CN112766248B

CN112766248B - Structured prescription picture identification method and device

Info

Publication number: CN112766248B
Application number: CN202110381096.0A
Authority: CN
Inventors: 赵鑫; 焦小斌
Original assignee: Suzhou Iron Technology Co Ltd
Current assignee: Suzhou Iron Technology Co Ltd
Priority date: 2021-04-09
Filing date: 2021-04-09
Publication date: 2021-07-09
Anticipated expiration: 2041-04-09
Also published as: CN112766248A

Abstract

The invention discloses a structured prescription picture identification method and a structured prescription picture identification device, wherein the method comprises the following steps: acquiring a prescription picture to be entered; carrying out area marking and cutting on a prescription picture to be input to obtain a plurality of first area images to be identified; classifying the first to-be-recognized area image according to the relevance of the area mark; splicing the first to-be-identified area image according to the category to obtain a second to-be-identified area image; performing character recognition on the second to-be-recognized area image; and binding the text information of the second to-be-identified area image with the position corresponding to the electronic document according to a preset matching mode. By splicing the prescription image mark areas again, the recognition accuracy and the recognition integrity are improved, so that the places of the whole prescription, which need to be recognized, are not easily recognized, the places, which do not need to be recognized, are not easily recognized, and the automation degree and the processing efficiency of the prescription image recognition and entry are improved.

Description

Structured prescription picture identification method and device

Technical Field

The invention relates to the technical field of image processing, in particular to a structured prescription picture identification method and device.

Background

At present, prescriptions are entered in two ways, one is manually entered, and the other is to identify the prescriptions by a text visual Recognition technology, namely an Optical Character Recognition (OCR) technology, and then enter the prescriptions automatically.

However, for OCR, only characters are recognized, and subsequent entry still requires manual processing and accurate information matching, so that the degree of automation is not high enough, and the processing efficiency is low.

Disclosure of Invention

In view of this, embodiments of the present invention provide a structured prescription image recognition method and apparatus, so as to solve the problem in the prior art that the efficiency of OCR recognition for automatic entry of a prescription image is not high enough.

The embodiment of the invention provides a structured prescription picture identification method, which comprises the following steps:

acquiring a prescription picture to be entered;

carrying out area marking and cutting on a prescription picture to be input to obtain a plurality of first area images to be identified;

classifying the first to-be-recognized area image according to the relevance of the area mark;

splicing the first to-be-identified area image according to the category to obtain a second to-be-identified area image;

performing character recognition on the second to-be-recognized area image;

and binding the text information of the second to-be-identified area image with the position corresponding to the electronic document according to a preset matching mode.

Optionally, the preset matching pattern is:

when the text data contained in the second to-be-recognized area image is smaller than a preset value, matching is carried out according to the intersection ratio of the second to-be-recognized area image and the first template area;

if the ratio of the cross-over ratio exceeds a preset threshold value, judging that the matching is successful;

when the text data contained in the second area image to be recognized is larger than a preset value, judging whether the second area image to be recognized is contained in a marking area in the second template area;

and if the second area image to be identified is contained in the marked area in the second template area, judging that the matching is successful.

Optionally, the preset threshold is 60%.

Optionally, before performing region marking and cutting on the picture of the prescription to be entered to obtain a plurality of region images to be identified, the method further includes:

and performing perspective and correction on the input prescription picture according to the template prescription to align the content of the input prescription picture with the template prescription.

Optionally, the method further comprises: marking the template prescription, and endowing character attributes corresponding to character information at different positions; wherein the text attribute includes at least one of a hospital name, a department name, patient information, and a diagnosis result.

Optionally, the stitching the multiple images of the area to be identified according to the relevance of the area markers includes:

the images of the areas to be identified are arranged in lines, so that each line only has one continuous identification area.

Optionally, the step of stitching the first to-be-recognized region image according to the category to obtain a second to-be-recognized region image further includes: and filling the right side, the lower side and the right lower side area of the second to-be-recognized area image with an image with a pixel value of 255, so that the proportion of the number of pixels of a single to-be-recognized character in the second to-be-recognized area image to the total number of pixels is reduced, and the resolution of the to-be-recognized character is improved to 300 dpi-400 dpi.

In the present embodiment, it is preferred that,

optionally, the preset value is twenty characters.

The embodiment of the invention also provides a structured prescription picture recognition device, which comprises:

the template marking module is used for marking the template prescription and endowing the text attributes corresponding to the text information at different positions;

the image processing module is used for carrying out area marking and cutting on the picture of the prescription to be input to obtain a plurality of first area images to be identified; the image processing module is also used for splicing a plurality of first to-be-identified area images according to the relevance of the area mark of the prescription of the template to obtain a plurality of second to-be-identified area images;

the character recognition module is used for carrying out character recognition on the second to-be-recognized area image;

and the template matching module is used for binding the text information of the second to-be-identified area image with the corresponding meaning of the electronic document according to the position information and/or the identified text content.

The embodiment of the invention provides a structured prescription picture identification method, which has the following beneficial effects:

1. by splicing the prescription image mark areas again, the recognition accuracy and the recognition integrity are improved, so that the places of the whole prescription, which need to be recognized, are not easily recognized, the places, which do not need to be recognized, are not easily recognized, and the automation degree and the processing efficiency of the prescription image recognition and entry are improved.

2. Through template matching, matching accuracy is improved, recognized characters can be accurately matched with represented meanings, the phenomena of wrong matching, missing matching and multiple matching are obviously reduced, secondary processing of recognized prescriptions by people is obviously reduced, and working efficiency is greatly improved.

Drawings

The features and advantages of the present invention will be more clearly understood by reference to the accompanying drawings, which are illustrative and not to be construed as limiting the invention in any way, and in which:

FIG. 1 shows a flow diagram of a structured prescription picture identification method;

FIG. 2 shows a block diagram of a structured prescription picture recognition apparatus;

fig. 3 shows a block diagram of a structured prescription picture recognition terminal.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Example 1

As shown in fig. 1, an embodiment of the present invention provides a structured prescription picture recognition method, including:

and step S10, acquiring a picture of the prescription to be entered.

In this embodiment, a preprocessing step is further included to reduce the noise of the image and remove the image noise for the subsequent operation.

And step S20, carrying out area marking and cutting on the prescription pictures to be entered to obtain a plurality of first area images to be identified.

In this embodiment, the cutting method is implemented by extracting a minimum area quadrangle containing prescription information in the picture to be entered: finding out two points on the image contour containing the prescription information, and marking the two points as a point A and a point B; connecting point A and point B to form a first line segment; searching a point C farthest from the first line segment and a point D farthest from the first line segment on the contour; the point a, the point D, the point B, and the point C are connected in this order to form a minimum area quadrangle ADBC. Polygonal approximation of the contour refers to: a polygon is used to approximate a contour. The purpose of polygon approximation is to reduce the number of vertices of the contour. The result of the polygon approximation is still a contour, but this contour is relatively coarse.

In step S30, the first to-be-recognized region image is classified according to the relevance of the region label.

In the present embodiment, the degree of relevance ranks the partitions according to the setting of template matching.

Step S40, stitching the first to-be-recognized area images according to the categories to obtain second to-be-recognized area images.

In the present embodiment, the second to-be-recognized region image is composed of at least one first to-be-recognized region image of the same category. And splicing the plurality of smaller first to-be-identified area images according to the category to obtain a plurality of larger second to-be-identified area images. In the subsequent character recognition process, a template with a fixed size is used for recognition, and the more pixels contained in the image, the higher the resolution, that is, the clearer the image, the easier the character is to recognize.

In step S50, character recognition is performed on the second to-be-recognized region image.

In the present embodiment, character recognition is performed by OCR.

And step S60, binding the text information of the second to-be-recognized area image with the position corresponding to the electronic document according to the preset matching mode.

In this embodiment, each recognition area obtained after OCR character recognition is matched with the position of each mark block in the image newly generated by the image processing module, and if the matching is successful, the character content of each recognized area is bound with the corresponding meaning of the area. In a specific embodiment, the data bound by the template matching module is formatted into a json format commonly used in network communication, so that the data can be processed, processed and transmitted conveniently.

According to the method and the device, the identification accuracy and integrity are improved by splicing the prescription image marking areas again, so that the places of the whole prescription, which need to be identified, are not easily identified, the places, which do not need to be identified, are not easily identified, and the automation degree and the processing efficiency of the prescription image identification and entry are improved.

As an optional implementation, the preset matching mode is:

if the ratio of the cross-over ratio exceeds a preset threshold value, judging that the matching is successful; specifically, the preset threshold is 60%.

In this embodiment, if the marked area to be recognized is smaller, matching is performed through the intersection ratio of the second area image to be recognized and the first template area, so that the effect of no recognition is achieved. In a specific embodiment, the predetermined value is twenty characters. The setting can be carried out according to actual needs.

In this embodiment, if the mark area to be recognized is large, for example, the text content of the diagnosis result is large, the content of the recognition area is not fixed, and the number of characters is large in a floating manner, the matching accuracy is improved by comparing whether the recognition area is included in the mark area or not, and if the recognition area is completely included, the matching is successful, so that the recognized text and the represented meaning can be accurately matched, and the phenomena of wrong matching, missing matching and multiple matching are remarkably reduced.

As an optional implementation manner, before step S20, the method further includes:

and step S11, perspective and correction are carried out on the entered prescription picture according to the template prescription, so that the content of the entered prescription picture is aligned with the template prescription.

In this embodiment, the template recipe is generated as follows: and storing the position information and the meaning name of each marking area, and generating a corresponding template file. If the template exists in the current prescription image to be recognized, the area to be recognized can be cut out quickly through perspective and correction.

As an optional implementation manner, step S11 further includes: marking the template prescription, and endowing character attributes corresponding to character information at different positions; wherein the text attribute includes at least one of a hospital name, a department name, patient information, and a diagnosis result.

In the embodiment, the template prescription is marked and divided in advance, so that the area to be identified is conveniently cut.

As an alternative embodiment, step S40 includes:

in step S41, the plurality of first to-be-recognized region images are arranged in lines such that each line has only one continuous recognition region.

In the embodiment, the recombined to-be-identified region images list the to-be-identified regions in a row, so that the situation that front and back information of a plurality of to-be-identified regions is mixed is avoided.

As an optional implementation manner, step S40 further includes: and filling the right side, the lower side and the right lower side area of the second to-be-recognized area image with an image with a pixel value of 255, so that the proportion of the number of pixels of a single to-be-recognized character in the second to-be-recognized area image to the total number of pixels is reduced, and the resolution of the to-be-recognized character is improved to 300 dpi-400 dpi.

Assuming that the whole prescription picture is taken as a character recognition object, the resolution ratio is 300dpi, which just meets the requirement of optical character recognition. If the prescription picture is cut and identified, the proportion of characters on the picture is too large, the picture pixels are reduced, and the character identification precision is affected, so in the embodiment, the picture pixels are increased by filling the pixels around the image of the area to be identified, so that the picture pixels meet the requirement of character optical identification, namely, the resolution reaches more than 300 dpi. In addition, since the position of the character to be recognized needs to be normalized in advance in the optical character recognition, pixels are filled in the right side, the lower side and the right lower side of the image of the region to be recognized, so that the characters are concentrated in the upper left corner region, and the position normalization is realized in advance.

Example 2

As shown in fig. 2, an embodiment of the present invention further provides a structured prescription picture recognition apparatus, including: template mark module, image processing module, character recognition module and template matching module, wherein: the template marking module is used for marking the template prescription and endowing the text attributes corresponding to the text information at different positions; the image processing module is used for carrying out area marking and cutting on the picture of the prescription to be input to obtain a plurality of first area images to be identified; the image processing module is also used for splicing a plurality of first to-be-identified area images according to the relevance of the area mark of the prescription of the template to obtain a plurality of second to-be-identified area images; the character recognition module is used for carrying out character recognition on the second to-be-recognized area image; and the template matching module is used for binding the text information of the second to-be-identified area image with the corresponding meaning of the electronic document according to the position information and/or the identified text content.

In this embodiment, the template marking module is configured to mark the identification area and obtain a location and a name of the marked area; the image processing module is used for perspective and correction of the picture, marking area segmentation, splicing of the segmented areas and generation of a new image to be identified; the character recognition module is used for realizing OCR character recognition and acquiring a recognition area and character information; the template matching module is used for matching the marking area with the identification area. The device also comprises a data processing module which is used for Json matching data, returning the data and storing the data.

An embodiment of the present invention further provides a structured prescription picture recognition terminal, as shown in fig. 3, the recognition terminal may include a processor 31 and a memory 32, where the processor 31 and the memory 32 may be connected by a bus or in another manner, and fig. 3 illustrates an example of connection by a bus.

The processor 31 may be a Central Processing Unit (CPU). The Processor 31 may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or combinations thereof.

The memory 32, which is a non-transitory computer readable storage medium, may be used to store non-transitory software programs, non-transitory computer executable programs, and modules. The processor 31 executes various functional applications and data processing of the processor by running non-transitory software programs, instructions and modules stored in the memory 32, namely, implements the structured prescription picture recognition method in the above method embodiment.

The memory 32 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created by the processor 31, and the like. Further, the memory 32 may include high speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 32 may optionally include memory located remotely from the processor 31, and these remote memories may be connected to the processor 31 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The one or more modules are stored in the memory 32 and, when executed by the processor 31, perform a structured prescription picture recognition method as in the embodiment shown in fig. 1-2.

The specific details of the structured prescription image recognition terminal can be understood by referring to the corresponding related descriptions and effects in the embodiments shown in fig. 1 to fig. 2, which are not described herein again.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk (Hard Disk Drive, abbreviated as HDD), a Solid State Drive (SSD), or the like; the storage medium may also comprise a combination of memories of the kind described above.

Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A structured prescription picture recognition method, comprising:

acquiring a prescription picture to be entered;

carrying out area marking and cutting on the picture of the prescription to be input by extracting a minimum area quadrangle containing prescription information in the picture to be input to obtain a plurality of first area images to be identified;

classifying the first to-be-identified region image according to the relevance of the region mark;

splicing the first to-be-identified area images according to categories, arranging a plurality of to-be-identified area images in rows, and enabling each row to have only one continuous identification area to obtain a second to-be-identified area image;

performing character recognition on the second to-be-recognized area image;

2. The method of claim 1, wherein the predetermined matching pattern is:

if the ratio of the intersection ratio exceeds a preset threshold value, judging that the matching is successful;

when the text data contained in the second area image to be recognized is larger than the preset value, judging whether the second area image to be recognized is contained in a marking area in a second template area;

3. The structured prescription picture recognition method of claim 1, wherein the preset threshold is 60%.

4. The structured prescription picture recognition method of claim 1, wherein before performing region marking and cutting on the prescription picture to be entered to obtain a plurality of region images to be recognized, the method further comprises:

and performing perspective and correction on the input prescription picture according to a template prescription to align the content of the input prescription picture with the template prescription.

5. The structured prescription picture recognition method of claim 4, further comprising: marking the template prescription, and endowing character attributes corresponding to character information at different positions; wherein the textual attributes include at least one of a hospital name, a department name, patient information, and a diagnosis result.

6. The method for identifying the structured prescription picture as claimed in claim 1, wherein stitching the first to-be-identified region image according to the category to obtain a second to-be-identified region image further comprises: and filling the right side, the lower side and the right lower side area of the second to-be-recognized area image with an image with a pixel value of 255, so that the proportion of the number of pixels of a single to-be-recognized character in the second to-be-recognized area image to the total number of pixels is reduced, and the resolution of the to-be-recognized character is improved to 300 dpi-400 dpi.

7. The structured prescription picture recognition method of claim 2, wherein the preset value is twenty characters.

8. A structured prescription picture recognition apparatus, comprising:

the image processing module is used for carrying out area marking and cutting on the picture of the prescription to be input to obtain a plurality of first area images to be identified; the image processing module is further used for splicing a plurality of first to-be-identified area images according to the relevance of the template prescription area mark to obtain a plurality of second to-be-identified area images;