Image processing method and device, electronic equipment and readable storage medium | |
| CN109657665B (en) | Invoice batch automatic identification system based on deep learning | |
| US7657091B2 (en) | Method for automatic removal of text from a signature area | |
| CN102782703B (en) | Page layout determination of an image undergoing optical character recognition | |
| US5857034A (en) | System for inputting character data | |
| CN1184987A (en) | Electronic transaction system | |
| CN102222241A (en) | Apparatus and method for character string recognition | |
| JP2004139484A (en) | Form processing device, program for implementing it, and program for creating form format | |
| CN102156864A (en) | Bill processing system | |
| EP0546343A2 (en) | Diagram recognizing system | |
| US20010051965A1 (en) | Apparatus for rough classification of words, method for rough classification of words, and record medium recording a control program thereof | |
| CN110427853A (en) | A kind of method of smart tickets information extraction processing | |
| CN105590367B (en) | Processing method for transaction abnormity of IC card and acceptance terminal for realizing processing method | |
| CN1185614A (en) | Electronic trade system | |
| CN103136544A (en) | Image judging device | |
| JP4983464B2 (en) | Form image processing apparatus and form image processing program | |
| CN114387604B (en) | Paper invoice information extraction method and device | |
| CN109559235A (en) | Asset reallocation method and device, storage medium and electronic equipment | |
| JP6757152B2 (en) | Banknote processing system and banknote processing method | |
| JP4521466B2 (en) | Form processing device | |
| KR101001693B1 (en) | Method character recognition in a GIRO paper teller machine | |
| JP2005165916A (en) | Magnetic ic card processor | |
| US20220262140A1 (en) | Number plate information specifying device, billing system, number plate information specifying method, and program | |
| CN106558139A (en) | Special certificate recognition methodss, special certificate identifying device and self-service equipment | |
| CN101354717B (en) | Document extracting method and document extracting apparatus |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130605 |