JP6704016B2 - Answer scoring system, program and method - Google Patents

Description

The present invention relates to an automatic scoring technique using an image acquired from an answerer's answer sheet such as a selective answer or a descriptive answer.

Regarding answer scoring, it has been put into practical use that the answer sheet is passed through an image scanner to capture an image, and the image is displayed on a display device of a computer for scoring.
With regard to the scoring technique for such an answer, it is known to arrange the scoring image by information processing based on the relative position of the answer writing position and the scoring image and the answer writing position (for example, Patent Document 1).
Regarding this information processing, a large number of answer sheets are captured by an image scanner, a large number of answer images are cut out from one answer sheet, filed for each question, and the answer image is displayed on the monitor by the grader based on the question number input. It is known to provide it (for example, Patent Document 2).

JP, 2005-167001, A JP-A-2000-122517

By the way, if a correct answer or an answer is taken in from an answer sheet with an image scanner and is made into an image, it is possible to perform information processing such that the answer data is dispersed and collated. However, if the image data is displayed on the display device of the PC and the grader visually scores it, it is not much different from scoring the answer sheet directly. That is, there is a problem that it is not possible to reduce the degree of fatigue of the grader or the error.
The answerers have a wide variety of characters on the answer sheet, and it is not easy to convert the images into characters or data for direct evaluation even if this is converted into an image, and the scoring accuracy will be reduced. There is a problem called.
Therefore, in view of the above problems, an object of the present invention is to reduce the labor of a scoring person, improve the scoring accuracy, and speed up the scoring.

In order to achieve the above object, according to one aspect of the answer scoring system of the present invention, a scanner for scanning an answer sheet and converting it into an answer image, and a rectangle surrounded by a straight line in the answer image captured from the scanner. extracts and answer column in the rectangle issued extract is set, by calculating the number of pixels in the inside of the answer column for each the set the answer column, before the number of pixels is less than a predetermined value SL It determines that wrong answer recognizes the answer column blank, for the answer column of pixels exceeds a predetermined value, for each of the answer columns, rows physician expansion, contraction processing on the edge portion detected from the answers images After deleting, as an unnecessary element, a straight line having a predetermined length or more and an angle within a predetermined range from the answer image, the character element in the answer image in which the difference between the characters recognized as the answer image is smaller than a threshold value is registered. Then, the character elements are converted to obtain answer data, the correct answer data obtained from the answer image and the answer data are compared to determine match or mismatch, and correct/incorrect information is generated in association with the answer column. A processing unit and an information presenting unit that presents the answer image with the correctness information attached thereto are provided.
In this answer scoring system, the processing unit further evaluates the lightness and shade of the answer image, and performs binarization processing of blackening if the pixel brightness is below a threshold value and whitening if the pixel brightness exceeds the threshold value. The grayscale distribution of the answer image may be emphasized .
In this answer scoring system, further, the set threshold value from the answer image conversion to the answer data, matching the threshold value to the correct data of the answer data, may be changed according to mismatch decision result ..

In order to achieve the above object, according to one aspect of the program of the present invention, it is a program to be realized by a computer, wherein a rectangle surrounded by a straight line is extracted from an answer image captured by a scanner, and the extracted rectangle is extracted. If answer column in a square is set, by calculating the number of pixels in the inside of the answer column for each the set the answer column, erroneous confirmed number of pixels and blank pre SL answer column of less than a predetermined value determines that the answer for the answer column of pixels exceeds a predetermined value, for each of the answer columns, the expansion for the detected edge portion from the answer image, shrinking processes have rows, wherein the answer images predetermined length After deleting a straight line that is above and within an angle of a predetermined range as an unnecessary element, the character element in the answer image in which the degree of difference between the recognized character and the character recognized in the answer image is smaller than a threshold value is registered, and the character element is converted. A function of acquiring answer data, a function of comparing the correct answer data and the answer data acquired from the answer image to determine whether they match or not, a function of generating correctness information in association with the answer column, and the correctness The function of presenting answer image data with information is realized by the computer.
In this program, further, the function of evaluating the shade of the answer image, and the binarization processing of blackening if the pixel brightness is less than the threshold value, and whitening if the pixel brightness exceeds the threshold value, and the density distribution of the answer image May be included .
In this program, further the ability to set the threshold value for the conversion of answers image to the answer data, matching the threshold value to the correct data of the answer data, a function of changing in accordance with a mismatch determination result May be included.

According to one aspect of the answer scoring method of the present invention for achieving the above object, there is provided a computer-implemented answer scoring method, wherein the computer scans an answer sheet and converts the answer sheet into an answer image. And a step of extracting a quadrangle surrounded by a straight line from the answer image captured by the scanner, and an answer column is set in the extracted quadrangle , the answer column is set for each of the set answer columns. calculates the number of pixels in the inner, determines that wrong answer by recognizing the number of pixels and blank pre SL answer column of less than a predetermined value, for the answer column of pixels exceeds a predetermined value, for each of the answer column the expansion to the edge portion detected from answer image, it has rows contraction processing, after deleting a straight line at an angle of predetermined length or more and a predetermined range from the answer image as an unnecessary element was recognized as the answer image The character element in the answer image in which the degree of dissimilarity of the characters is smaller than the threshold value is registered, the answer data is acquired by converting the character element, and the correct answer data and the answer data acquired from the answer image are compared and matched. , Each step including a step of determining a mismatch and generating correctness/incorrectness information in association with the answer column, and a step of presenting an answer image with the correctness/incorrectness information attached thereto.

According to the present invention, any one of the following effects can be obtained.
(1) Regardless of whether it is a descriptive answer or a selective answer, the correct answer image or the answer image is acquired from the answer sheet, and the characters are converted into characters for scoring, so the labor of the grader can be reduced and the scoring efficiency can be improved. You can
(2) The correct answer and answer obtained from the answer sheet are compared to determine whether they match or not, and as a result of the determination, the answer is associated with the question and the correct or incorrect information is added to the answer, so the grader can quickly grasp the graded result. be able to.

It is a figure which shows the scoring system which concerns on 1st embodiment. It is a figure which shows the processing sequence of a scoring system. It is a figure which shows the scoring system which concerns on 2nd embodiment. It is a figure which shows the processing sequence of a scoring system. It is a figure which shows the scoring system which concerns on 3rd embodiment. It is a figure which shows the processing sequence of a scoring system. It is a figure which shows the scoring system which concerns on one Example. It is a figure which shows the front side of an answer sheet. It is a figure which shows the back surface side of an answer sheet. It is a figure which shows the binarization process of an answer image. It is a flow chart which shows the processing procedure of automatic scoring of answer data. It is a figure which shows an answer folder. It is a flow chart which shows the processing procedure of scoring. A is a diagram showing an answer column correct answer character type folder, and B is a diagram showing an answer column correct/incorrect folder. It is a flow chart which shows a deletion processing procedure of an unnecessary element. It is a figure which shows an original image. It is a figure which shows the image after removing the line from the original image. It is a figure which shows the image after edge detection. It is a figure which shows the image after an expansion process. It is a figure which shows the image after a contraction process. It is a figure which shows the image used for character recognition. It is a figure which shows the image before and after removal of an unnecessary element. It is a figure which shows expansion processing and contraction processing.

[First Embodiment]
FIG. 1 shows a scoring system according to the first embodiment. The configuration shown in FIG. 1 is an example, and the present invention is not limited to such a configuration.
The scoring system 2 is, for example, a system for automatically scoring answers including descriptive answers and selection answers. The scoring system 2 includes an image scanner 4, a processing unit 6, a storage unit 8, an operation input unit 10 and an information presenting unit 12, which may be connected via a network such as the Internet as an example.

The image scanner 4 is controlled by the processing unit 6, scans the answer sheet, converts it into answer data of image data, and provides this answer data to the processing unit 6. This answer data includes both correct answer data representing the correct answer and answer data of the answer submitted by the answerer.
The processing unit 6 is, for example, an information processing device such as a personal computer (PC). The information processing by the processing unit 6 includes a) image capture processing 14, b) registration processing 16, c) data conversion processing 18, d) determination processing 20, e) correct/incorrect information generation processing 22, and the like.

In addition to the OS (Operating System), the storage unit 8 stores information used in the scoring system 2. The storage unit 8 stores f) answer folder 32, g) scoring folder 34, and the like.
The operation input unit 10 is used for inputting information necessary for the scoring system 2. The operation input unit 10 includes, for example, a keyboard and a mouse, and a touch panel installed on the display screen of the information presenting unit 12.
The information presenting unit 12 displays the input image and the scoring result image in the scoring system 2, and the information presenting unit 12 uses, for example, an LCD (Liquid Crystal Display).

<Information processing by the processing unit 6>
a) Image capture processing 14
The image scanner 4 scans and reads the answer sheet and converts it into answer data consisting of image data. In the image capturing process 14, the answer data output from the image scanner 4 is captured by the processing unit 6.

b) Registration process 16
This registration processing 16 is executed under the control of the processing unit 6 and the answer data taken into the processing unit 6 from the image scanner 4 is registered in the storage unit 8. The registration of the answer data is performed in association with the tester, the answerer, or the question.
c) Data conversion process 18
In this data conversion process 18, the processing unit 6 converts the answer data (image data) into character data. In this data conversion, the correct answer data representing the correct answer character is generated from the correct answer image, and the answer data representing the answer character is generated from the answer image. The correct answer data is stored in the storage unit 8 in association with the question, and the answer data is stored in the storage unit 8 in association with the answerer.

d) Judgment processing 20
In this determination processing 20, the processing unit 6 determines whether the answer in the answer to the question is a correct answer. That is, for a specific question, if the answer data matches the correct answer data, the judgment is "correct answer", and if the answer data does not match the correct answer data, judgment is "wrong answer". That is, the answer data is a character code converted from the answer image, and the correct answer data is also a character code. If these character codes match, the judgment is "correct answer", and if they do not match, the judgment is "wrong answer".
The result of the determination process 20 is stored in the storage unit 8 in association with the answerer.

e) Correct/wrong information generation process 22
The correctness/incorrectness information generation process 22 is executed by the processing unit 6, and correctness/incorrectness information is generated in association with the determination result of the determination process 20. The generation of the correctness/incorrectness information includes generation of information indicating correctness/incorrectness as a determination result and generation of correctness/incorrectness display. The former is generated in association with the answerer and the question of the answer, and is stored in the storage unit 8. The latter is a correct/incorrect display displayed in association with the answer column of the answer image.

<memorized information>
f) Answer folder 32
The answer folder 32 stores answer data captured by the image scanner 4.
g) Scoring folder 34
In the scoring folder 34, correct answer character data, answer character data, determination result information, etc. are stored.

<Processing sequence>
FIG. 2 shows a processing sequence of the scoring system 2. This processing sequence is an example of the answer scoring processing program or scoring method. In this processing sequence, S is a processing step, and the numbers given to S are examples of the order of steps (the same applies hereinafter).
In the scoring system 2, upon scoring, the processing unit 6 and the image scanner 4 are activated, the image scanner 4 scans and reads the answer sheet (S101), and converts it into answer data composed of image data representing a correct answer image or an answer image. (S102).

The processing unit 6 acquires the answer data from the image scanner 4 (S103) and performs a data conversion process (S104). In this data conversion process, the answer data is converted into character data, and correct answer data is generated from the correct answer image and answer data is generated from the correct answer image. The answer data is stored in the storage unit 8 by the processing unit 6.
The processing unit 6 determines whether the correct answer data and the answer data match or not (S105), and generates a correct/incorrect display together with the determination result information (S106). After the generation of the correct/incorrect display, the processing unit 6 generates the presentation information with the correct/incorrect display and provides it to the information presenting unit 12 (S107).
The information presentation unit 12 presents the presentation information provided by the processing unit 6 (S108).

<Effects of First Embodiment>
According to the first embodiment, the following effects can be obtained.
(1) Acquire answer data (image data) from the answer sheet, generate correct answer data from the correct answer image or answer data from the answer image, and share the correct answer data and answer data with a common character code, for example It is possible to make a judgment and automatically make a correct answer or an incorrect answer.
(2) Since the correct/incorrect display is generated in association with the answer column along with the judgment result, the grader can easily recognize whether the question is correct or not.

[Second embodiment]
FIG. 3 shows a scoring system according to the second embodiment. 3, the same parts as those in FIG. 1 are designated by the same reference numerals. The configuration shown in FIG. 3 is an example, and the present invention is not limited to such a configuration.
The answer written on the answer sheet submitted by the answerer is inconsistent, such as shading and collapse of letters. In the answer data acquired from such answer sheets, unstable elements such as shading and collapse of characters by the answerer reduce the conversion accuracy of the answer data from the answer image, and the answer is incorrect, but the answer is incorrect. You cannot ignore the case. In order to avoid such an inconvenience, in the scoring system 2 according to the second embodiment, the information processing of the processing unit 6 is added with h) shade evaluation processing 24 and i) binarization processing 26.

<Information processing by the processing unit 6>
h) Shading evaluation processing 24
The processing unit 6 executes the lightness/darkness evaluation processing 24 on the answer data before the data conversion processing 18, and judges the lightness/darkness of the answer image. Here, regarding the light and shade evaluation, “dark” means that the contrast of the character image with respect to the background image is outstanding, and “light” means that the contrast of the character image with respect to the background image is low and confusing.

i) Binarization processing 26
The processing unit 6 performs binarization processing 26 based on the result of the grayscale evaluation processing 24. The binarization process 26 is an example of a contrast enhancement process for answer data. That is, a step value is set for the pixel brightness for a gradation image from “light” to “dark”. For example, if the pixel brightness exceeds the threshold value, it is white, and if the brightness is less than the threshold value, it is black.

<Processing sequence>
FIG. 4 shows a processing sequence of the scoring system 2 including the gradation evaluation processing 24 and the binarization processing 26.
In this scoring system 2, after the answer sheet is read by the image scanner 4 (S201) and converted into answer data (S202), the processing unit 6 acquires the answer data from the image scanner 4 (S203).

The processing unit 6 evaluates the shade of the acquired answer data (S204), and binarizes the data according to the result of the shade evaluation (S205). That is, the answer data is converted into the answer data in which the shade is outstanding by the binarization process.
After the binarization process, the processing unit 6 performs a data conversion process (S206). This data conversion process is the same as the above-described process (S104), and therefore its explanation is omitted.
The processing unit 6 determines whether the correct answer data and the answer data match or not (S207), generates a correct/incorrect display together with the determination result information (S208), and, after the correct/incorrect display is generated, generates the presentation information with the correct/incorrect display. Then, the presentation information is provided to the information presenting unit 12 (S209). The information presentation unit 12 presents the presentation information provided by the processing unit 6 (S210).

<Effects of Second Embodiment>
According to the second embodiment, the following effects can be obtained in addition to the effects of the first embodiment.
(1) It is possible to avoid the influence of shading in the answer data, improve the accuracy of converting the answer data into a character code, and as a result, increase the accuracy of determining a correct answer or an incorrect answer.
(2) Since the grader is freed from the work of checking the influence of the shade of the answer sheet in advance, it is possible to reduce the scoring labor and to improve the efficiency of the scoring process.

[Third Embodiment]
FIG. 5 shows a scoring system according to the third embodiment. 5, the same parts as those in FIG. 1 are designated by the same reference numerals. The configuration shown in FIG. 5 is an example, and the present invention is not limited to such a configuration.
The answer written on the answer sheet submitted by the answerer may include unnecessary elements other than the answer character, such as a part of the character due to incomplete deletion. In the answer data acquired from such an answer sheet, an unnecessary element deteriorates the conversion accuracy of the answer image into the answer data, and it is not possible to ignore the case where the answer is a correct answer although it is a correct answer. In order to avoid such an inconvenience, in the scoring system 2 according to the third embodiment, j) unnecessary element deletion processing 30 is added to the information processing of the processing unit 6.

<Information processing by the processing unit 6>
j) Unnecessary element deletion processing 30
By executing the unnecessary element deletion processing 30, the processing unit 6 deletes unnecessary elements other than the character elements to be recognized from the answer data (answer image) from the answer data, and executes the data conversion processing 18 with this answer data.

<Processing sequence>
FIG. 6 shows a processing sequence of the scoring system 2 including the unnecessary element deletion processing 30.
In this processing sequence, unnecessary elements are deleted from the answer data through the processing unit 6, and the subsequent data is registered. This registration includes the processing performed in the storage unit 8. The unnecessary elements may be assumed to be stains other than the remaining answer characters, such as a part of characters due to incomplete deletion in the answer sheet.
The processing unit 6 reads the answer sheet by the image scanner 4 (S301) and converts it into answer data (S302), and then the processing unit 6 acquires the answer data from the image scanner 4 (S303).

The processing unit 6 deletes unnecessary elements from the answer data and registers the answer data (S304). A data conversion process in which this unnecessary element is deleted is performed (S305). This data conversion process is the same as the above-described process (S104), and therefore its explanation is omitted.
The processing unit 6 determines whether the correct answer data and the answer data match or not (S306), generates a correct/incorrect display together with the determination result information (S307), and, after generating the correct/incorrect display, generates the presentation information with the correct/incorrect display. Then, the presentation information is provided to the information presenting unit 12 (S308). The information presentation unit 12 presents the presentation information provided by the processing unit 6 (S309).

<Effects of the third embodiment>
According to the third embodiment, the following effects can be obtained in addition to the effects of the first embodiment.
(1) Unnecessary elements such as stains can be deleted from the answer data by information processing, the accuracy of converting the answer data into a character code can be improved, and as a result, the accuracy of determining a correct answer or an incorrect answer can be improved.
(2) Since the grader is freed from the work of checking in advance the effects of unnecessary elements such as stains on the answer sheet, the scoring effort can be reduced and the scoring process can be made more efficient.

FIG. 7A shows a scoring system according to one embodiment. In FIG. 7A, the same parts as those in FIG. 1 are designated by the same reference numerals.
The scoring system 2 includes an image scanner 4 and a PC (personal computer) 36. The PC 36 includes a processing unit 6, a keyboard 38, a monitor 40, etc., which are linked by a network 42.
The scoring system 2 of this example is an example of the above-described first embodiment, second embodiment, and third embodiment, and has the functions described in each embodiment.
The image scanner 4 interlocks with the PC 36, scans the answer sheet, acquires answer data which is image data from the answer sheet, and provides the answer data to the PC 36. The image scanner 4 may be an image reading unit of a multifunction machine having a copying function and a fax function.
The keyboard 38 is an example of the operation input unit 10 (FIG. 1). The operation input unit 10 may be a user interface such as a mouse together with the keyboard 38 or instead of the keyboard 38, or a touch panel installed on the display screen of the monitor 40.
The monitor 40 is an example of the information presenting unit 12, and for example, an LCD may be used, or a printer may be used together with the monitor.

<Hardware of scoring system 2>
B of FIG. 7 shows the hardware of the scoring system 2. 7B, the same parts as those in FIG. 7A are denoted by the same reference numerals.
The PC 36 includes a processing unit 6, a storage unit 8, and an input/output unit 44. The processing unit 6 executes the OS and the scoring program stored in the storage unit 8.
The storage unit 8 is provided with a storage element that stores an OS, a scoring program, and the like. The storage element includes a ROM (Read-Only Memory) and a RAM (Random-Access Memory). An answer folder 32, a grading folder 34, etc. are stored in the storage unit 8.

<answer sheet 46>
FIG. 8 shows the front side of the answer sheet 46, and FIG. 9 shows the back side thereof. The answer sheet 46 is an example of the answer sheet, and the present invention is not limited to such a configuration.
The front surface 46-1 of the answer sheet 46 constitutes a so-called mark sheet. On this surface 46-1, taper cut 48, positioning mark 50, identifier 52, name field 54, date field 56, attribute fields 58 and 60, answerer identification field 62, answer fields 64-1, 64-2 and 64. -3 is printed.

The taper cut 48 is used to identify the answer sheet 46 upside down. The positioning mark 50 is the identifier 52 of the answer sheet 46, the name column 54, the date column 56, the attribute columns 58 and 60, the answerer identification column 62, the answer columns 64-1 and 64-2 in the answer data 46 when scanned. It is used for alignment information for specifying the position of 64-3.
The identifier 52 is identification information such as the model number of the answer sheet 46. The name of the answerer is entered in the name field 54. The date field 56 is filled in with the date of implementation of the test. In the attribute columns 58 and 60, attributes for identifying whether the answer sheet 46 is the model answer (correct answer) or the examinee's answer are entered.
The answerer identification field 62 stores identification information such as the affiliation and number of the answerer. The answer columns 64-1, 64-2, 64-3 are provided with an answer number and a mark writing section for each question, and the answerer can answer by filling in the corresponding answer number.

The back surface 46-2 of the answer sheet 46 constitutes a so-called fill-in type sheet surface. A taper cut 48, a positioning mark 50, an identifier 52, an area frame 66, and fill-in type answer columns 68-1, 68-2, 68-3 are printed on the back surface 46-2. Since the taper cut 48, the positioning mark 50, and the identifier 52 are as described above, the description thereof will be omitted. The area frame 66 is an envelope line that represents an area that specifies the fill-in-type answer columns 68-1, 68-2, and 68-3. In the area frame 66, "paper vertical top" is displayed as identification information for specifying the upper side, and "paper horizontal top" is displayed as identification information for specifying the lateral side.
The fill-in type answer columns 68-1, 68-2, 68-3 are provided with numbers and entry portions representing questions, and letters are entered by the answerer.

<Mechanism of processing of answer sheet 46>
If the answer sheet 46 is scanned by the image scanner 4, the model number of the answer sheet 46 can be selected by the identifier 52. If this model number is registered, the answer sheet 46 can be read. In this case, if both the front surface 46-1 and the back surface 46-2 are on both sides, it is judged from the image on the front surface 46-1 and the image on the back surface 46-2 that the paper is double-sided, and if the combination is different, the reading is performed. It can be judged as an error. In the reading process of the answer sheet 46, it is possible to process the answer sheet 46 upside down without any problem.

The following information can be acquired by using the model number of the answer sheet 46.
A) Distinction between single-sided and double-sided a) In the case of single-sided, the analysis value of the identifier of the front surface 46-1 c) In the case of double-sided, the analysis value of the identifier of the front surface 46-1 and the analysis value of the rear surface 46-2 Both) The position of the answer column, the number of answer columns selected, the number of answer columns, the attributes of the answer column, the individual identification, the elliptical mark and the check mark, and the descriptive answer.

<Binary processing of handwritten characters in answer data>
10A shows handwritten characters, FIG. 10B shows answer data, FIG. 10C shows intermediate stage data in the binarization process, and D in FIG. 10 shows the binarized character data.
Regarding the answer "A" in the entry type answer field 68-1 on the back surface 46-2 of the answer sheet 46, the answerer is in a state in which the answer "A" is written in blurred characters as shown in A of FIG. The answer "A" is reminiscent of the katakana "A", but each element has a gradation state from an element close to white to an element recognized as black.
When such a handwritten character is scanned and taken into the processing unit 6 from the image scanner 4, it is converted into image data having a gradation of gradation as shown in B of FIG. The image data is shown in four shades (luminance) for the sake of explanation. Regarding "a", when each part is shown by a brightness level, it is composed of brightness level L1 (=lightest), L2 (=lightest), L3 (=lightest), and L4 (=darkest).

When this image data is binarized, the brightness level L1 is assimilated to the background color, and the brightness levels L2 and L3 are converted into dark areas like the brightness level L4. As a result, as shown in C and D of FIG. 10, the portion 70-1 of the luminance level L1 is assimilated to the background color, while the portion 70-2 of the luminance level L2 is black that is distinctive to the background color. Changes to.
By this binarization processing, the image data is converted into image data having a remarkable contrast.

<Delete unnecessary elements from image data>
In order to improve the accuracy of character recognition of the answer data, straight line information is acquired from the image obtained by binarizing the original image. For example, if the acquired straight line has a length of 15 mm or more and the extracted line has an angle of 85 degrees to 95 degrees or −5 degrees to 5 degrees, the unnecessary element is deleted. It is used as the target line information and the unnecessary line is deleted from the original image.

<Threshold operation for character recognition and judgment>
The answer data is binarized, and character recognition of the answer is performed on the binarized answer data. It is determined that the smaller the recognized character data and the dissimilarity (a value representing the difference between the image and the character recognized character), the closer the image and the recognized character are.
For example, when the initial value of the threshold for each character is set as the threshold for using the character recognition result, for example, 150 is set for each character, and when the dissimilarity is less than the threshold, the character recognition result is automatically scored. To use. If the result of automatic scoring is incorrect when the degree of difference for each character is less than or equal to the threshold value on the grading screen, the character threshold value may be lowered. If the degree of difference for each character on the grading screen is greater than or equal to the threshold value and automatic scoring is not performed, but the result is correct, the character threshold value may be increased.

<Correct answer data registration and automatic scoring processing>
FIG. 11 shows a procedure of correct answer data registration and automatic scoring. This processing procedure includes the following processing.
The correct answer image (correct answer image 1) on the answer sheet 46 is saved in the answer folder 32A (A in FIG. 12) with a file name "ttt.jpg", for example.
Generate an answer column candidate list. That is, when the image of the answer data is displayed on the monitor 40, the rectangle generated by the binarization process and surrounded by the straight line is extracted from the image to create the answer column candidate list.

In the image displayed on the screen, if a rectangle surrounded by a straight line is selected (for example, the inside of the rectangle is double-clicked with the mouse), answer column candidates are extracted from the answer column candidate list. This is set as the answer column, and a straight line representing a rectangle is displayed as a red line as the answer column. The process of selecting a rectangle and shifting to red display is repeated for all answer columns.
Then, in the automatic scoring setting, if an answer column is selected with a mouse and “automatic scoring setting” is selected, an automatic scoring setting screen is displayed. Create an image with lines removed from the image when the first automatic scoring setting screen is displayed. This image is saved in the answer folder 32B (B in FIG. 12) with a file name "ttt.bmp", for example.
An image obtained by performing binarization processing on the answer column cut out from the image is created, character recognition is performed on that part, correct answer candidates are displayed in the setting screen, and after confirmation, registration is performed. The result is stored in the answer folder 32C (C in FIG. 12).

This processing will be described according to the processing procedure shown in FIG. 11. The image scanner 4 and the processing unit 6 are activated, and the image scanner 4 scans the answer sheet 46 to read the answer data (S401).
The processing unit 6 determines whether the identifier 52 is correct from the answer data (S402). If the identifier 52 is different from the registered identifier, it is determined that there is a read error (S403), and this processing ends. If the identifier 52 matches the registered identifier, the identifier 52 is determined to be correct, and the answer data which is image data is saved (S404).
The processing unit 6 evaluates the shade of the image of the stored answer data, then performs binarization processing on the answer data (S405), extracts a rectangle from the answer data (S406), and answers box candidate list. Is created (S407), and the image of the answer data is displayed on the monitor 40 (S408).

As the answer column setting by selecting the rectangle displayed on the screen, for example, by placing the cursor on the rectangle and double-clicking the mouse, the processing unit 6 selects the rectangle (S409) and sets this as the answer column. (S410). The processing unit 6 displays the answer column in red (S411), thereby clearly indicating that the answer column has been set. The answer column is set by repeating the operations of S409 to S411. The processing unit 6 determines whether all the answer columns have been set for the answer data (S412), and if all the answer columns have not been set, S409 to S412 are repeated.
If the processing unit 6 determines that all of the answer columns of the answer data have been set, it determines whether the score is automatic (S413). Whether or not automatic scoring depends on user settings. If it is not the automatic grading, that is, if it is the hand grading, this processing is ended (S414).
If the automatic scoring is performed, the processing unit 6 displays an automatic scoring setting screen on the monitor 40 (S415), binarizes the image (S416), and removes unnecessary elements such as noise including removal of lines that are unnecessary elements. Are removed (S417), the image that has undergone these processes is saved (S418), the characters are recognized from the answer data (S419), the answer is registered (S420), and this process ends.

<Answer folder 32A, 32B, 32C>
As shown in A of FIG. 12, the answer folder 32A includes an identification unit 72 and an image storage unit 74. The identifier of the image stored in the image storage unit 74 is stored in the identification unit 72. The image information specified by the identifier is stored in the image storage unit 74, and in addition to the correct answer image, the answer image of the student aaa who is the examinee, the answer image of the student bbb,...

As shown in B of FIG. 12, the answer folder 32B includes an identification unit 76 and an image storage unit 78. The identifier of the image stored in the image storage unit 78 is stored in the identification unit 76. Image information specified by an identifier is stored in the image storage unit 78, and in addition to the correct answer image, the answer image of the student aaa who is the examinee, the answer image of the student bbb,... The image from which is removed is stored.

As shown in C of FIG. 12, the answer folder 32C includes an identification unit 80 and an image storage unit 82. The identifier of the image stored in the image storage unit 82 is stored in the identification unit 80. The image storage unit 82 stores the answer image specified by the identifier for each of the examinees (students) and the answer columns 1, 2, 3,... That is, the image storage unit 82 stores the answer images of the students aaa, the students bbb,...

<Automatic scoring process>
FIG. 13 shows a processing procedure of automatic scoring processing. In this processing procedure, the processing unit 6 reads the answer sheet with the image scanner 4 (S501) and acquires the answer data. The answer data is stored in the answer folder 32 with a file name of, for example, "aaa.jpg".
The processing unit 6 confirms the identifier 52 read from the stored answer data (S502). If this identifier 52 does not match the registered identifier, a read error (S503) occurs and this processing ends. If this identifier matches the registered identifier, the scoring process is continued and the image (answer data) is saved (S504).

The processing unit 6 confirms the attribute of the answer column (S505) and determines whether the attribute of the answer column is a mark type answer (selection type answer) or a descriptive type answer.
If it is a mark type, the processing unit 6 determines whether it is correct or incorrect (S506). In this determination, which mark is filled is recognized, and if the preset correct answer and the recognition result match, the correct answer is determined, and if they do not match, the incorrect answer is determined. If it is a correct answer, the correct answer is held (S507), and if it is an incorrect answer, the incorrect answer is held (S508). At this time, the correctness information or the score in the answer column is held, and the marking of the mark type answer is finished.

If the answer is a descriptive answer, the processing unit 6 determines whether the answer column is blank as attribute determination of the answer column (S509). The number of pixels inside the answer column is used for this determination. The number of pixels inside the answer column is calculated, and if the number of pixels is a fixed value or less, for example, 3 pixels or less, it is determined that the answer column is blank. When it is determined that the answer column is blank, the answer column is determined as an incorrect answer, and the correct answer of the answer column is stored in the scoring result as “wrong answer” and “score 0” (S510) and reflected in the answer column. (S520).
If the answer column is not determined to be blank, that is, if the answer column is “answer available”, the processing unit 6 creates an image in which unnecessary elements such as lines are removed from the answer column (S511), and this image is generated. After the binarization processing, the image is stored in the answer folder 32B (B in FIG. 12) with the file name as "aaa.bmp" (S512).

The processing unit 6 confirms whether the processing is automatic scoring (S513). In the case of automatic scoring, the answer in the image is acquired (S514), the image in the answer column is binarized, and the character is recognized (S515).
The processing unit 6 confirms the degree of difference in character recognition (S516). In the confirmation of the dissimilarity (S516), it is determined whether or not the dissimilarity is smaller than a threshold value for each character, and automatic scoring is performed using the result of character recognition (S517). If the character matches the correct answer character in the answer column, it is determined to be the correct answer. The answer column/correct answer character type folder 32D (A in FIG. 14) is used for this determination. The processing unit 6 holds, for example, a correct answer score of 5 in the correctness/incorrectness information scoring result in the answer column, determines that it is an incorrect answer when the scores do not match, and determines the answer column/correctness folder 32E (B in FIG. 14). The erroneous answer and the score of 0 are held in the scoring result of the correctness information in the answer column (S518).
The processing unit 6 confirms whether the answer character is dark or light (S519), and when it is light, it is reflected in the answer column (S520) of the image, and when it is dark, the image is displayed on the screen (S521). To finish.

<answer box/correct answer character type folder 32D>
14A shows an example of the answer column/correct answer character type folder 32D.
The answer column/correct answer character type folder 32D includes an answer identification field 84, a character type field 86, a correct answer field 88, and an automatic scoring field 90. The answer identification column 84 stores identification information representing the answer column. Character type information such as katakana and numbers is stored in the character type column 86 as the characters described in the answer column. Character data such as characters is stored in the correct answer column 88 as correct answer information. Correct or incorrect identification information is stored in the automatic scoring column 90.

<Answers/Correct folder 32E>
B of FIG. 14 shows an example of the answer column/correctness information folder 32E.
The answer column/correctness folder 32E includes an answerer column 92, an answer identification column 94, an answer column 96, a scoring result column 98, and a score column 100. The answerer column 92 stores the answerer's identification information. The identification information of the answer column is stored in the answer identification column 94. Answer information is stored in the answer column 96. The scoring result column 98 stores scoring result information such as whether the answer by the answerer is correct or incorrect. The score column 100 stores the score obtained by the solver.

<Delete unnecessary elements of answer data, etc.>
FIG. 15 shows a processing procedure such as an unnecessary element deletion processing performed before scoring the answer data. This processing procedure includes, as unnecessary elements, processing such as noise removal, sloughing dirt removal, small noise removal, and connection of discontinuous lines.
The processing unit 6 creates an image (FIG. 17) in which lines including ruled lines are removed from the original image (FIG. 16) (S601), and creates an image of only edges from the image in which lines are removed (FIG. 17). An image (FIG. 18) is created by emphasizing a clear portion of a character line and removing blurred stains and light-colored noise (S602).
The processing unit 6 further expands the edge portion of the edge-only image (FIG. 18) to create an image (FIG. 19) in which the discontinuous portion is processed into a continuous line, and the expanded image ( An image (FIG. 20) contracted from FIG. 19) is created, and the thickness of the line is restored (S603). Further, the processing unit 6 erases small size elements having a certain size or less, for example, 1 mm ² or less as a threshold, as noise (S604).
Then, the processing unit 6 performs black-and-white inversion again on the image in which black-and-white is inverted at the time of edge detection to generate an image for character recognition (FIG. 21).

<Image processing from the original image to the image for character recognition>
FIG. 16 shows an original image of answer data. Lines and answers are clearly shown in this original image.
FIG. 17 shows an image in which lines have been removed from the original image. The line is removed from this original image, and only the answer information is specified.
FIG. 18 shows a black-and-white inverted image of the image with the lines removed. Edges are emphasized in this image.
FIG. 19 shows a black and white inverted image after the expansion processing. In this image, edges are emphasized, and discontinuous portions are continuous lines.
FIG. 20 shows a black-and-white inverted image after contraction processing. In this image, edges are emphasized and only answer information is clearly shown.
FIG. 21 shows an image for character recognition through black and white inversion.
22A and 22B show “M” and “N” as characters on which noise is superimposed. As for "M" and "N", frictional stains on the roller occur in the original image. On the other hand, C and D in FIG. 22 show “M” and “N” with noise removed.

<Edge detection, expansion/contraction processing>
A known algorithm may be used to detect the edge of the image.
The expansion process performed on the image is a process of enlarging the figure in the image by one pixel. On the other hand, the contraction process is the opposite of the expansion process and is a process of narrowing the figure in the image by one pixel. For this process, for example, as shown in A of FIG. 23, the process of “4 vicinity” is used.
B of FIG. 23 shows the original image. In the expansion process for the original image, the central portion N of the pixel shown in FIG. 23A is applied to the pixels in the vicinity of each pixel of the original image, and if any one of the neighborhoods is “1” in the original image, the central portion is expanded. The four neighboring pixels of N are moved to “1” and enlarged. As a result, an expanded image is generated as the original image as shown in C of FIG.
Further, in the contraction processing for the original image (B in FIG. 23), the central portion N of the pixel shown in A of FIG. 23 is applied to the pixels in the vicinity of each pixel of the original image, and any one of the neighborhoods is the original image. If it is "0", the four neighboring pixels in the central portion N are moved to "0" and contracted. As a result, a contracted image is generated as the original image as shown in D of FIG.

<Effect of Example>
According to this embodiment, the following effects can be obtained.
(1) It is possible to read the answer data (image) from the answer sheet with an image scanner and perform automatic scoring based on the answer data.
(2) Since the binarization processing is performed by gray scale, the character recognition accuracy can be improved without being affected by the light and shade, the efficiency of automatic scoring and the scoring accuracy can be improved, and the reliability of the scoring processing can be improved. In addition to being improved, the labor of the grader can be reduced and the scoring efficiency can be enhanced.

As described above, the scoring system, the program, and the method of the present invention have been described with respect to the most preferable embodiments. The present invention is not limited to the above description. Various modifications and changes can be made by those skilled in the art based on the gist of the invention described in the claims or the mode for carrying out the invention. It goes without saying that such modifications and changes are included in the scope of the present invention.

According to the present invention, by reading an answer sheet containing a selective answer or a descriptive answer with an image scanner, automatic scoring of a selective answer and automatic scoring by character recognition of a descriptive answer can be used together, and scoring is easy and short. Build a system that can be processed in time.

2 scoring system 4 image scanner 6 processing unit 8 storage unit 10 operation input unit 12 information presenting unit 14 image capturing process 16 registration process 18 data conversion process 20 determination process 22 correct/incorrect information generation process 24 grayscale evaluation process 26 binarization process 30 unnecessary Element deletion process 32 Answer folder 34 Grading folder 36 PC
38 keyboard 40 monitor 42 network 44 input/output section 46 answer sheet 46-1 front surface 46-2 back surface 48 taper cut 50 positioning mark 52 identifier 54 name column 56 date column 58, 60 attribute column 62 answerer identification column 64-1, 64 -2, 64-3 answer column 66 area frame 68-1, 68-2, 68-3 fill-in-type answer column 70-1 L1 portion 70-2 L2 portion

Claims (7)

A scanner that scans answer sheets and converts them into answer images,
When a quadrangle surrounded by a straight line is extracted from the answer image captured from the scanner and an answer column is set in the extracted quadrangle , inside the answer column for each of the set answer columns calculates the number of pixels, it is determined that a wrong answer by recognizing the number of pixels and blank pre SL answer column of less than a predetermined value, for the answer column of pixels exceeds a predetermined value, for each of the answer columns, the answer expansion to the edge portion detected from the image, have rows contraction processing, after deleting a straight line at an angle of predetermined length or more and a predetermined range from the answer image as an unnecessary element of characters recognized as the answer image A character element in the answer image whose dissimilarity is smaller than a threshold value is registered, the character element is converted to obtain answer data, and the correct answer data and the answer data acquired from the answer image are compared to each other, and the two do not match. And a processing unit that generates correctness information in association with the answer column,
An information presenting unit that presents an answer image with the correctness information,
An answer scoring system comprising: Furthermore, the processing unit evaluates the shade of the answer image, performs binarization processing of blackening if the pixel brightness is less than or equal to a threshold value and whitening if the pixel brightness exceeds the threshold value, and the shading distribution of the answer image. The answer scoring system according to claim 1, characterized by: Further, a threshold value is set for conversion from the answer image to the answer data, and the threshold value is changed according to a determination result of whether the answer data matches the correct answer data or not. A written answer scoring system. A program to be realized by a computer,
When a quadrangle surrounded by a straight line is extracted from the answer image captured by the scanner and an answer column is set in the extracted quadrangle , the number of pixels inside the answer column for each of the set answer columns is calculated, and determines that the wrong answer by recognizing the number of pixels and blank pre SL answer column of less than a predetermined value, for the answer column of pixels exceeds a predetermined value, for each of the answer columns, from the answer image expands against the detected edge portion, have rows contraction processing, after deleting a straight line at an angle of predetermined length or more and a predetermined range from the answer image as required elements, character dissimilarity recognized with the answers image A function of registering a character element in the answer image in which is smaller than a threshold value and converting the character element to obtain answer data,
A function of comparing the correct answer data and the answer data acquired from the answer image, and determining a mismatch,
A function for generating correctness information in association with the answer column,
A function of presenting the answer image data with the correctness information,
A program for realizing the above in the computer. Further, a function of evaluating the shade of the answer image,
If the pixel brightness is less than or equal to a threshold, blackening, if the pixel brightness exceeds the threshold, whitening is binarized, and a function to highlight the density distribution of the answer image,
The program according to claim 4 , comprising: Furthermore, a function of setting a threshold value for conversion from the answer image to the answer data,
A function of changing the threshold value according to the judgment result of whether the answer data is the correct answer data or not,
The program according to claim 4 , comprising: A computer-implemented answer scoring method,
The computer is
Scanning the answer sheet and converting the answer sheet into an answer image;
When a quadrangle surrounded by a straight line is extracted from the answer image captured from the scanner, and an answer column is set in the extracted quadrangle , a pixel inside the answer column is set for each of the set answer columns. and calculates the number, determines that the wrong answer by recognizing the number of pixels and blank pre SL answer column of less than a predetermined value, for the answer column of pixels exceeds a predetermined value, from the answer image for each of the answer column expands against the detected edge portion, have rows contraction processing, after deleting a straight line at an angle of predetermined length or more and a predetermined range from the answer image as required elements, character dissimilarity recognized with the answers image Is registered in the answer image smaller than a threshold value, the character element is converted to obtain answer data, the correct answer data obtained from the answer image and the answer data are compared to determine a match or a mismatch. And a step of generating correctness information in association with the answer column,
Presenting the answer image with the correctness information,
An answer scoring method characterized by performing each step including.

Images