TWI612479B - Character image recognition system and method for recognizing character image - Google Patents

Abstract

A text image recognition system includes a word database, an image capture device, a display screen, and a processor. The word database stores the first word data set and the second word data set. The display screen displays a plurality of instant preview images according to the image data captured by the image capturing device. The processor causes the display screen to mark the target block in the plurality of instant preview images, selects the target image from the plurality of instant preview images, and identifies the text image corresponding to the target block in the target image according to the first word data group. And identifying, according to the second word data group, the text image corresponding to the target block in the target image. The first word data group and the second word data group are different word data groups.

Description

Text image recognition system and method for recognizing text images

The invention relates to a character image recognition system, in particular to a character image recognition system capable of recognizing a text image of a different block in an image by using a different word data group.

In general, in order to be able to manage files efficiently, it is a common and important process to visualize documents and electronically digitize the images. In the prior art, when the text in the image is recognized, the image features obtained after analyzing the image content are compared and recognized by the word features stored in the general word library, and the words stored in the general word library are stored. Words may contain a variety of commonly used words or even proper nouns in various fields. However, since the word data in the general word dictionary does not correspond to the image content, directly comparing the image features obtained by analyzing the image with the words in the general word database may not only consume unnecessary computing resources, but also consume unnecessary computing resources. It may not be able to achieve satisfactory results.

In addition, in many applications, the content of the image may be a standard form file, and different fields in the form often record different types of information. For example, envelopes or packages often use different fields to record and post. Passing relevant information, such as the recipient field and the recipient address field, will record the name and address information separately. However, the words commonly used in these two kinds of information are quite different. For example, the words commonly found in address information may include "XX City", "XX Area", "XX Road", etc., and these words. Words are generally less common in names. In this case, if you use the words recorded in the general-purpose word library to recognize the text of different fields in the table, it is easy to waste unnecessary computing resources in the unrelated words, and may also The misjudgment is caused, and after the identification result is obtained, the correction is also required by manual means, which causes inconvenience in use. Therefore, how to effectively recognize the text in the image becomes a problem to be solved.

An embodiment of the present invention provides a character image recognition system. The character image recognition system includes a word database, an image capture device, a display screen, and a processor.

The word database stores the first word data group and the second word data group, and the first word data group and the second word data group are different word data groups. The image capturing device can capture the target image. The display screen is coupled to the image capturing device and can display the target image. The processor is coupled to the image capturing device, the display screen and the word database, and the processor can mark the target screen by the display screen in the target image, and identify the target image corresponding to the target block according to the first word data group. The text image inside, and the character image corresponding to the target block in the target image is identified according to the second word data group.

Another embodiment of the present invention provides a method for recognizing a text image. The method for recognizing a text image includes an image capturing device capturing a target image, displaying a screen display target image, and causing the display screen to mark the target block in the target image. The processor selects the target image from the plurality of instant preview images, and the processor identifies the text image content corresponding to the target block in the target image according to the first word data group, and the processor identifies the target according to the second word data group. The image corresponds to the text image content outside the target block. The first word data group and the second word data group are different word data groups.

FIG. 1 is a schematic diagram of a character image recognition system 100 according to an embodiment of the present invention. The character image recognition system 100 includes a word database 110, an image capturing device 120, a display screen 130, and a processor 140. The display screen 130 is coupled to the image capturing device 120. The processor 140 is coupled to the image capturing device 120, the display screen 130, and the word database 110.

The image capturing device 120 can capture image data, and the display screen 130 can display a plurality of instant preview images according to the image data captured by the image capturing device 120, and the processor 140 can display the display screen 130 on the display screen. The target block is marked in the live preview image. In other words, the display screen 130 can instantly display the image captured by the image capturing device 120, and mark the target block of the instant preview image on the display screen.

FIG. 2 is a display screen of the display screen 130 according to an embodiment of the present invention. In the second figure, the display screen 130 displays the instant preview image IMG1 captured by the image capturing device 120, and according to the control of the processor 140, the display screen 130 also displays the boundary of the target block T1 in the instant preview image IMG1. The target block T1 is marked. The target block T1 is displayed in the display screen 130 to occupy a predetermined ratio, and the predetermined ratio may be 30 to 50%, wherein 40% is the optimal ratio. The shape of the target block T1 is preferably rectangular, but there is no limitation to match the customary address of the recipient, and a variety of habits can be provided for the user to select, and the selected shape becomes preset after selection. .

In some embodiments of the present invention, the display screen 130 can be a touch screen. The processor 140 can mark the target tile T1 in the instant preview image IMG1 according to the touch operation of the user on the display screen 130, and adjust the target area according to the touch operation of the user on the display screen. The size of block T1. For example, the user can click on the display screen 140 to display the boundary of the marked target block T1, adjust the position of the target block T1 by dragging, or can be closed by the finger on the display screen 140 or Extending the size of the target block T1, of course, in other embodiments of the present invention, the user may control the processor 140 through different touch modes, and enable the processor 140 to adjust the target block T1 correspondingly. Position and/or adjust the size of the target block T1.

In this way, the user can align the target block T1 marked on the display screen 130 with a specific area of the image to be processed, and enable the target block T1 to contain a specific area of the image to be processed. For example, in FIG. 2, the content of the instant preview image IMG1 is the postal data form on the package, and the specific area of the instant preview image IMG1 may be the recipient address field A1 in the postal data form. Since in the second figure, the target block T1 does not completely contain the address field A1 of the instant preview image IMG1, the user may need to continuously adjust the angle of the image capturing device 120 to preview the image in real time, and may also It is necessary to adjust the size of the target block T1 by the above method. Since the target block T1 can completely contain the address field A1 of the instant preview image IMG1 and does not contain information other than the address field A1 (for example, a trademark pattern), unnecessary waste is not wasted. Computational resources are not easy to cause misjudgment in terms that are not related to the comparison.

After the user adjusts the image capturing device 120 and the display screen 130, and the target block T1 can include the content of the recipient address field A1 in the postal data form retrieved by the image capturing device 120, the user The processor 140 can be controlled to select the instant preview image displayed on the display screen 130 at this time as the target image for character image recognition. In other words, the user can dynamically adjust the position of the image capturing device 120 and the size of the target block T1 according to the live preview screen displayed on the display screen 130, and after the adjustment is completed, cause the processor 140 to perform the function of capturing the image. At this time, the image obtained by the shooting is the target image. In some embodiments of the present invention, the processor 140 may also store the selected target image into the memory.

In addition, in other embodiments of the present invention, the user may further adjust the position and/or size of the target block T1 through the processor 140 after extracting the appropriate target image TIMG1.

FIG. 3 is a target image TIMG1 selected by the processor 140 according to an embodiment of the present invention. In the target image TIMG1, the target block T1 will contain a specific block to be processed, that is, for example, the aforementioned recipient address field A1. The image content processor 140 corresponding to the target image TMG1 corresponding to the image content in the target block T1 and the target image TIMG1 corresponding to the target block T1 can be identified by using different word data sets respectively.

For example, after the user's adjustment, the content of the target image TIMG1 corresponding to the target block T1 should already contain the data of the address field A1 of the recipient. At this time, the processor 140 may identify the destination address corresponding to the target block T1 in the target image TIMG1 according to the plurality of word data related to the address, and identify the target image TIMG1 according to a plurality of word data commonly used by the standard. The other mail related material corresponding to the target block T1, such as the recipient name or the company line number.

In some embodiments of the present invention, the word database 110 may store the first word data group 112 and the second word data group 114, and the first word data group and the second word data group are different. The word data group. In the above embodiment, the first word data group 112 may be a plurality of word materials related to the address, and the second word data group 114 may be a plurality of word materials commonly used by the standard, and the processor 140 The character image corresponding to the target block T1 in the target image TIMG1 may be identified according to the first word data group 112, and the text image corresponding to the target block T1 in the target image TIMG1 may be identified according to the second word data group 114.

In some embodiments of the present invention, the processor 140 may perform image processing on the image corresponding to the target block T1 in the target image TIMG1, for example, by using edge detection to separate the characters in the image. Then, the image features of each character are calculated. For example, the number of boundaries of each character, the number of intersections of the boundary and the boundary, and the angle between the boundary and the boundary can be calculated as the image features of each character.

In this case, the first word data group 112 may include image feature data of a plurality of words related to the address obtained according to the above manner, such as "Taipei", "Kaohsiung", "Yonghe", "Yongkang" The image features of the words "Zhongxiao East Road" and "Zhonghua Road", so when the processor 140 recognizes the text image corresponding to the target block T1 in the target image TIMG1 according to the first word data set 112, By comparing the character image features corresponding to the target block T1 in the target image TIMG1 with the image features of other types of words, the computing resources can be reduced. In addition, when the character image feature corresponding to the target image block T1 in the target image TIMG1 is compared with the word image feature in the first word data group 112, a plurality of possible candidate words may be compared, for example, "He" and "Taiwan", if the next character image is recognized and the possible candidates are "North" and "Mega", the first word data group 112 has the word "Taipei". Therefore, the weights of "Taiwan" and "North" will be improved. At this time, the processor 140 may preferentially select the words with higher weights, that is, "Taiwan" and "North". Since the words in the first word data group 112 are all related to the address, when selecting possible candidate words, the character image recognition system 100 can accurately determine the text image in the target block T1 and reduce the misjudgment. The case of other categories of words.

In other embodiments of the present invention, if a particular community wants to utilize the text image recognition system 100 to process the mail and package received by the particular community, the first word data set 112 may also include only that particular The plural word data associated with the household address of the community, or a plurality of serial number data associated with the household address of the particular community. For example, if the address of a particular community is from 50 to 54 Dong Yi Street and 22 to 24 North 2nd Road, and the floor of the household is from 1st to 10th floor, the first word data set 112 may contain " Image features of words such as "One Street", "North Second Road", "50", "52", "54", "22", "24", "3rd Floor" and "5th Floor"... It will include other road names or numbers that are not related to that particular community, such as "60", "100" or "Zhongxiao East Road"...etc. Even when the address of a particular community is limited to a single road segment, for example, when the address is from No. 50 to No. 54 of Dongyi Street, the first word data group 112 may only contain a plurality of serial numbers related to the household address of the specific community. Image data such as "50", "52" and "54". In this way, when the processor 140 compares the character image in the target block T1 with the words in the first word data group 112 to adjust the weight, it is possible to more accurately determine the target block T1. What is the text image and reduce the misjudgment into other categories of words.

In some embodiments of the present invention, since the text image corresponding to the target block T1 in the target image TIMG1 has no specific content, the second word data group can be a plurality of words commonly used in the standard, and the data processing The device 140 can identify the text image corresponding to the target block T1 in the target image TIMG1 according to the second word data group 114.

In addition, the present invention is not limited to the first word data group 112 being a plurality of word materials related to an address and the second word data group 114 is a plurality of word materials commonly used as standards, and the other in the present invention. In an embodiment, the specific block that the user wants to process may also contain other content, and the first word data group 112 may also be the word data related to the content. For example, if the specific block to be processed by the user is the patient's medical record data, the first word data set 112 may be a plurality of word data related to the disease, and if the specific block is external to the patient, For the data, the second word data group 114 can still be a plurality of word materials commonly used in the standard, or a plurality of word materials related to general basic materials such as name, blood type, height, etc., so that the data can be efficient. And accurately identify the text in the target image.

FIG. 4 is a flow chart of a method 200 for recognizing a text image according to an embodiment of the present invention. Method 200 can include, but is not limited to, steps S210 through S260, and can be applied to text image recognition system 100.

S210: The image capturing device 120 captures the target image TIMG1;

S220: The display screen 130 displays the target image TIMG1;

S230: The processor 140 causes the display screen 130 to mark the target block T1 in the target image TIMG1;

S240: The processor 140 identifies, according to the first word data group 112, the text image content corresponding to the target image block T1 in the target image TIMG1;

S250: The processor 140 identifies, according to the second word data group 114, the target image TIMG1 corresponding to the text image content outside the target block T1.

In step S210, the image capturing device 120 captures the target image TIMG1, and in step S220, the display screen 130 displays the target image TIMG1 captured by the image capturing device 120.

Next, in step S230, the processor may cause the display screen 130 to mark the target block T1 in the target image TIMG1. In some embodiments of the present invention, the shape of the target block may be preset to be narrow and long to match a specific block in the instant preview image, such as the recipient address field A1 in the postal data form. In addition, in some embodiments of the present invention, the display screen 130 may be a touch screen. In this case, the method 200 may further include adjusting the size of the target block T1 according to a touch operation of the user on the display screen 130. The step is to facilitate the user to adjust the target block T1 and enable the target block T1 to include a specific block in the target image TIMG1.

In some embodiments of the present invention, the display screen 130 can also display a plurality of instant preview images according to the image data captured by the image capturing device 120, and the user can adjust according to the instant preview image displayed on the display screen 130. The image capture device 120 captures the angle and/or the size and/or position of the target block T1, and after the adjustment is completed, causes the processor 140 to perform a function of capturing an image, and the captured image is the target image. Then, in steps S240 and S250, the processor can identify, according to the first word data group 112 and the second word data group 114, the text image content corresponding to the target image TIMG1 in the target block T1, respectively. The word data included in the word data group 112 is related to the text content in the target block T1 corresponding to the target image TIMG1, so the method 200 can accurately determine the text image in the target block T1 and reduce false positives. In the case of other categories of words.

In summary, the text image recognition system and the method for recognizing text images provided by the embodiments of the present invention allow the user to mark the target block for the user to align when the target image to be recognized is captured. Preview a specific area in the image, and after determining the selected target image, identify the image content in the target image corresponding to the inside and outside of the target block by using different word data groups. Since the words in the word data group can be limited to be related to the content in the target block, the character image recognition system and the method for recognizing the character image provided by the embodiments of the present invention can improve the recognition accuracy of the text image, and It is possible to reduce the case where a character image is mistakenly judged as a word unrelated to the video content. The above are only the preferred embodiments of the present invention, and all changes and modifications made to the scope of the present invention should be within the scope of the present invention.

100 text image recognition system 110 word database 112 first word data group 114 second word data group 120 image capturing device 130 display screen 140 processor IMG1 instant preview image A1 recipient address field T1 target block TIMG1 Target image 200 method S210 to S250 steps

1 is a schematic diagram of a character image recognition system according to an embodiment of the present invention. FIG. 2 is a display screen of a display screen according to an embodiment of the present invention. FIG. 3 is a target image selected by a processor according to an embodiment of the present invention. FIG. 4 is a flow chart of a method for recognizing a character image according to an embodiment of the present invention.

200 Method S210 to S250 Steps

Claims (15)

A text image recognition system includes: a word database for storing a first word data group and a second word data group; an image capturing device for capturing a target image; and a display screen The image capturing device is coupled to the image capturing device for displaying the target image; and a processor coupled to the image capturing device, the display screen and the word database, the processor is configured to: display the display screen Marking a target block in the target image; identifying a text image corresponding to the target block in the target image according to the first word data group; and identifying the target image according to the second word data group Corresponding to the text image outside the target block; wherein the first word data group and the second word data group are different word data groups; and the first word data group includes a specific community The plurality of word data related to the household address, and the processor identifies, according to the first word data group, a receiving address in the target image corresponding to the target block. The character image recognition system of claim 1, wherein the first word data group further comprises a plurality of word materials related to the address. The character image recognition system of claim 1, wherein the first word data group further comprises a plurality of serial number data related to a household address of a specific community. The character image recognition system of claim 1, wherein the second word data group is a standard And a plurality of word data that is commonly used, and the processor identifies, according to the second word data group, a mail related material corresponding to the target block in the target image. The character image recognition system of any one of claims 1 to 4, wherein the display screen is a touch screen. The character image recognition system of claim 5, wherein the processor adjusts the position of the target block according to a touch operation of the display screen by the user. The character image recognition system of claim 5, wherein the processor is further configured to adjust a size of the target block according to a touch operation of the display screen by the user. The character image recognition system of any one of claims 1 to 4, wherein the processor is further configured to store the target image to a memory. A method for recognizing a text image, comprising: an image capturing device capturing a target image; a display screen displaying the target image; and a processor causing the display screen to mark a target block in the target image; the processor Identifying, according to a first word data group, text image content corresponding to the target image in the target image; and the processor identifying, according to a second word data group, the target image corresponding to the text outside the target block Image content; wherein the first word data group and the second word data group are different word data groups; and The first word data group includes a plurality of word data related to a household address of a specific community, and the processor identifies, according to the first word data group, the target image corresponding to the target block. A receiving address. The method of claim 9, wherein the first word data set further comprises a plurality of word materials associated with the address. The method of claim 9, wherein the first word data set further comprises a plurality of serial number data associated with a household address of a particular community. The method of claim 9, wherein the second word data group is a plurality of word materials commonly used by the standard, and the processor identifies the target image corresponding to the target according to the second word data group. A mail-related material outside the block. The method of any one of claims 9 to 12, wherein the display screen is a touch screen; and the processor adjusts the position of the target block according to a touch operation of the display screen by the user. The method of any one of claims 9 to 12, wherein: the display screen is a touch screen; and the method further comprises the processor adjusting the target block according to a touch operation of the user on the display screen the size of. The method of any one of claims 9 to 12, further comprising the processor storing the target image Save to a memory.