JP2004005070A - Character recognition system and character recognition program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system capable of precisely recognizing a character displayed with a low resolution without being limited to an object having a character shown thereon. <P>SOLUTION: This character recognition system 1 comprises a digital camera 4 for taking an image of a character; and a recognition device 2 having an input means 7 for inputting the taken image, an imaging information acquiring means 9 for acquiring imaging information 8, a pattern generating means 11 for generating a correction pattern 10 suitable to the taken image, a storage means 13 for storing character definition data 12 or the like, an emphasis correcting means 15 for performing the emphasis correction of the character by contrast correction, brightness correction, sharpness correction and negating correction of the taken image, a pattern detecting means 16 for detecting a designated pattern 5, a range selecting means 18 for selecting a range image 17, a range recognizing means 19 for performing a character recognition, and an output means 21 for outputting the recognized character. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a character recognition system and a character recognition program, and more particularly, to a character recognition system and a character recognition program that perform character recognition after capturing a character to be recognized as an image, and enhancing and correcting the captured image. It is about.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, systems capable of reading characters and symbols printed on a print medium such as paper as electronic data using an optical device have been developed. For example, a so-called “POS system” that manages point-of-sale information by reading a barcode displayed or affixed to the packaging of a product with an optical barcode reader in order to simplify product management and accounting processing of stores and the like. There are places to adopt. Further, as a use other than product management, a member card distributed to a member registered in advance at a rental video store or the like is provided with the barcode described above, and the information of the barcode is read at the time of rental, thereby being a registered member. Make sure that.
[0003]
In addition, a document manuscript is obtained as an image using a scanner equipped with an image pickup device such as a CCD (Charge Coupled Device), and characters are automatically input on a computer using OCR software (optical character reading software). The recognition process is being performed. These are capable of rapidly and electronically converting a large number of characters displayed on the surface of a paper medium or the like into electronic data, and are widely used in government offices and offices where large amounts of materials and documents need to be digitized. are doing.
[0004]
In addition, for example, a license plate of a passing vehicle installed on a highway or a general national road is photographed using an infrared camera, and a vehicle registration number (number) is automatically recognized from an image input as an image. A vehicle number automatic reading system called a so-called "N system" is known. With this, when a vehicle used for criminal activity or a vehicle used by a criminal to escape has passed the road where the N system is installed, the number of the vehicle concerned is recognized, and the position of each vehicle is recognized at each police station. Information can be reported, and criminal arrests can be made promptly.
[0005]
[Problems to be solved by the invention]
However, the system for reading and recognizing information using the barcode described above has the following problems. That is, an information recognition system using a barcode reads the line thickness, interval, and arrangement of the barcode by the reflection of the laser beam emitted from the barcode reader, and associates the barcode with the prestored data. It was intended to recognize information such as price and product number. In this case, in order to avoid erroneous recognition, it is necessary to display the bar code line thickness and interval precisely.
[0006]
For this reason, a barcode with a relatively low resolution displayed on a monitor screen, a liquid crystal screen, or the like may cause erroneous recognition or may take a long time for the recognition process. Therefore, in order to solve such a problem, measures such as sticking a sticker on which a barcode is printed at a high resolution to a product have been taken, but extra costs such as printing cost for printing the sticker are required. Was. In addition, in the barcode printed on the sticker, if the content of the information to be recognized is changed, it is necessary to prepare a new sticker, and the already printed sticker is wasted, and the time required for printing , It was not possible to respond quickly to the change.
[0007]
On the other hand, a character recognition system using a scanner or the like can easily recognize a character displayed on a flat object such as a document manuscript, but causes the document manuscript to be read in close contact with the contact glass. Therefore, it is difficult to recognize a character displayed on a three-dimensional object having irregularities. Further, it sometimes takes a long time to input characters as images by a scanner and to perform character recognition processing using OCR software. For this reason, an operator who is not good at computers sometimes takes time to operate and cannot perform character recognition in a short time.
[0008]
On the other hand, the N system is a system lacking in versatility specialized in recognizing a vehicle number, and it has been difficult to apply the system to general uses.
[0009]
In view of the above circumstances, the present invention provides a highly versatile character recognition system and a character recognition program capable of recognizing characters displayed at a low resolution with high accuracy without being limited to recognition targets. It is an issue.
[0010]
[Means for Solving the Problems]
In order to solve the above-mentioned problem, a character recognition system according to the present invention comprises a storage unit for storing character definition data defining a character to be recognized, and an imaging unit for imaging the character and the vicinity of the character. And enhancement correction means for performing at least one of contrast correction, brightness correction, sharpness correction, and negative correction on a captured image including the character captured by the imaging means, The image processing apparatus includes a character recognizing unit that detects an image that matches the character definition data from the captured image and performs character recognition, and an output unit that outputs the recognized character.
[0011]
Here, the character definition data defines characters, including numbers, alphabets (upper / lowercase), hiragana, katakana, symbols, kanji, etc., to be recognized in advance by defining their shapes, outlines, intersections of lines, and the like. It is a database and is basic data for recognizing an image in a captured image as a character. If the recognition target is known in advance (for example, a combination of numerals and English characters), the type and number of characters corresponding to the stored character definition data may be limited. In addition, it may include data corresponding to the typeface of each character.
[0012]
Further, the imaging means can use an image input device such as a video camera or a digital camera, and can capture a character and the vicinity of the character as a captured image. Here, in the captured image, a character to be recognized and the vicinity of the character are captured.
[0013]
At this time, the character may not be clearly shown in the captured image due to the reflection of light at the time of image capturing or the difference in shading between the character and the surroundings, which may hinder the subsequent processing related to character recognition.
[0014]
Therefore, correction for emphasizing the character portion included in the captured image so that it can be easily recognized is performed by the emphasis correction unit. Specifically, at least one of image processing among contrast correction, brightness correction, sharpness correction, and negative correction is performed on a captured image. Here, the contrast correction adjusts the contrast of the captured image, such as lightness and darkness, and emphasizes the color and contrast between the character to be recognized and its surroundings, and clarifies the difference between the character and the surroundings. It is. The brightness correction is to adjust the brightness (brightness) of a captured image, to adjust the brightness of the entire captured image, and to display characters clearly. On the other hand, the sharpness correction is for adjusting the sharpness of a captured image, and for sharpening a difference between character boundary portions and the like. In addition, the negative correction is a process of inverting the gradation of the captured image, that is, inverting the image from a positive image to a negative image, so that a character portion in the captured image becomes prominent.
[0015]
Therefore, according to the character recognition system of the first aspect of the present invention, a character portion is emphasized by an enhancement correction such as a contrast correction on an image captured by the imaging unit, and character recognition is performed based on the character definition data. That is, the captured image is corrected by emphasizing the character portion, and after clarifying the outline and shape of the image portion corresponding to the character included in the captured image, and performing character recognition, the captured image is captured. Compared with the case where the character is recognized as it is, the recognition accuracy of the character is increased, and the possibility that the character is erroneously recognized is reduced. It should be noted that the output means includes not only those output on the monitor screen but also those transmitted through the network as character information.
[0016]
A character recognition system according to a second aspect of the present invention is the character recognition system according to the first aspect, wherein the character recognition unit detects a designated graphic designated in advance from the emphasized corrected captured image. A range selecting means for selecting a range surrounded by the detected designated figure as a range image; and a range for detecting an image matching the character definition data from the selected range image and performing character recognition. And recognition means.
[0017]
Here, the designated graphic is selected in advance that exists in the captured image and that is already known that the character to be recognized is within the range surrounded by the designated graphic. For example, when character recognition is to be performed on a vehicle number, a horizontally long rectangular shape of a license plate is selected as a designated graphic, and character recognition processing is performed only in a range surrounded by the rectangle. Note that, as the designated figure, various figures such as a circle, an ellipse, a triangle, and a rhombus can be selected in addition to the above-described rectangle.
[0018]
Therefore, according to the character recognition system of the second aspect of the present invention, in addition to the operation of the character recognition system of the first aspect of the present invention, the designated figure is detected from the captured image, and Performs character recognition. As a result, the target range of character recognition is limited, the possibility of erroneous character recognition is reduced, and the processing load on the system is reduced.
[0019]
The character recognition system according to a third aspect of the present invention is the character recognition system according to the first or second aspect, wherein the emphasis correction unit includes the contrast correction, the brightness correction, the sharpness correction, and the negative correction. Is performed in accordance with the order of the above.
[0020]
Therefore, according to the character recognition system of the third aspect of the present invention, in addition to the operation of the character recognition system of the first or second aspect of the present invention, enhancement correction of a captured image is performed by contrast correction, brightness correction, sharpness correction, and negative correction. The correction is performed in the order. Here, the order of each correction is determined based on the magnitude of the influence on the captured image after correction.
[0021]
That is, if the contrast correction is performed strongly, the entire captured image tends to be too black or vice versa. That is, the change in the captured image is greatly affected. Therefore, if the contrast correction is performed after the sharpness correction is performed first, the already performed sharpness correction may be completely meaningless due to the influence of the contrast correction. Further, since the negative correction is to invert the gradation of the captured image from positive to negative, even if the brightness or the sharpness is corrected after the grayscale inversion, the character is hardly emphasized. Therefore, the order of correction to the captured image is determined in advance in accordance with the magnitude of the influence of the correction, and the correction of the captured image can be performed appropriately by performing the correction in accordance with the order.
[0022]
A character recognition system according to a fourth aspect of the present invention is the character recognition system according to any one of the first to third aspects, wherein the character image and the captured image including the vicinity of the character are captured. Imaging information acquisition means for acquiring imaging information based on the image information; pattern generation means for generating and storing a correction pattern based on the imaging information; and a pattern for performing enhancement correction of the captured image using the generated correction pattern. And application means.
[0023]
Here, the captured image has different color tone, saturation, and the like depending on the imaging conditions and the captured object. For this reason, even when the emphasis correction is performed under certain conditions, there are cases where characters are emphasized and corrected, and cases where characters are hardly emphasized. Therefore, the state of the captured image, and more specifically, the degree of projection of light and shadow based on the conditions at the time of image capture, or the color tone and color type of characters and parts other than characters (difference between similar colors and different colors) ), Etc., as well as various kinds of information such as the material of the object on which the character is displayed, the degree of reflection on the surface, etc., are acquired as imaging information, and a correction pattern based on the information is applied, so that enhancement correction suitable for each captured image is obtained. Is done.
[0024]
Therefore, according to the character recognition system of the fourth aspect, in addition to the operation of the character recognition system of any one of the first to third aspects, imaging information is obtained from an imaging image to be enhanced and corrected. And a correction pattern based thereon is applied. For example, when the captured image has a dark color tone as a whole, the contrast and the brightness are adjusted to be stronger, and an image with increased brightness is performed. Conversely, in the case of a captured image in which the whole is expressed in a bright color tone, the correction is performed with the brightness being slightly suppressed. In this case, the intensity of the contrast correction and the intensity of the brightness correction are determined depending on the brightness of each captured image.
[0025]
In addition, when the character and the part other than the character are represented by similar colors, in other words, when the character is easily confused around the character and it is difficult to recognize the character, the sharpness correction strength is increased. The emphasis correction is performed so that the boundary (outline) of the character becomes clear. That is, by performing a correction suitable for each captured image based on the captured information and the correction pattern, characters included in the captured image are further emphasized, and subsequent character recognition processing can be performed easily.
[0026]
A character recognition system according to a fifth aspect of the present invention is the character recognition system according to the fourth aspect, wherein at least one of the imaging information acquisition unit, the pattern generation unit, and the pattern application unit is capable of learning. A neural network is applied.
[0027]
Therefore, in the character recognition system according to the fifth aspect of the invention, in addition to the operation of the character recognition system according to the fourth aspect of the present invention, a neural network is applied to the enhancement correction. In other words, by using a neural network having a learning function, the enhancement correction of the captured image is optimized based on the learning content that is previously learned and stored, and is further learned by generating imaging information and a correction pattern. You. In other words, by performing a number of processes related to the enhancement correction, the neural network with a higher learning effect makes the enhancement correction for the captured image more optimal, thereby shortening the processing time and improving the recognition accuracy. Will be.
[0028]
7. The character recognition program according to claim 6, wherein the storage unit stores character definition data defining a character to be recognized, an imaging unit that captures an image of the character and the vicinity of the character, and the image that is captured by the imaging unit. Enhancement correction means for performing at least any one of contrast correction, brightness correction, sharpness correction, and negative processing correction on a captured image including characters, and the character definition data from the emphasized corrected captured image The computer functions as a character recognizing unit that detects an image that matches the character string and performs character recognition and an output unit that outputs the recognized character.
[0029]
Therefore, according to the character recognition program of the invention of claim 6, when the program is executed, the computer has the effect of claim 1.
[0030]
A character recognition program according to a seventh aspect of the present invention is the character recognition program according to the sixth aspect, wherein the figure detection means detects a designated figure designated in advance from the emphasized-corrected picked-up image. Range selection means for selecting a range surrounded by a figure as a range image; detecting, from the selected range image, an image matching the character definition data; and performing range recognition on the computer as range recognition means for character recognition. To make it work.
[0031]
Therefore, when the character recognition program according to the seventh aspect of the present invention executes the program in addition to the action of the character recognition program according to the sixth aspect of the invention, the computer performs the action according to the second aspect.
[0032]
The character recognition program according to an eighth aspect of the present invention is the character recognition program according to the sixth or seventh aspect, wherein the imaging information based on the imaging image captured including the character and the vicinity of the character is stored. The computer as an imaging information acquisition unit to acquire, a pattern generation unit to generate and store a correction pattern based on the imaging information, and a pattern application unit to perform enhancement correction of the captured image using the generated correction pattern. Is to function further.
[0033]
Therefore, when the character recognition program of the invention of claim 8 executes the program in addition to the operation of the character recognition program of the invention of claim 6 or 7, the computer performs the operation of claim 4.
[0034]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, a character recognition system 1 according to an embodiment of the present invention will be described with reference to FIGS. FIG. 1 is an explanatory diagram showing a configuration of a character recognition system 1 according to an embodiment of the present invention, FIG. 2 is a block diagram showing a functional configuration of the character recognition system 1, and FIGS. 6 is a flowchart illustrating a flow of a second process.
[0035]
As shown in FIG. 1, the character recognition system 1 of the present embodiment converts a character L consisting of a number shown on the surface of a recognition target (here, a license plate N of a vehicle C) into a vicinity of the character L. And a digital camera 4 that captures the captured image 3 as a captured image 3, and captures the captured image 3 captured by the digital camera 4 and processes the captured image 3 in a state suitable for character recognition by performing enhancement correction, detection of a designated figure 5, and the like. And a recognition device 2 that performs character recognition and outputs and displays a recognition result. Here, the digital camera 4 corresponds to an imaging unit in the present invention.
[0036]
Further, as shown in FIG. 2, the recognition device 2 is connected to a digital camera 4 via a cable 6 as a functional configuration, and an input unit 7 capable of inputting a captured image 3 captured by the digital camera 4. An imaging information acquisition unit 9 for acquiring imaging information 8 for performing appropriate enhancement correction from a captured image 3 by applying a neural network capable of learning, a correction pattern corresponding to the imaging information 8 and suitable for the imaging image 3 The apparatus includes a pattern generation unit 11 that generates a character 10 by applying a neural network, character definition data 12 that defines a character L to be recognized, and a storage unit 13 that stores the generated correction pattern 10.
[0037]
Further, the recognition device 2 applies a pattern learning unit 14 that applies a previously learned or generated correction pattern 10 to enhancement correction, and performs image processing of contrast correction, brightness correction, sharpness correction, and negative correction on the captured image 3. An emphasis correction unit 15 that performs emphasis correction for emphasizing a character L included in the captured image 3; a graphic detection unit 16 that detects a designated graphic 5 specified in advance from the corrected captured image 3; Range selection means 18 for selecting an enclosed range as a range image 17; range recognition for detecting an image (part) matching the stored character definition data 12 from the range image 17 and performing character recognition on the image; Means 19, and an output control means 20 and an output means 21 for signal-controlling and outputting the recognized character L. . Here, the figure detecting means 16, the range selecting means 18, and the range recognizing means 19 correspond to the character recognizing means in the present invention.
[0038]
Here, a commercially available personal computer is used as the recognition device 2, and among the components in this embodiment, a USB (universal serial bus) as the input means 7, a hard disk drive as the storage means 13, a monitor as the output means 21, Graphic boards are used as the output control means 20, respectively. Each unit for performing enhancement correction, graphic detection, and detection and selection of the range image 17 on the captured image 3 is controlled by hardware such as a CPU 22, a ROM 23, and a RAM 24 mounted on the recognition device 2, respectively. Have been. Here, a recognition program 27 for performing character recognition is stored in the ROM 23, and a command is issued to the CPU 22 and the like to execute processing relating to recognition. On the other hand, the data of the designated figure 5 detected by the figure detecting means 16 is stored in the RAM 24 in advance. Further, a keyboard 25 and a mouse 26 for performing operations of the recognition device 2 and imaging operations of the digital camera 4 are attached. In addition, as shown in FIG. 1, a monitor screen corresponding to the output unit 21 includes a captured image display unit 28 that displays a captured image 3 captured by the digital camera 4 and a character display unit that displays recognized characters L. 29.
[0039]
Next, a processing flow of the recognition device 2 in the character recognition system 1 of the present embodiment will be described with reference to FIGS. In the present embodiment, a description will be given of a case where a number displayed on a license plate N of a vehicle C is recognized as a character L to be recognized. Here, steps S1 to S16 in FIGS. 3 and 4 are executed by the recognition program 27 stored in the ROM 23. Here, the recognition program 27 corresponds to a character recognition program in the present invention.
[0040]
First, the recognition device 2 detects whether or not the captured image 3 captured by the connected digital camera 4 has been input via the input unit 7 (Step S1). At this time, the detection of the timing at which the traffic of the vehicle C is detected and the image is captured is determined by applying a detection means such as a conventionally known sensor technology. Further, the focus of the digital camera 4 and the size of the captured image 3 (portion surrounded by a broken line on the right side of the paper in FIG. 1) include the vicinity of the license plate N attached to the lower part of the passing vehicle C. The adjustment has been performed in advance.
[0041]
When the input of the captured image 3 from the digital camera 4 is detected (YES in step S1), the recognition device 2 receives the input of the captured image 3 (step S2). On the other hand, when the input of the captured image 3 is not detected (NO in step S1), the subsequent processing is canceled, and the process proceeds to step S15. The recognition device 2 that has received the input of the captured image 3 acquires the imaging information 8 from the captured image 3 (Step S3).
[0042]
Here, the imaging information 8 is a basis for extracting or generating a correction pattern 10 applied at the time of the enhancement correction performed on the captured image 3 in the subsequent steps. Various information such as the degree of shadow projection, the color tones of characters and parts other than characters, and the number and types of colors included are acquired. Thereby, the correction pattern 10 to be applied is selected.
[0043]
At this time, it is determined whether or not the correction pattern 10 corresponding to the acquired imaging information 8 has already been generated (step S4). In other words, those corresponding to the experience of performing the enhancement correction of the captured image 3 captured under the same conditions, the arrangement of the characters L to be recognized, the color tone, and the like, or the imaging information 8 previously acquired through learning and storage. In some cases, for example, a four-digit number (letter L) with a green letter L on a white background, such as a license plate N of a vehicle C shown in the present embodiment (in the case of an ordinary car) is targeted. In this case, if there is experience in performing character recognition on the license plate N before, the correction pattern 10 for the imaging information 8 of the license plate N has already been generated and stored. Furthermore, it is recognized from the imaging information 8 acquired by the learning effect of the neural network that the captured image 3 includes the license plate N. Therefore, when the corresponding correction pattern 10 has been generated (YES in step S4), the correction pattern 10 stored in the storage unit 13 is extracted and the enhancement correction of the captured image 3 is performed (step S5).
[0044]
On the other hand, when the corresponding correction pattern 10 has not been generated, that is, when there is no experience of executing the captured image 3 having the similar imaging information 8 (NO in step S4), the correction pattern suitable for the imaging information 8 10 is newly generated (step S6) and stored in the storage unit 13 (step S7).
[0045]
Here, the generation of the correction pattern 10 will be specifically described. Based on the acquired imaging information 8, contrast correction, brightness correction, sharpness correction, and negative correction performed in the subsequent steps (details are described in detail). The degree of strength is determined by the pattern generation unit 11. For example, when the captured image 3 includes the above-described license plate N, the green portion of the character L and the white portion serving as the background By contrast, the strength of the contrast is determined. In this case, since the color tone is different between green and white, the contrast between the character and the background is clear, and it is determined that the contrast correction is performed to a weak degree.
[0046]
Further, the intensity of the brightness correction is determined based on the overall brightness of the captured image 3. For example, when the captured image 3 is entirely composed of a bright color tone, that is, when the entire image is a whitish image, correction for lowering (darkening) the brightness is performed. Also, if the image 3 captured by the digital camera 4 is blurred, the boundary between green and white may not be clear. Therefore, the boundary between green and white, that is, the boundary between green and white, may be considered. The intensity of the sharpness correction is determined according to the sharpness of the image. In addition, since the negative correction is for inverting the gradation, it is not possible to add strength such as contrast correction, and therefore, the correction pattern 10 includes only a determination as to whether or not to perform the negative processing. ing. As a result, the correction pattern 10 in which the intensity of the contrast correction, the brightness correction, and the sharpness correction suitable for the captured image 3 and whether or not to perform the negative correction is determined based on the imaging information 8 is generated.
[0047]
As a result, a correction pattern 10 suitable for the captured image 3 is generated and can be used for enhancement correction. In addition, when the enhancement of the captured image 3 having similar imaging information 8 is subsequently performed by the learning effect of the neural network, the correction pattern 10 is newly added. The process of generating the correction pattern 10 can be omitted. Thereafter, the process proceeds to step S5, in which the captured image 3 is subjected to enhancement correction using the generated correction pattern 10.
[0048]
Then, in step S5, enhancement correction of the captured image 3 to which the correction pattern 10 is applied is performed. At this time, the degree of prominence such as lightness and darkness in the captured image 3 is adjusted based on the strength of each correction included in the applied correction pattern 10 (contrast correction: step S5a), and brightness and luminance are adjusted (brightness). Correction: step S5b), adjustment of sharpness (sharpness correction: step S5c), and inversion of the tone of the captured image 3 from positive to negative (negative correction: step S5d) are sequentially performed.
[0049]
Here, the order of each correction is determined based on the magnitude of the influence on the captured image 3. That is, since the contrast correction S5a has a large effect on the captured image 3, the effect of the correction (for example, the sharpness correction) performed before that may be lost. Therefore, by taking into account the magnitude of the effect on the captured image 3 and performing the contrast correction, the brightness correction, the sharpness correction, and the negative correction in this order, a more appropriate correction can be made while utilizing the effect of the enhancement correction performed in the previous stage. Will be able to do. As a result, the corrected captured image 3 that has been finally emphasized and corrected has the character L clearly emphasized compared to the captured image 3 before correction.
[0050]
Then, the designated figure 5 set in advance is detected from the corrected captured image 3 (step S8). Here, in the present embodiment, a horizontally long rectangle which is the shape of the license plate N attached to the vehicle C is selected as the designated graphic 5. That is, since the range surrounded by the rectangle of the license plate N always includes the four-digit number (character L) of the vehicle number to be recognized, the range of the designated figure 5 is selected by Recognition processing of other parts can be omitted. As a result, the recognition accuracy of the character L can be increased, and the processing time for character recognition can be reduced.
[0051]
When the designated figure 5 is detected from the captured image 3 (YES in step S8), a range surrounded by the detected designated figure 5 is selected as the range image 17 (step S9). On the other hand, if the designated figure 5 is not detected from the captured image 3 (NO in step S8), it is recognized that the character L to be recognized does not exist in the captured image 3, and the subsequent processing is canceled. The process proceeds to S15. Here, the range image 17 and the designated figure 5 have the same shape and size (see FIG. 1).
[0052]
Then, the presence or absence of an image having conditions such as a shape and an outline matching the character definition data 12 stored in the storage means 13 is detected from the selected range image 17 (step S10). Here, when an image of a condition matching the character definition data 12 is detected (YES in step S10), the image at the corresponding location is recognized as a character L (step S11). On the other hand, if there is no image satisfying the condition that matches the character definition data 12 (NO in step S10), it is determined that the character L does not exist in the range image 17 surrounded by the designated graphic 5, and the subsequent processing is cancelled. Then, the process proceeds to step S15.
[0053]
Then, after recognizing the character L, the presence or absence of an image of a condition that matches the character definition data in the range image 17 is detected in addition to the recognized character L (step S12). At this time, if an image satisfying the condition that matches the character definition data 12 is not detected (YES in step S12), the character recognition process ends (step S13), and the process proceeds to the next step.
[0054]
On the other hand, when an image with a condition that matches the character definition data 12 is further detected (NO in step S12), the process returns to the previous step and the corresponding image is recognized as a character L (step S11). Thus, even when a plurality of characters L exist in the range image 17, all the characters L can be recognized by repeating Steps S11 and S12. The order of character recognition is not particularly limited, but when the character to be recognized is written horizontally as in English, that is, when written from left to right, character recognition is performed sequentially from the left, and when written vertically, What is necessary is just to set so that character recognition may be performed sequentially from top to bottom.
[0055]
Then, after recognizing all the characters included in the range image 17, one or a plurality of recognized characters L are signal-controlled by the output control means 20 and displayed on the monitor screen as the output means 21 ( Step S14). Thereby, the operator of the character recognition system 1 can check the result of the character recognition on the screen. The recognized character L can be output to another system as information and used. For example, when the number of the license plate N of the vehicle C described above is recognized, the system is applied to a system for managing parking lot exits, and the parking time is calculated. The present invention can be applied to a calculation system and the like.
[0056]
Thereafter, the presence or absence of an instruction to end the processing of the system is detected (step S15). Here, when an instruction to end is input (YES in step S15), the system is ended (step S16). On the other hand, when there is no input of an instruction to end the process (NO in step S15), the process returns to step S1, and the presence or absence of the input of a new captured image 3 is continuously detected.
[0057]
As described above, according to the character recognition system 1 of the present embodiment, the captured image 3 including the character L and the vicinity of the character L is captured by the digital camera 4 and a learnable neural network is applied. By generating a correction pattern 10 optimized for enhancement correction of the captured image 3 and applying the correction pattern 10 to perform enhancement correction, the captured image is corrected to a state suitable for character recognition, and Can be emphasized in the captured image 3. As a result, character recognition with little erroneous recognition and high recognition accuracy can be performed.
[0058]
In addition, since the target for character recognition is limited to the range image 17 surrounded by the designated figure 5 specified in advance, the accuracy of character recognition can be improved and the processing time for recognition can be shortened. Further, it is not necessary to print at a high resolution unlike a conventional barcode, and characters on a monitor screen or a liquid crystal screen output and displayed at a relatively low resolution can be easily recognized. Therefore, it is possible to save physical costs and labor such as sticking and printing the barcode. Even if the character L is changed, it is only necessary to change the character L displayed on the screen, so that the cost and time for printing a new sticker are not required, and a prompt response can be taken.
[0059]
As described above, the present invention has been described with reference to the preferred embodiments. However, the present invention is not limited to these embodiments, and various improvements can be made without departing from the gist of the present invention as described below. And design changes are possible.
[0060]
That is, in the character recognition system 1 of the present embodiment, an example is described in which the number indicated on the license plate N of the vehicle C is recognized as the character L to be recognized. However, the present invention is not limited to this. In addition, characters L written on printed matter and various characters L displayed on the screen can be recognized. Further, although the digital camera 4 is used as the image pickup means, another image input device such as a video camera may be used. Further, the recognized character L can be transmitted as character information through a network or the like.
[0061]
【The invention's effect】
As described above, the character recognition system according to the first aspect of the invention can correct the contrast or the like of the captured image including the character, and can emphasize the included character. Then, a character recognition process is performed on the captured image in which the characters are emphasized. As a result, it is possible to perform character recognition with higher accuracy than before.
[0062]
According to a second aspect of the present invention, in addition to the effects of the first aspect of the present invention, the character recognition system according to the first aspect of the present invention detects a designated figure specified in advance, and performs character recognition within a range surrounded by the detected designated figure. Be done. Thereby, the range to be recognized is limited, and the burden on the recognition process is reduced. Therefore, for example, the display of the recognition result can be shortened.
[0063]
According to a third aspect of the present invention, in addition to the effects of the first or second aspect of the present invention, a character recognition system enhances a captured image in the order of contrast correction, brightness correction, sharpness correction, and negative correction. The character can be further emphasized while utilizing the effect of the correction that has been performed. As a result, the characters included in the corrected captured image become clearer, and the accuracy of character recognition can be improved.
[0064]
According to the character recognition system of the fourth aspect, in addition to the effect of the character recognition system of any one of the first to third aspects, a correction pattern is generated based on imaging information of a captured image. That is, enhancement correction optimized for each captured image is performed based on the correction pattern. Thereby, the characters in the captured image are clearly shown, and the character recognition is more easily performed.
[0065]
A character recognition system according to a fifth aspect of the present invention has the effect of the character recognition system according to the fourth aspect of the present invention, and further includes a correction that is learned in advance or obtained in a character recognition process by applying a learnable neural network. The enhancement correction of the captured image can be optimized based on the pattern. As a result, the recognition accuracy is improved, and the recognition processing time is reduced.
[0066]
According to the character recognition program of the invention of claim 6, by executing the program, it is possible to cause a computer to exert the effects described in claim 1.
[0067]
A character recognition program according to a seventh aspect of the present invention can cause a computer to achieve the effects described in the second aspect by executing the program in addition to the effects of the character recognition program according to the sixth aspect of the present invention.
[0068]
The character recognition program according to the eighth aspect of the present invention can cause the computer to achieve the effects described in the fourth aspect by executing the program in addition to the effects of the character recognition program according to the sixth or seventh aspect of the present invention. it can.
[Brief description of the drawings]
FIG. 1 is an explanatory diagram illustrating a configuration of a character recognition system according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating a functional configuration of a recognition device in the character recognition system.
FIG. 3 is a flowchart illustrating a flow of processing of the recognition device.
FIG. 4 is a flowchart showing a flow of processing of the recognition device.
[Explanation of symbols]
1 Character recognition system
2 Recognition device
3 Imaging image
4 Digital camera (imaging means)
5 Designated figure
8 Imaging information
9 Imaging information acquisition means
10 Correction pattern
11 Pattern generation means
12 Character definition data
13 Storage means
14 Pattern application means
15 Emphasis correction means
16 Figure detection means (character recognition means)
17 Range image
18 Range selection means (character recognition means)
19 Range recognition means (character recognition means)
20 Output control means
21 Output means
27 Recognition program (character recognition program)
L letter

Claims (8)

Storage means for storing character definition data defining a character to be recognized;
Imaging means for imaging the character and surroundings of the character;
For a captured image including the character captured by the image capturing unit, contrast correction, brightness correction, sharpness correction, and enhancement correction unit that performs at least one of the negative correction,
A character recognition unit that detects an image that matches the character definition data from the captured image that has been enhanced and corrected, and performs character recognition.
Output means for outputting the recognized character. The character recognition means,
A figure detecting means for detecting a designated figure designated in advance from the emphasized corrected captured image,
Range selection means for selecting a range surrounded by the detected designated figure as a range image,
2. The character recognition system according to claim 1, further comprising: a range recognition unit that detects an image matching the character definition data from the selected range image and performs character recognition. The emphasis correction means,
3. The character recognition system according to claim 1, wherein the contrast correction, the brightness correction, the sharpness correction, and the negative correction are performed in the order of the enhancement of the captured image. Imaging information obtaining means for obtaining imaging information based on the image captured including the vicinity of the character and the surroundings of the character,
A pattern generation unit that generates and stores a correction pattern based on the imaging information;
4. The character recognition system according to claim 1, further comprising a pattern application unit configured to perform enhancement correction of the captured image using the generated correction pattern. 5. The character recognition system according to claim 4, wherein at least one of the imaging information acquisition unit, the pattern generation unit, and the pattern application unit employs a learnable neural network. Storage means for storing character definition data defining a character to be recognized; imaging means for imaging the character and the vicinity of the character; contrast correction and brightness for an image including the character imaged by the imaging means Enhancement means for performing at least any one of correction, sharpness correction, and negative processing correction, and detects an image matching the character definition data from the emphasized corrected captured image and performs character recognition. A character recognition program for causing a computer to function as character recognition means and output means for outputting the recognized characters. A figure detecting means for detecting a designated figure designated in advance from the emphasized corrected picked-up image; a range selecting means for selecting a range surrounded by the detected designated figure as a range image; The computer-readable storage medium according to claim 6, wherein the computer further functions as a range recognition unit that detects an image that matches the character definition data and performs character recognition. Imaging information acquisition means for acquiring imaging information based on the character and the captured image including the vicinity of the character; pattern generation means for generating and storing a correction pattern based on the imaging information; The computer-readable storage medium according to claim 6, wherein the computer further functions as a pattern application unit that performs the enhancement correction of the captured image using the correction pattern.

Images