JP2021093225A - Information processing device, program, and information processing method - Google Patents

Abstract

To make character recognition processing applicable to an image in which accurate detection of an outline of a character is difficult.SOLUTION: In accordance with this invention, a search region is set to a recognition object image, and a plurality of cutout regions are set within the search region, and images corresponding to the plurality of set cutout regions respectively are extracted, and each of the extracted images is compared with dictionary data to detect candidate character information and position information of a cutout region corresponding to the candidate character. Then, a piece of candidate character information having the highest evaluation value, out of the detected candidate character information is outputted as a recognition result. Further, a search region for the next character is set on the basis of position information of the cutout region corresponding to the recognition result.SELECTED DRAWING: Figure 12

Description

The present invention relates to character recognition processing.

Conventionally, the character recognition process for a document image obtained by scanning a paper document detects the outline (outline) of a character from the document image, cuts out the character image for each character, and obtains the cut out character image. , Character recognition processing was performed to identify what character it was. Further, since the correct character is not recognized if the character cutout position is incorrect, a technique for correcting the character cutout position according to the user's instruction is also provided. For example, if one character image is cut out as a plurality of characters (for example, if one kanji is cut out by dividing it into a bias and a right component), there is a technique for correcting them as one character. .. Patent Document 1 discloses a technique in which when a user corrects a character recognition result, a part in which the same erroneous recognition is performed is searched from an unmodified part and the same correction is applied.

Further, in recent years, with the spread of smartphones and digital cameras, it has become possible to easily obtain image information including text information. This is opening up a large market for capturing character information by performing character recognition processing from a wider variety of observation environments. For example, at a quarry site such as a mine, there is a use case in which a serial number stamped on a tire is used to manage the tire used for a dump truck. Therefore, it is conceivable to take a picture of the serial number engraved on the tire with a smartphone, a digital camera, or the like, perform character recognition processing on the taken image, and manage using the serial number of the character recognition result. However, if the contrast between the characters and the background is small, or if the surface is dirty and noisy, such as the serial number engraved on the tire, the outline of the characters will be detected accurately as in the conventional technology. The thing itself is difficult.

Japanese Unexamined Patent Publication No. 11-143983

If the conventional technique of cutting out characters based on the outline of characters is applied to an image in which the outline of characters cannot be detected accurately, the frequency of erroneous cutting out positions of characters increases, and the user corrects the recognition result. The burden of doing it also increases.

An object of the present invention is to make it possible to apply character recognition processing to an image in which it is difficult to accurately detect the outline of characters.

In order to solve the above problems, the information processing apparatus of the present invention has a first setting means for setting a search area for a recognition target image, and a first setting means for setting a cutout area at a plurality of places in the search area. By extracting the images corresponding to each of the setting means 2 and the plurality of cutout areas set by the second setting means and comparing each extracted image with the dictionary data, the candidate character information can be obtained. It has a character detecting means that detects the position information of the cutout area corresponding to the candidate character and outputs the candidate character information having the highest evaluation value from the detected candidate character information as a recognition result. The first setting means further sets a search area for the next character based on the position information of the cutout area corresponding to the recognition result output by the character detecting means, thereby causing the second setting means. It is characterized in that the processing by the character detecting means and the character detecting means is repeatedly executed.

For an image in which it is difficult to accurately detect the outline of characters, multiple areas are cut out while shifting the position in the search area where characters are presumed to be present, and character recognition processing is performed based on the cut out multiple areas. By applying, the accuracy of character recognition processing can be improved.

An example of the appearance of a mobile device An example of hardware configuration An example of software configuration in mobile terminal 100 An example of character image information (dictionary data) Conceptual diagram of character recognition processing Recognition result display screen example Example of display screen when instructing to correct the recognition result Display screen example after correction of recognition result An example showing how the cutout area is reset after the correction process An example of the data structure of character image information (dictionary data) An example of the data structure of the character recognition result Flowchart showing details of character recognition processing Flowchart of processing executed after the character recognition result is corrected Flowchart showing details of character detection processing Flowchart showing details of character recognition processing

(Example 1)
As an example of the information processing device according to the present embodiment, a mobile terminal (mobile terminal) will be described as an example. A mobile terminal is a terminal capable of communicating with the outside by using a wireless communication function or the like.

FIG. 1 is a diagram showing the appearance of the mobile terminal 100 (front surface 101 and back surface 103 of the mobile terminal) and a tire serving as a subject 105. The mobile terminal front portion 101 is provided with a touch panel display 102, and has two functions of display and touch operation input. The back surface 103 of the mobile terminal is provided with a camera unit 104 for photographing a subject and capturing an image. In the present embodiment, the user of the mobile terminal 100 can take a picture of the subject 105 and execute the character recognition process by using the mobile application (details will be described later) operated by the CPU of the mobile terminal. Subject 105 is an example of a tire. The captured image 106 can be acquired by photographing the portion where the serial ID (also referred to as the serial number or the serial number) of the tire is described by using the camera unit of the mobile terminal. The serial ID 107 indicates a serial number stamped on the tire, and is an ID for uniquely identifying the tire.

In the present embodiment, the subject 105 will be described by taking a tire as an example, but the subject is not limited to the tire. The mobile application described later can capture an image of the subject 105 and output the image to the touch panel 102.

FIG. 2 is a diagram showing an example of the hardware configuration of the mobile terminal 100. The CPU (Central Processing Unit) 201 is a processing unit that realizes various functions by executing various programs. The RAM (Random Access Memory) 202 is a unit used for storing various types of information and as a temporary working storage area of the CPU 201. The non-volatile memory (for example, ROM) 203 is a unit that stores various programs, data, and the like. The CPU 201 loads the program stored in the non-volatile memory 203 into the RAM 202 and executes the program. That is, the CPU (computer) of the mobile terminal functions as each processing unit as described with reference to FIG. 3 by executing the program, and executes each step of the sequence described later. The non-volatile memory 203 may be a flash memory, an HDD (Hard Disk Drive), an SSD (Solid State Disk), or the like. It should be noted that each function of the mobile terminal 100 and all or part of the processing related to the sequence described later may be realized by using dedicated hardware. The Input / Output interface 204 transmits / receives data to / from the touch panel 102. The NIC (Network Interface Card) 205 is a unit for connecting the mobile terminal 100 to a network (not shown). The camera interface 206 is connected to the camera unit 104 and captures an image of the subject 105 into the mobile terminal 100. The unit described above can send and receive data via the bus 207.

Next, the software configuration in the mobile terminal 100 will be described. FIG. 3 is a conceptual diagram showing an example of the software configuration of the mobile terminal 100. The CPU of the mobile terminal functions as each processing unit (each processing module) 303 to 308 by executing the mobile application (application program for the mobile terminal) 302. Further, the OS (Operating System) (not shown) of the mobile terminal 100 functions as the image management unit 301.

The image management unit 301 manages images and application data. The OS provides a control API (Application Programming Interface) for using the data management unit 301. Each application uses the control API to perform image and application data acquisition processing and storage processing on the data management unit 301.

The mobile application 302 is an application that can be executed by downloading and installing using the OS installation function of the mobile terminal 100. The mobile application 302 performs various data processing on the image of the subject 105 captured via the camera interface 206.

The main control unit 303 gives instructions and manages each module unit (303 to 308) described later.

The information display unit 304 controls the user interface (UI) of the mobile application 302 as shown in FIGS. 6 to 8 to be displayed on the touch panel in accordance with the instruction from the main control unit 303.

6 to 8 are diagrams showing an example of a screen (mobile terminal screen 600) of the UI (UI for a mobile terminal) of the mobile application 302. The mobile terminal screen 600 is displayed on the touch panel 102 of the mobile terminal 100. The mobile terminal screen 600 displays an image captured by the camera 104 in the area 601 and also accepts a user operation (user operation) on the image, UI, or the like. The shutter button 602 is a button for saving the image input from the camera unit in the RAM 202 or the data management unit 301, and hereinafter, the saved image will be referred to as a captured image. The zoom button 603 is a button for scaling the displayed image. Reference numerals 604 to 607 are guides that serve as a guide for the position at which the recognition target should be photographed. The user adjusts the shooting position so that the serial ID 107 to be recognized fits within the rectangular area surrounded by the four guides, and shoots the tire. 608 is a display area for displaying the character recognition result of the serial ID 107. When the recognition result is incorrect, the user touches the character to be corrected in the recognition result display area 608 to correct the recognition result. When the user touches the part of the character to be corrected in the recognition result display area 608 on the screen of FIG. 6, the correction candidate character for the touched character is displayed in the candidate character areas 701 to 703 as shown in the screen of FIG. Will be done. When any of the candidate character areas 701 to 703 is touched on the screen of FIG. 7, the characters in the recognition result display area 608 are updated to the selected candidate characters (the candidate character area 702 is touched on the screen of FIG. 8). An example is shown after the correction has been made).

The UI form (position, size, range, arrangement, display content, etc.) of the mobile application 302 is not limited to the form shown in the figure, and can realize the functions of the mobile terminal 100 as appropriate. The configuration can be adopted.

Each module will be described by returning to FIG. The operation information acquisition unit 305 acquires information related to the user operation performed on the UI of the mobile application, and notifies the main control unit 303 of the acquired information. For example, when the user touches the area 601 by hand, the operation information acquisition unit 305 senses the information on the touched position on the screen and transmits the information on the sensed position to the main control unit 303.

The image processing unit 306 performs image processing necessary for character recognition such as grayscale conversion, edge extraction, and feature amount extraction on the captured image of the subject 105 captured via the camera motor unit 206.

The character recognition unit 307 cuts out a plurality of areas in which characters are presumed to be described from the image processed by the image processing unit 306, compares the image of each area with the character image information (dictionary data) to be compared, and determines the image. Determine the most similar characters.

The character image management unit 308 manages character image information (information used as dictionary data of a so-called character recognition dictionary) used as a comparison target when the character recognition unit 307 recognizes characters. FIG. 4 is an example of character image information used as a comparison target when the character recognition unit 307 recognizes a character from an image. Character image information is prepared for each type of character used in the tire to be recognized. The character image information 401 to 410 is an example of a numerical image, but the serial ID 107 of the tire to be recognized in this embodiment includes a character image (not shown) of an uppercase alphabet in addition to the number.

The character image information (dictionary data) managed by the character image management unit may be characteristic information indicating the characteristics of each character created based on the font of the characters engraved on the tire. It may be the image itself of each character. What kind of dictionary data is used may be determined according to the algorithm used when collating the image to be recognized with the dictionary data.

FIG. 5 is a diagram illustrating the character recognition process in this embodiment. The recognition target image 501 is an image obtained by cutting out a part of the image of the subject 105 captured via the camera unit 104 and the camera interface 206. As described with reference to FIG. 6, the user adjusts the shooting position so that the serial ID 107 fits exactly in the guide (604 to 607 of FIG. 6) presented in the UI of the mobile application 302, and shoots the tire. The mobile application 302 cuts out an image of a portion surrounded by a guide from the captured image and sets it as a recognition target image 501.

The format of the tire serial ID 107 is determined for each manufacturer, and in this embodiment, the number of digits is 9 and the tire serial ID 107 is described as being composed of a number and an uppercase alphabet.

It is difficult to accurately detect the outline of characters in an image of a serial ID stamped on a tire because the contrast between the characters and the background is small and the surface of the tire (subject) is dirty. It is an image. Therefore, if the conventional technique of cutting out a character based on the outline of the character is applied, there is a high possibility that the character cutting position is erroneous, and as a result, the accuracy of the character recognition process deteriorates. Therefore, in the present invention, first, a region in which characters are considered to exist is set as a search region, and within the set search region, a plurality of cutout regions are set while changing the position and size, and a plurality of region images Repeat the cutting out. Then, each of the cut out plurality of region images is compared with the dictionary data (character image information to be compared, which is managed by the character image management unit), and the character recognition result for each region image and its similarity are obtained. After that, the character recognition result having the highest degree of similarity among those results and the cutout area used to obtain the character recognition result are set as the recognition result in the search area. Then, based on the position of the cutout area of the recognition result, the search area for the next character is set, and the same process is repeated. In this embodiment, the 9-digit serial ID 107 included in the recognition target image 501 is recognized in order from the first digit (the leftmost character).

The search area 502 for the first digit character is set at a position separated by a predetermined coordinate from the left end of the recognition target image 501 cut out based on the guides 604 to 607. The position of the first search area 502 is set in advance as an area in which there is a high possibility that the leftmost character exists when the image is taken so as to fit in the guide. Then, a cutout area 505 is set in the search area 502, an image of the cutout area 505 is extracted, compared with dictionary data relating to characters that may appear in the first digit, and each included in the dictionary data. Evaluate the similarity between letters. Further, the cutout area 505 is set at a plurality of positions in the search area 502 that are shifted in the horizontal direction (x-axis direction) and the vertical direction (y-axis direction), and dictionary data is obtained for the image of the cutout area at each position. Evaluate the similarity in comparison with. That is, the cutout area of a predetermined size is set at a plurality of places so as to cover the entire search area 502, and the image of the cutout area at each position is compared with the dictionary data. After that, the width and height of the cutout area 505 are further changed, the cutout areas are set at a plurality of places so as to cover the entire search area 502, and the image data is extracted and compared with the dictionary data. .. For example, when the width of the cutout area 505 is changed by 3 patterns and the height is changed by 2 patterns, the size of the cutout area 505 is 3 × 2 = 6 patterns in total. Further, when the cutout area 505 is set at each of the positions where the cutout area 505 is slid four times in the horizontal direction and four times in the vertical direction, the cutout area 505 is set to (4 + 1) × (4 + 1) = 25 places with respect to the search area 502. It will be set. Since the size of the cutout area is changed in 6 patterns and the setting positions are 25 places, the image of the cutout area is cut out from the search area 502 a total of 6 × 25 = 150 times. Then, each time the image is cut out, it is compared with the dictionary data (character image information of the character to be compared) of the character that may appear in the first digit, and the similarity to each character is evaluated.

Of the results of evaluating the images of all the cutout areas, the character with the highest similarity is determined as the recognition result of the first digit, and the position of the cutout area when the similarity is the highest is the first digit. The position of the character. In 504, "B" having the highest degree of similarity is determined as the recognition result of the first digit character, and indicates the cutout position at that time.

After that, the search area 503 of the next adjacent character (second character from the left) is set. The search area 503 is set at a relative position from the position 504 of the recognition result of the first digit. For the second digit character, as in the case of the first digit, a plurality of cutout areas 506 are set in the search area 503 and each of them is evaluated to determine the character having the highest degree of similarity. In the third and subsequent digits as well, the search area is set, the cutout area is set, and the similarity comparison with the dictionary data is sequentially performed to determine the characters of the recognition result.

It is desirable to make the search area 502 for the first digit character slightly wider in consideration of the fact that the characters are shifted to the left and right during shooting. On the other hand, since the space between characters is predetermined according to the character string of the subject, the search area 503 for the characters in the second and subsequent digits may be set narrower than the search area 502.

FIG. 10 is an example of a data structure of character image information (dictionary data) managed by the character image management unit 308. The character image information list includes a plurality of character image information. The character image information (dictionary data of the character recognition dictionary) includes character information (character code) of each character and feature information extracted from the character image of each character. As the feature information of each character, for example, HOG (Histograms of Oriented Gradients) feature amount may be used, but other feature amounts may be used.

FIG. 11 is an example of a data structure of result information of the character recognition process executed by the character recognition unit 307. The recognition result information includes a plurality of recognition result character information. The recognition result character information corresponds to the recognition result for each character, and includes a plurality of candidate character information for one recognition result character information. Since the serial ID 107 of this embodiment has 9 digits, the recognition result information includes 9 recognition result character information corresponding to each digit. Further, each candidate character information includes rectangular information (position and size of a cutout area corresponding to each candidate character), character information (character code), and an evaluation value. The evaluation value is a correlation coefficient (similarity) as a result of comparing the feature information of the character image information of FIG. 10 with the feature information extracted from the image cut out in the cutout region.

FIG. 12 is a flowchart showing details of the character recognition process executed by the character recognition unit 307 of the mobile application 302 after shooting the tire.

In step S1201, the character recognition unit 307 sets the search area for the first digit character (502 in FIG. 5) for the recognition target image 501 cut out from the captured image based on the guide.

In step S1202, the character recognition unit 307 cuts out an image in the search area.

In step S1203, the character recognition unit 307 sequentially sets the cutout area and compares the similarity with the dictionary data for the image of the cutout search area, and sets the position of the cutout area and the candidate characters at each position. (Character detection processing). The details of the process in step S1203 will be described later with reference to FIG.

In step S1204, it is determined whether or not the character is the last digit (9th digit), and if it is determined to be the character of the last digit, the process proceeds to step S1207. If it is determined that the character is not the last digit, the process proceeds to step S1205.

In TEP S1205, the character recognition unit 307 searches for the candidate character information having the highest evaluation value (similarity) from the recognition result character information shown in FIG. 11, and the rectangular information (the cutout area corresponding to the candidate character information). Position information) is acquired.

In step S1206, a search area for the next digit is set based on the rectangular information acquired in step S1205, and the process proceeds to step S1202.

In step S1207, the character recognition unit 307 displays the recognition result in the recognition result display area 608 of the screen via the information display unit 304, and ends.

FIG. 13 is a flowchart of processing executed by the mobile application 302 after correcting the character recognition result according to the user's instruction as described with reference to FIGS. 6 to 8.

In step S1301, the character recognition unit 307 calculates the sum of the evaluation values (highest evaluation value in the candidate character information) of each character from the next digit to the last digit of the corrected character.

In step S1302, the character recognition unit 307 searches for candidate character information in which the corrected character and the character information (character code) match from the recognition result character information corresponding to the character to be corrected.

In step S1303, the character recognition unit 307 acquires the rectangular information included in the candidate character information searched in S1302.

In step S1304, the character recognition unit 307 resets the search area of the next digit based on the rectangular information (position information of the cutout area corresponding to the corrected character) acquired in S1303.

In step S1305, the character recognition unit 307 cuts out an image in the search area.

In step S1306, the character recognition unit 307 sequentially sets the cutout area and compares the similarity with the dictionary data for the image of the cutout search area, and sets the position of the cutout area and the candidate characters at each position. (Character detection processing). The details of the process of step S1306 are the same as the process of S1203, and will be described later with reference to FIG.

In step S1307, the character recognition unit 307 determines whether or not the character is the last digit (9th digit), and if it is determined to be the last digit character, proceeds to step S1310 and determines that the character is not the last digit character. If so, the process proceeds to step S1308.

In step S1308, the character recognition unit 307 searches for the candidate character information having the highest evaluation value from the recognition result character information, and acquires the rectangle information. In step S1309, the search area for the next digit is set from the rectangular information acquired in step S1308, and the process proceeds to step S1305.

As described above, since the processing of S1305 to S1309 is re-executed based on the search area reset in S1304, the recognition result after the next digit of the corrected character may be different from the result in FIG. There is sex.

In step S1310, the character recognition unit 307 evaluates each character from the next digit to the last digit of the corrected character (highest evaluation among the candidate character information) based on the processing result of S1305 to S1309. Value) sum is calculated.

In step S1311, the character recognition unit 307 calculates the sum of the evaluation values (after executing the recognition processes S1305 to S1309 after the correction) in step S1310 from the sum of the evaluation values calculated in step S1301 (before the correction). Determine if it is high. If it is determined in step S1311 that the sum of the evaluation values calculated in step S1310 is higher, the process proceeds to step S1312. On the other hand, if it is determined that the sum of the evaluation values calculated in step S1310 is not higher, the process ends as it is.

In step S1312, the character recognition unit 307 uses the candidate characters obtained in the processing of S1305 to S1309 for the recognition result from the next digit to the last digit of the corrected character in the recognition result display area 608 of the screen. And update.

FIG. 14 is a flowchart showing details of the character detection process of step S1203 of FIG. 12 and step S1306 of FIG. In particular, it shows the details of the process of setting the cutout area at a plurality of positions in the search area and performing the recognition process while changing the size of the cutout area.

In step SS1401, the character recognition unit 307 sets the width of the cutout area (505, 506 in FIG. 5) to the minimum value, and proceeds to step S1402.

In step S1402, it is determined whether or not the width of the cutout region exceeds a predetermined maximum value. If it is determined that the width of the cutout area exceeds the maximum value, the process ends. If it is determined that the width of the cutout area does not exceed the maximum value, the height of the cutout area is set to the minimum value in step SS1403, and the process proceeds to step S1404.

In step S1404, it is determined whether or not the height of the cutout region exceeds a predetermined maximum value. If it is determined that the height of the cutout region exceeds the maximum value, the width of the cutout region is increased by a predetermined amount in step S1413, and the process proceeds to step S1402. If it is determined in step S1404 that the height of the cutout region does not exceed the maximum value, the process proceeds to step SS1405.

In step S1405, the character recognition unit 307 sets the x-coordinate of the left end of the cutout area to the initial value (x-coordinate of the left end of the search area), and proceeds to step S1406. In step S1406, it is determined whether or not the x-coordinate of the right end of the cutout area exceeds the x-coordinate of the right end of the search area. If it is determined that the x-coordinate at the right end of the cut-out area exceeds the x-coordinate at the right end of the search area, the height of the cut-out area is increased by a predetermined amount in step S1412, and the process proceeds to step S1404. If it is determined in step S1406 that the x-coordinate of the right end of the cut-out area does not exceed the x-coordinate of the right end of the search area, the y-coordinate of the upper end of the cut-out area is set to the initial value (in the search area) in step SS1407. It is set to the y coordinate of the upper end), and the process proceeds to step S1408.

In step S1408, it is determined whether or not the y-coordinate of the lower end of the cutout area exceeds the y-coordinate of the lower end of the search area. If it is determined that the y-coordinate of the lower end of the cut-out area exceeds the y-coordinate of the lower end of the search area, the cut-out area is slid in the x-axis direction (the x-coordinate is increased) in step S1411, and the process proceeds to step S1406. .. If it is determined in step S1408 that the y-coordinate of the lower end of the cut-out area does not exceed the y-coordinate of the lower end of the search area, the character image information (dictionary data) is added to the image of the cut-out area in step S1409. Comparison processing (character recognition processing) is performed. Details of the process in step S1409 will be described with reference to FIG. In step S1410, the cutout region is slid in the y-axis direction (y-coordinate is increased), and the process proceeds to step S1408.

FIG. 15 is a detailed flowchart of the character recognition process in step S1409 of FIG.

In step S1501, the character recognition unit 307 cuts out an image of the cutout region (505, 506 in FIG. 5), and in step S1502, the image processing unit 306 extracts feature information (HOG feature amount) from the cut out image. ..

In step S1503, the character recognition unit 307 acquires the character image information (dictionary data) at the beginning of the character image information list shown in FIG. In step S1504, the feature information included in the acquired character image information is compared with the feature information extracted in step S1502, and the correlation coefficient (similarity) is obtained as an evaluation value.

In step S1505, the character recognition unit 307 creates the candidate character information shown in FIG. 11 and sets the correlation coefficient of the comparison result in step S1504 as an evaluation value. At this time, the character information (character code) of the candidate character information is set to the character information of the character image information, and the rectangular information is set to the position and size of the cutout area.

In step S1506, the character recognition unit 307 searches for the candidate character information of the recognition result character information (shown in FIG. 11) with respect to the character of the digit being processed, and the candidate character information created in step S1505 and the character information are obtained. Determine if matching candidate character information already exists. If it is determined in step S1506 that there is no candidate character information that matches the character information, the process proceeds to step S1509. If it is determined that the candidate character information that matches the character information already exists, the process proceeds to step S1507.

In step S1507, the character recognition unit 307 determines whether or not the evaluation value of the candidate character information created in step S1505 is higher than the evaluation value of the candidate character information that already exists. If it is not determined that the candidate character information created in step S1505 is higher, the process proceeds to step S1510. If it is determined that the candidate character information created in step S1505 is higher, the process proceeds to step S1508, and the candidate character information already existing in the recognition result character information is deleted. Then, in step S1509, the candidate character information created in step S1505 is stored in the recognition result character information, and the process proceeds to step S1510.

In step S1510, it is determined whether or not it is the last of the character image information list, and if it is determined that it is not the last, in step S1511, the next character image information of the character image information list is acquired. If it is determined in step S1510 that the end of the character image information list is reached, the process ends.

FIG. 9 is a diagram showing an example in which the cutout area is reset after the character recognition result of this embodiment is corrected.

901 is an image of the serial ID 107 of the photographed tire. By executing the process of FIG. 12 on this image, it is assumed that the result as shown in 608 of FIG. 6 is obtained as the first character recognition result. It is assumed that the position of the cutout region corresponding to the first character recognition result 902 is the position of 903. After that, as described with reference to FIG. 7, when the third character from the left is corrected by the user's instruction, the rectangular area corresponding to the corrected character is searched, and the digits after the corrected character are searched. The search area is reset, the cutout area is set, and the recognition process is re-executed. Reference numeral 904 and 905 indicate a character recognition result as a result of executing the process described with reference to FIG. 13 and a cutout area thereof, respectively. As a result of erroneously determining the cutout area of the third digit character in the first cutout area 903, the cutout area of the fourth and subsequent digits is also invalid, so that the character recognition result is also incorrect. Then, when the user corrects the third digit, the process of FIG. 13 is executed, and as a result, the fourth and subsequent digits are also corrected in the corrected cutout area 905.

As described above, when the user corrects the recognition result, the search area is reset and the cutout area is set in the reset search area again for the characters after the corrected character. Then, the recognition result is corrected. Therefore, the misrecognition of the subsequent characters caused by the misrecognition of one character is also corrected, and the misrecognition of the subsequent characters is also corrected when the one character is corrected. Therefore, it is possible to reduce the burden on the user to correct the error in the recognition result. In addition, in order to compare the evaluation value before correction and the evaluation value after correction of all characters following the corrected character, and to reflect the recognition result after correction on the screen when the evaluation value after correction is high. , It also prevents the screen from being updated with a recognition result that is worse than the recognition result before correction.

(Other Examples)
The present invention supplies a program that realizes one or more functions of the above-described embodiment to a system or device via a network or storage medium, and one or more processors in the computer of the system or device reads and executes the program. It can also be realized by the processing to be performed. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

In order to solve the above-mentioned problem, the information processing apparatus of the present invention includes a first cutout area provided at a predetermined position in the search area and a cutout area provided at a position different from the predetermined position. A first setting means for setting a plurality of cutout areas including a first cutout area and a second cutout area having a size different from the first cutout area in the search area, and the plurality of pieces set by the first setting means. Image corresponding to each of the cutout areas of is extracted, candidate character information of each image is acquired by comparing each extracted image with dictionary data, and evaluation is performed from the candidate character information of each acquired image. It is characterized by having a character detecting means for outputting the candidate character information having the highest value as a recognition result.

Claims (5)

The first setting means for setting the search area for the image to be recognized, and
A second setting means for setting a cutout area at a plurality of places in the search area, and
By extracting an image corresponding to each of the plurality of cutout regions set by the second setting means and comparing each extracted image with dictionary data, the candidate character information and the candidate character are supported. It has a character detecting means that detects the position information of the cutout area and outputs the candidate character information having the highest evaluation value from the detected candidate character information as a recognition result.
The first setting means further sets a search area for the next character based on the position information of the cutout area corresponding to the recognition result output by the character detecting means, thereby setting the second setting. An information processing device characterized in that processing by the means and the character detecting means is repeatedly executed. A display means for displaying the recognition result output by the character detection means, and a display means for displaying the recognition result.
Based on the user's instruction, the correction means for executing the correction of the recognition result displayed by the display means, and the correction means.
An acquisition means for acquiring the position information of the cutout area corresponding to the character after the correction by the correction means, and
With more
Based on the position information acquired by the acquisition means, the first setting means resets the search area for the next character of the modified character, based on the reset search area. The information processing apparatus according to claim 1, wherein the processing by the second setting means and the character detecting means is executed. By resetting the search area for the character next to the corrected character, the result of processing by the second setting means and the character detecting means is executed based on the reset search area. The information processing apparatus according to claim 2, wherein the evaluation value is compared with the evaluation value before the correction to determine whether or not to correct the recognition result of the character next to the corrected character. A program for causing a computer to function as each means of the information processing apparatus according to any one of claims 1 to 3. The first setting step of setting the search area for the image to be recognized, and
A second setting step for setting a cutout area at a plurality of locations in the search area, and
By extracting the images corresponding to each of the plurality of cutout regions set in the second setting step and comparing each extracted image with the dictionary data, the candidate character information and the candidate character are supported. It has a character detection step that detects the position information of the cutout area and outputs the candidate character information having the highest evaluation value from the detected candidate character information as a recognition result.
In the first setting step, the second setting is further performed by setting a search area for the next character based on the position information of the cutout area corresponding to the recognition result output by the character detecting means. An information processing method characterized in that the processes in the step and the character detection step are repeatedly executed.

Images