CN113762274A - Answer sheet target area detection method, system, storage medium and equipment - Google Patents
Answer sheet target area detection method, system, storage medium and equipment Download PDFInfo
- Publication number
- CN113762274A CN113762274A CN202111323174.8A CN202111323174A CN113762274A CN 113762274 A CN113762274 A CN 113762274A CN 202111323174 A CN202111323174 A CN 202111323174A CN 113762274 A CN113762274 A CN 113762274A
- Authority
- CN
- China
- Prior art keywords
- frame
- answer sheet
- area
- identification
- missed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method, a system, a storage medium and equipment for detecting a target area of an answer sheet, wherein the method comprises the following steps: acquiring and preprocessing an answer sheet picture to be detected, dividing the preprocessed answer sheet picture to be detected to obtain a subarea of the answer sheet to be detected, acquiring frame information of the subarea, and judging whether the subarea has a missing identification frame; and if the missed identification frame exists, acquiring and completing the frame of the missed identification frame. According to the answer sheet target area detection method, the system, the storage medium and the equipment, the pre-processed answer sheet picture to be detected is divided through the pre-trained answer sheet division detection model, so that a plurality of sub-areas of the answer sheet to be detected are obtained, further, whether the sub-areas have the missed identification frame or not is judged according to the frame information and the answer sheet division detection model, if the missed identification frame exists, the frame of the missed identification frame is obtained and supplemented, the situations of missed detection and false detection are avoided, and the accuracy is improved.
Description
Technical Field
The invention relates to the technical field of image recognition, in particular to a method, a system, a storage medium and equipment for detecting a target area of an answer sheet.
Background
With the rise of artificial intelligence technology, the artificial intelligence technology has been applied to a plurality of industry fields, and people also increasingly use automation equipment to go over examination papers of students, so that the examination paper going over efficiency is higher.
In an examination, an answer sheet is often used to answer the question. The answer sheet generally comprises a positioning point, a basic information bar, a two-dimensional code, a test number filling area and a test question area, and examinees fill in or answer the answer sheet according to test paper questions. After the examination is finished, the answers on the answer sheet need to be interpreted through machine identification or manual assistance. Before the answers on the answer sheet are interpreted, the area of the answer sheet generally needs to be divided so as to better identify the corresponding answer area. Generally, area positioning is carried out based on positioning point identification, so that the requirement on the quality of a scanned answer sheet picture is high, and the answer sheet area cannot be correctly identified often due to problems of image inclination, resolution and the like; meanwhile, because the positioning points are relied on for carrying out the area positioning, the situation that the positioning points in the answer sheet picture do not exist can not be processed.
In the prior art, with the rise of artificial intelligence, the answer sheet detection method based on deep learning is also applied to answer sheet area detection scenes, but the existing target detection deep learning model has high complexity, high requirement on GPU load, low detection speed and difficult deployment and application, so that the conditions of missing detection and false detection exist in the deep learning model identification result, and the accuracy is low.
Disclosure of Invention
Based on this, the invention aims to provide an answer sheet target area detection method, system, storage medium and device, which are used for solving the technical problem of low accuracy of an identification result caused by the situations of missed detection and false detection of the identification result in the prior art.
The invention provides a method for detecting a target area of an answer sheet on one hand, which comprises the following steps:
acquiring an answer sheet picture to be detected, preprocessing the answer sheet picture to be detected, and segmenting the preprocessed answer sheet picture to be detected through a pre-trained answer sheet segmentation detection model to obtain a plurality of sub-regions of the answer sheet to be detected, wherein the plurality of sub-regions comprise option regions and subjective question regions;
acquiring frame information of the subarea, and judging whether the subarea has a missing identification frame or not by combining the frame information with the answer sheet segmentation detection model;
if the missed identification frame exists, acquiring and completing the frame of the missed identification frame;
wherein, mend the step of missing the frame of discerning the frame includes:
in the option area:
acquiring option features in the option areas, clustering the option areas according to the option features to obtain a plurality of area clusters, collecting and comparing option borders of the same area clusters, acquiring missed identification options, and completing the borders of the missed identification options;
in the main topic area:
acquiring the area frame of the subjective question area and calculating the frame interval between two adjacent subjective question areas;
judging whether the frame distance is larger than a distance threshold value or not;
and if the frame interval is larger than the interval threshold, a missing identification area exists between two adjacent subjective question areas, and the frame of the missing identification area is obtained and completed. According to the answer sheet target area detection method, a pre-processed answer sheet picture to be detected is divided through a pre-trained answer sheet division detection model, so that a plurality of sub-areas of the answer sheet to be detected are obtained, the traditional scheme that area positioning is carried out based on positioning point identification and the quality requirement of the scanned answer sheet picture is high is avoided, further, whether the sub-areas have missed identification frames or not is judged according to frame information and the answer sheet division detection model, if the missed identification frames exist, the frames of the missed identification frames are obtained and supplemented completely, the situations of missed detection and false detection are avoided, the accuracy is improved, specifically, for an option area, a plurality of area clusters are obtained through clustering the option area, option frames of the same area clusters are collected and compared, missed identification options are obtained, and the frames of the missed identification options are supplemented completely; for the subjective question area, the area frame of the subjective question area is obtained, the frame interval of two adjacent subjective question areas is calculated, whether the frame interval is larger than an interval threshold value is judged, if the frame interval is larger than the interval threshold value, an identification missing area exists between the two adjacent subjective question areas, the frame of the identification missing area is obtained and supplemented completely until the complete frame of the subjective question area is obtained, and therefore more accurate detection information of the target area of the answer sheet is obtained, the technical scheme that in the traditional technology, the answer sheet area cannot be identified correctly due to poor picture quality is avoided, and the technical problem that in the prior art, the identification result is missed and mistakenly detected, and the accuracy of the identification result is low is solved.
In addition, the answer sheet target area detection method according to the present invention may further have the following additional technical features:
further, if there is a missing identification frame, the step of obtaining and completing the frame of the missing identification frame includes:
identifying the frame of the supplemented missed identification frame, and acquiring the frame line of the supplemented missed identification frame;
and combining the frame lines of the frames which are not missed to be identified, and adaptively adjusting the frame lines of the missed to be identified after completing the frame lines so that the frame lines of the missed to be identified after completing the frame lines are mutually attached to the frame lines of the frames which are not missed to be identified, thereby obtaining the complete frame of the subjective question area.
Further, the step of obtaining the frame information of the sub-area and determining whether the sub-area has a missing identification frame according to the frame information and the answer sheet segmentation detection model includes:
performing self-adaptive adjustment on the obtained frame of the subregion by combining the pre-trained rectangular and linear detection models and the answer sheet segmentation detection model to obtain a self-adaptive subregion;
and performing column division processing on the answer sheet according to the adaptive frame information of the subareas to obtain an answer sheet layout structure.
Further, the step of obtaining the answer sheet picture to be detected comprises:
acquiring a certain number of original answer sheet pictures;
performing data enhancement on the original answer sheet picture through a pre-trained generated confrontation network model to generate a data-enhanced answer sheet picture;
marking each subarea of the answer sheet picture according to the original answer sheet picture and the data-enhanced answer sheet picture, wherein the subareas comprise an examination admission card number area, a question selecting area, a blank question filling area, a subjective question area, a bar code and a question selecting area;
and training the deep learning neural network model through the answer sheet pictures marked with the sub-regions to obtain the answer sheet segmentation detection model.
Further, if the frame interval is greater than the interval threshold, a missing identification region exists between two adjacent subjective problem regions, and the step of obtaining and completing the frame of the missing identification region includes:
identifying the area content between two adjacent subjective question areas, and judging whether a potential subarea exists between the two adjacent subjective question areas according to the area content, wherein handwritten characters, question numbers or scores are arranged in the potential subarea;
and if a potential sub-area exists between two adjacent main topic areas, the potential sub-area is a missing identification area.
Further, the preprocessing includes de-drying, data enhancement, and size variation.
Further, the step of obtaining and completing the border of the missed identification border further includes:
and (4) associating and storing each sub-area of the missed identification frame after the frame is completed with the answer sheet file, generating a teacher paper-marking template and storing the template as a question bank.
In another aspect, the present invention provides a system for detecting a target area of an answer sheet, where the system includes:
the acquisition and segmentation module is used for acquiring an answer sheet picture to be detected, preprocessing the answer sheet picture to be detected, and segmenting the preprocessed answer sheet picture to be detected through a pre-trained answer sheet segmentation detection model to obtain a plurality of sub-regions of the answer sheet to be detected, wherein the plurality of sub-regions comprise option regions and subjective question regions;
the judging module is used for acquiring the frame information of the subarea and judging whether the subarea has a missing identification frame or not by combining the frame information with the answer sheet segmentation detection model;
the first execution module is used for acquiring and completing the frame of the missed identification frame when the missed identification frame exists;
wherein, in the option area, the first execution module includes:
the first completion unit is used for acquiring option features in the option areas, clustering the option areas according to the option features to acquire a plurality of area clusters, acquiring and comparing option frames of the same area clusters, acquiring missed identification options, and completing the frames of the missed identification options;
in the main topic area, a first execution module comprises:
the calculation unit is used for acquiring the area frame of the subjective question area and calculating the frame interval between two adjacent subjective question areas;
the judging unit is used for judging whether the frame interval is larger than an interval threshold value;
and the second completion unit is used for obtaining and completing the frame of the missed identification area if the frame distance is larger than the distance threshold value, wherein the missed identification area exists between two adjacent subjective question areas.
Another aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the answer sheet target area detection method as described above.
The invention also provides a data processing device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the answer sheet target area detection method.
Drawings
Fig. 1 is a flowchart of a method for detecting an answer sheet target area according to a first embodiment of the present invention;
FIG. 2 is a flowchart illustrating the step S103 according to the first embodiment of the present invention;
fig. 3 is a flowchart of a method for detecting an answer sheet target area according to a second embodiment of the present invention;
FIG. 4 is a flowchart illustrating the step S202 in the second embodiment of the present invention;
FIG. 5 is a flowchart illustrating the step S203 in the second embodiment of the present invention;
FIG. 6 is a flowchart illustrating step S2033 according to the second embodiment of the present invention;
fig. 7 is a block diagram of an answer sheet target area detection system according to a third embodiment of the present invention;
fig. 8 is a style chart of the answer sheet in the application of the present invention.
The following detailed description will further illustrate the invention in conjunction with the above-described figures.
Detailed Description
To facilitate an understanding of the invention, the invention will now be described more fully with reference to the accompanying drawings. Several embodiments of the invention are presented in the drawings. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
The invention integrates the deep learning neural network model and the computer vision algorithm, can effectively detect the subareas of any answer sheet in any format, and stores all the subareas in the objective answer sheet area and the answer sheet file in a correlation manner. Specifically, the method includes annotating areas of batch sample answer sheet data, training sample data by combining a deep neural network to obtain an answer sheet segmentation detection model, and detecting each subarea by the answer sheet segmentation detection model when a new answer sheet to be detected is input, so that each subarea obtained through detection can be filled and judged. Meanwhile, whether the frame of each subregion detected by the answer sheet segmentation detection model has defects is judged by combining with a computer vision algorithm, if the frame has defects, the frame is similar to a frame which may miss identification of an individual subjective region or an option region, and the frame information and the model target detection result are utilized to modify and verify the model detection result by fusing the computer vision algorithm and logic judgment, so that more accurate target detection information is obtained. The method has the advantages that the region detection does not need to depend on positioning points, the limitation of the positioning points is eliminated, and meanwhile, the region frame is corrected by combining a computer vision algorithm, so that more accurate target detection information is obtained.
Example one
Referring to fig. 1, a method for detecting an answer sheet target area in a first embodiment of the present invention is shown, the method includes steps S101 to S103:
s101, obtaining an answer sheet picture to be detected, preprocessing the answer sheet picture to be detected, and segmenting the preprocessed answer sheet picture to be detected through a pre-trained answer sheet segmentation detection model to obtain a plurality of sub-regions of the answer sheet to be detected, wherein the plurality of sub-regions comprise option regions and subjective question regions.
In this embodiment, batch target answer sheet pictures are collected in advance, an answer sheet picture with an anti-network style migration is generated through a deep learning model, and a new answer sheet picture is generated, so that the number of pictures is increased, and the diversity of answer sheet data in a training model is improved. Generative Adaptive Networks (GAN) is a deep learning model. Further, pre-processing includes de-drying, data enhancement, and size change.
S102, frame information of the subareas is obtained, and whether the subareas have missed identification frames or not is judged according to the frame information and an answer sheet segmentation detection model.
If the missing identification frame exists, executing step S103;
and S103, acquiring and completing the frame of the missed identification frame.
In the option area, step S103 specifically includes:
acquiring option features in the option area, clustering the option area according to the option features to obtain a plurality of area clusters, acquiring and comparing option frames of the same area clusters, acquiring the missed identification option, and completing the frames of the missed identification option.
Further, as shown in fig. 2, in the main topic area, step S103 specifically includes steps S1031 to S1033:
and S1031, obtaining area borders of the subjective question areas and calculating the border interval between two adjacent subjective question areas.
S1032, whether the frame interval is larger than an interval threshold value is judged.
If the frame interval is greater than the interval threshold, a missing identification area exists between two adjacent main topic areas, and step S1033 is executed;
and S1033, obtaining and completing the frame of the missed identification area.
In summary, in the answer sheet target area detection method in the above embodiments of the invention, the pre-trained answer sheet segmentation detection model is used to segment the pre-processed answer sheet picture to be detected, thereby obtaining a plurality of subareas of the answer sheet to be detected, avoiding the traditional scheme of carrying out area positioning based on positioning point identification and leading the requirement on the quality of the scanned answer sheet picture to be higher, judging whether the sub-area has a missing identification frame according to the frame information and the answer sheet segmentation detection model, if so, the frame of the missed identification frame is obtained and supplemented, the conditions of missed detection and false detection are avoided, the accuracy is improved, and particularly, for the option area, clustering the option area to obtain a plurality of area clusters, then collecting and comparing option frames of the same area clusters to obtain missed identification options, and completing the frames of the missed identification options; for the subjective question area, the area frame of the subjective question area is obtained, the frame interval of two adjacent subjective question areas is calculated, whether the frame interval is larger than an interval threshold value is judged, if the frame interval is larger than the interval threshold value, an identification missing area exists between the two adjacent subjective question areas, the frame of the identification missing area is obtained and supplemented completely until the complete frame of the subjective question area is obtained, and therefore more accurate detection information of the target area of the answer sheet is obtained, the technical scheme that in the traditional technology, the answer sheet area cannot be identified correctly due to poor picture quality is avoided, and the technical problem that in the prior art, the identification result is missed and mistakenly detected, and the accuracy of the identification result is low is solved.
Example two
Referring to fig. 3, a method for detecting an answer sheet target area in a second embodiment of the present invention is shown, and the method includes steps S201 to S205:
s201, obtaining an answer sheet picture to be detected, preprocessing the answer sheet picture to be detected, and segmenting the preprocessed answer sheet picture to be detected through a pre-trained answer sheet segmentation detection model to obtain a plurality of sub-regions of the answer sheet to be detected, wherein the plurality of sub-regions comprise option regions and subjective question regions.
Specifically, the preprocessing comprises the steps of removing dryness of the answer sheet picture to be detected, performing data enhancement on the original answer sheet picture through generating the countermeasure network GAN, and adjusting the size of the answer sheet picture to be detected, so that the answer sheet picture to be detected is conveniently segmented, and the segmentation result is more accurate. Specifically, as shown in fig. 8, the overall effect diagram is a sample overall effect diagram of a teacher paper-reading template, the style of the answer sheet can be known through the diagram, and the choice question area in the diagram is the option area in the description of the present application.
As a specific example, before obtaining the picture of the answer sheet to be detected, the answer sheet segmentation detection model needs to be obtained in advance, which is specifically as follows:
acquiring a certain number of original answer sheet pictures; performing data enhancement on the original answer sheet picture through a pre-trained generated confrontation network model to generate a data-enhanced answer sheet picture; marking each subarea of the answer sheet picture according to the original answer sheet picture and the data-enhanced answer sheet picture, wherein the subareas comprise an examination admission card number area, a question selection area, a blank question filling area, a subjective question area, a bar code and a question selection area; and training the deep learning neural network model through the answer sheet pictures marked with the sub-regions to obtain an answer sheet segmentation detection model.
The training data set is enhanced by the aid of the deep learning model GAN, diversity of answer sheet sample pictures is improved, and robustness of the target detection model is enhanced.
Specifically, in a model training link, a lightweight deep learning network Yolo5 is modified, a backbone network is optimized to Transform from original CNN, and the whole network model structure is Yolov5x, so that the precision of the target detection model is improved, the robustness is high, and the answer sheet target detection model is obtained through transfer learning training, namely the answer sheet segmentation detection model. Furthermore, the deep learning network Yolo5 is a lightweight model, the model complexity is low, the requirement on the hardware GPU environment is low, the application deployment is relatively simple, the recognition speed is high, and the real-time detection is realized.
And storing the obtained answer sheet segmentation detection model in a database, and when an answer sheet picture to be detected is obtained, accurately segmenting the answer sheet through the answer sheet segmentation detection model in the database to obtain all sub-regions of the answer sheet.
S202, frame information of the sub-area is obtained, and whether the sub-area has a missing identification frame or not is judged according to the frame information and an answer sheet segmentation detection model.
Specifically, by combining with the layout structure of the answer sheet, comparing and identifying the subareas by combining with the answer sheet segmentation detection model through a computer vision algorithm, and judging whether the border result of the subareas has missing identification according to the identification result.
Further, as shown in fig. 4, step S202 specifically includes steps S2021 to S2022:
s2021, performing self-adaptive adjustment on the obtained frame of the sub-region by combining the pre-trained rectangular and linear detection models and the answer sheet segmentation detection model to obtain the self-adaptive sub-region.
And S2022, performing column division processing on the answer sheet according to the adaptive frame information of the subareas to obtain an answer sheet layout structure.
In some optional embodiments, the deep learning model may have some defects, which may cause the detection frame to be larger or smaller, so that the model frame is adaptively adjusted through a rectangle detection and straight line detection algorithm to obtain a more conformable detection frame; and then, the answer sheet is subjected to column division processing according to each adaptive subregion to obtain an answer sheet layout structure, so that the subregion can be conveniently contrasted and identified by combining a computer vision algorithm and an answer sheet segmentation detection model. Specifically, answer sheets are generally divided into a single column, two columns, three columns, and the like.
In some optional embodiments, because the deep learning model may have some defects, all detection frames may not be correctly obtained, and similarly, individual subjective regions or option regions may be missed to be identified, at this time, the model detection result is modified and verified by fusing the traditional computer vision algorithm and logic judgment by using the column information and the model target detection result, so that more accurate target detection information can be obtained.
If the missing identification frame exists, executing step S203;
and S203, acquiring and completing the frame of the missed identification frame.
Specifically, in the option area, step S203 specifically includes:
acquiring option features in the option area, clustering the option area according to the option features to obtain a plurality of area clusters, acquiring and comparing option frames of the same area clusters, acquiring the missed identification option, and completing the frames of the missed identification option.
Specifically, as shown in fig. 5, in the main topic area, step S203 specifically includes step S2031 to step S2034:
s2031, obtaining area frames of the subjective question areas and calculating the frame interval between two adjacent subjective question areas;
s2032, judging whether the frame interval is larger than an interval threshold value;
if the frame distance is larger than the distance threshold, judging that a missing identification area exists between two adjacent main topic areas, and executing the step S2033;
if the frame distance is not greater than the distance threshold, judging that no recognition missing area exists between two adjacent main topic areas, and executing a step S2034;
s2033, obtaining and completing the frame of the missed identification area.
And S2034, the frame of the missed identification area does not need to be supplemented completely.
Because the result of the model detection frame may have missing identification, the computer vision algorithm and the logic are fused to judge, modify and verify each area obtained by the model: extracting the characteristics of the selection questions in the option area through a clustering algorithm, judging whether the identification is missed according to the positions of the upper frame and the lower frame, and completing the selection questions according with clustering conditions if the identification is missed; and for the subjective question area, calculating the interval of each subjective question frame, if the interval proportion is larger than a set threshold value, judging that the identification is missed, completing the area, and performing self-adaptive adjustment after completing to finally obtain the frame of each area.
Further, as shown in fig. 6, step S2033 specifically includes steps S20331 to S20332:
s20331, identifying the area content between two adjacent main subject areas, and judging whether a potential sub-area exists between the two adjacent main subject areas according to the area content, wherein handwritten characters, question numbers or scores are arranged in the potential sub-area.
If a potential sub-area exists between two adjacent main topic areas, executing step S20332;
and S20332, the potential sub-area is a missing identification area.
If handwritten characters, an item number or a score and the like exist between two adjacent main subject areas, the fact that the area has contents is indicated, and therefore if the area is not scanned, the area is indicated to be a missed identification area, and further identification is needed.
And S204, identifying the frame of the supplemented missed identification frame, and acquiring the frame line of the supplemented missed identification frame.
S205, combining the frame lines of the frames which are not missed to be identified, and adaptively adjusting the frame lines of the missed to be identified after completion, so that the frame lines of the missed to be identified after completion and the frame lines of the frames which are not missed to be identified are mutually attached to obtain a complete frame of the subjective problem area.
Further, after a complete frame of the subjective question area is obtained, in order to facilitate later-stage reuse, each sub-area of the missed identification frame after the frame is completed is stored in association with the answer sheet file, and a teacher paper marking template is generated and stored as a question bank.
In the embodiment, each sub-area containing the complete frame is stored in association with the answer sheet file to generate the teacher paper-marking template, so that when the answer sheet to be corrected is identified by the system, the areas for the teacher to mark the paper can be quickly generated, the accuracy is improved, and the efficiency is also improved. Furthermore, each sub-area containing the complete frame is stored as a question bank, so that when the system identifies the answer sheet to be corrected, the system can quickly and accurately divide each area in the answer sheet for the teacher to read the paper.
In summary, in the answer sheet target area detection method in the above embodiments of the invention, the pre-trained answer sheet segmentation detection model is used to segment the pre-processed answer sheet picture to be detected, thereby obtaining a plurality of subareas of the answer sheet to be detected, avoiding the traditional scheme of carrying out area positioning based on positioning point identification and leading the requirement on the quality of the scanned answer sheet picture to be higher, judging whether the sub-area has a missing identification frame according to the frame information and the answer sheet segmentation detection model, if so, the frame of the missed identification frame is obtained and supplemented, the conditions of missed detection and false detection are avoided, the accuracy is improved, and particularly, for the option area, clustering the option area to obtain a plurality of area clusters, then collecting and comparing option frames of the same area clusters to obtain missed identification options, and completing the frames of the missed identification options; for the subjective question area, the area frame of the subjective question area is obtained, the frame interval of two adjacent subjective question areas is calculated, whether the frame interval is larger than an interval threshold value is judged, if the frame interval is larger than the interval threshold value, an identification missing area exists between the two adjacent subjective question areas, the frame of the identification missing area is obtained and supplemented completely until the complete frame of the subjective question area is obtained, and therefore more accurate detection information of the target area of the answer sheet is obtained, the technical scheme that in the traditional technology, the answer sheet area cannot be identified correctly due to poor picture quality is avoided, and the technical problem that in the prior art, the identification result is missed and mistakenly detected, and the accuracy of the identification result is low is solved.
EXAMPLE III
Referring to fig. 7, a system for detecting an answer sheet target area in a third embodiment of the present invention is shown, the system includes:
the acquisition and segmentation module is used for acquiring an answer sheet picture to be detected, preprocessing the answer sheet picture to be detected, and segmenting the preprocessed answer sheet picture to be detected through a pre-trained answer sheet segmentation detection model to obtain a plurality of sub-regions of the answer sheet to be detected, wherein the plurality of sub-regions comprise option regions and subjective question regions;
the judging module is used for acquiring the frame information of the subarea and judging whether the subarea has a missing identification frame or not by combining the frame information with the answer sheet segmentation detection model;
the first execution module is used for acquiring and completing the frame of the missed identification frame when the missed identification frame exists;
wherein, in the option area, the first execution module includes:
the first completion unit is used for acquiring option features in the option areas, clustering the option areas according to the option features to acquire a plurality of area clusters, acquiring and comparing option frames of the same area clusters, acquiring missed identification options, and completing the frames of the missed identification options;
in the main topic area, a first execution module comprises:
the calculation unit is used for acquiring the area frame of the subjective question area and calculating the frame interval between two adjacent subjective question areas;
the judging unit is used for judging whether the frame interval is larger than an interval threshold value;
and the second completion unit is used for obtaining and completing the frame of the missed identification area if the frame distance is larger than the distance threshold value, wherein the missed identification area exists between two adjacent subjective question areas.
In some optional embodiments, the first execution module may further include:
the identification and acquisition module is used for identifying the frame of the supplemented missed identification frame and acquiring the frame line of the supplemented missed identification frame;
and the self-adaptive module is used for self-adaptively adjusting the frame lines of the supplemented missed identification frame by combining the frame lines of the missed identification frame, so that the supplemented frame lines of the missed identification frame and the frame lines of the missed identification frame are mutually attached to obtain the complete frame of the subjective problem area.
In some optional embodiments, the determining module may further include:
the detection unit is used for carrying out self-adaptive adjustment on the obtained frame of the subregion by combining the answer sheet segmentation detection model through a pre-trained rectangular and linear detection model to obtain the adaptive subregion;
and the column processing unit is used for performing column processing on the answer sheet according to the adaptive frame information of the subareas to obtain the answer sheet layout structure.
In some optional embodiments, the obtaining and segmenting module may further comprise:
the original answer sheet acquisition module is used for acquiring a certain number of original answer sheet pictures;
the data enhancement module is used for performing data enhancement on the original answer sheet picture through a pre-trained countermeasures generation network model to generate a data enhanced answer sheet picture;
the marking module is used for marking each subarea of the answer sheet picture according to the original answer sheet picture and the data-enhanced answer sheet picture, wherein the subareas comprise an examination admission card number area, a question selecting area, a blank question filling area, a subjective question area, a bar code and a question selecting area;
and the segmentation detection module is used for training the deep learning neural network model through the answer sheet pictures marked with the sub-regions to obtain the answer sheet segmentation detection model.
In summary, in the answer sheet target area detection system in the above embodiments of the invention, the pre-trained answer sheet segmentation detection model is used to segment the pre-processed answer sheet picture to be detected, thereby obtaining a plurality of subareas of the answer sheet to be detected, avoiding the traditional scheme of carrying out area positioning based on positioning point identification and leading the requirement on the quality of the scanned answer sheet picture to be higher, judging whether the sub-area has a missing identification frame according to the frame information and the answer sheet segmentation detection model, if so, the frame of the missed identification frame is obtained and supplemented, the conditions of missed detection and false detection are avoided, the accuracy is improved, and particularly, for the option area, clustering the option area to obtain a plurality of area clusters, then collecting and comparing option frames of the same area clusters to obtain missed identification options, and completing the frames of the missed identification options; for the subjective question area, the area frame of the subjective question area is obtained, the frame interval of two adjacent subjective question areas is calculated, whether the frame interval is larger than an interval threshold value is judged, if the frame interval is larger than the interval threshold value, an identification missing area exists between the two adjacent subjective question areas, the frame of the identification missing area is obtained and supplemented completely until the complete frame of the subjective question area is obtained, and therefore more accurate detection information of the target area of the answer sheet is obtained, the technical scheme that in the traditional technology, the answer sheet area cannot be identified correctly due to poor picture quality is avoided, and the technical problem that in the prior art, the identification result is missed and mistakenly detected, and the accuracy of the identification result is low is solved.
Furthermore, an embodiment of the present invention also proposes a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps of the method in the above-described embodiment.
Furthermore, an embodiment of the present invention also provides a data processing apparatus, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the steps of the method in the above-mentioned embodiment.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
Claims (10)
1. A method for detecting an answer sheet target area is characterized by comprising the following steps:
acquiring an answer sheet picture to be detected, preprocessing the answer sheet picture to be detected, and segmenting the preprocessed answer sheet picture to be detected through a pre-trained answer sheet segmentation detection model to obtain a plurality of sub-regions of the answer sheet to be detected, wherein the plurality of sub-regions comprise option regions and subjective question regions;
acquiring frame information of the subarea, and judging whether the subarea has a missing identification frame or not by combining the frame information with the answer sheet segmentation detection model;
if the missed identification frame exists, acquiring and completing the frame of the missed identification frame;
wherein, mend the step of missing the frame of discerning the frame includes:
in the option area:
acquiring option features in the option areas, clustering the option areas according to the option features to obtain a plurality of area clusters, collecting and comparing option borders of the same area clusters, acquiring missed identification options, and completing the borders of the missed identification options;
in the main topic area:
acquiring the area frame of the subjective question area and calculating the frame interval between two adjacent subjective question areas;
judging whether the frame distance is larger than a distance threshold value or not;
and if the frame interval is larger than the interval threshold, a missing identification area exists between two adjacent subjective question areas, and the frame of the missing identification area is obtained and completed.
2. The method for detecting the target area of the answer sheet according to claim 1, wherein if the missing identification border exists, the step of obtaining and completing the border of the missing identification border comprises the following steps:
identifying the frame of the supplemented missed identification frame, and acquiring the frame line of the supplemented missed identification frame;
and combining the frame lines of the frames which are not missed to be identified, and adaptively adjusting the frame lines of the missed to be identified after completing the frame lines so that the frame lines of the missed to be identified after completing the frame lines are mutually attached to the frame lines of the frames which are not missed to be identified, thereby obtaining the complete frame of the subjective question area.
3. The method for detecting the answer sheet target area according to claim 2, wherein the step of obtaining the frame information of the sub-area and determining whether the sub-area has a missing identification frame according to the frame information and the answer sheet segmentation detection model comprises:
performing self-adaptive adjustment on the obtained frame of the subregion by combining the pre-trained rectangular and linear detection models and the answer sheet segmentation detection model to obtain a self-adaptive subregion;
and performing column division processing on the answer sheet according to the adaptive frame information of the subareas to obtain an answer sheet layout structure.
4. The method for detecting the target area of the answer sheet according to claim 1, wherein the step of obtaining the picture of the answer sheet to be detected comprises the following steps:
acquiring a certain number of original answer sheet pictures;
performing data enhancement on the original answer sheet picture through a pre-trained generated confrontation network model to generate a data-enhanced answer sheet picture;
marking each subarea of the answer sheet picture according to the original answer sheet picture and the data-enhanced answer sheet picture, wherein the subareas comprise an examination admission card number area, a question selecting area, a blank question filling area, a subjective question area, a bar code and a question selecting area;
and training the deep learning neural network model through the answer sheet pictures marked with the sub-regions to obtain the answer sheet segmentation detection model.
5. The method for detecting the target area of the answer sheet according to claim 1, wherein if the border interval is greater than the interval threshold, an identification missing area exists between two adjacent subjective question areas, and the step of obtaining and completing the border of the identification missing area comprises:
identifying the area content between two adjacent subjective question areas, and judging whether a potential subarea exists between the two adjacent subjective question areas according to the area content, wherein handwritten characters, question numbers or scores are arranged in the potential subarea;
and if a potential sub-area exists between two adjacent main topic areas, the potential sub-area is a missing identification area.
6. The answer sheet target area detection method of claim 1, wherein the preprocessing comprises drying, data enhancement and size change.
7. The method for detecting the target area of the answer sheet according to claim 1, wherein the step of obtaining and completing the border of the missed-recognition border further comprises the following steps:
and (4) associating and storing each sub-area of the missed identification frame after the frame is completed with the answer sheet file, generating a teacher paper-marking template and storing the template as a question bank.
8. An answer sheet target area detection system, the system comprising:
the acquisition and segmentation module is used for acquiring an answer sheet picture to be detected, preprocessing the answer sheet picture to be detected, and segmenting the preprocessed answer sheet picture to be detected through a pre-trained answer sheet segmentation detection model to obtain a plurality of sub-regions of the answer sheet to be detected, wherein the plurality of sub-regions comprise option regions and subjective question regions;
the judging module is used for acquiring the frame information of the subarea and judging whether the subarea has a missing identification frame or not by combining the frame information with the answer sheet segmentation detection model;
the first execution module is used for acquiring and completing the frame of the missed identification frame when the missed identification frame exists;
wherein, in the option area, the first execution module includes:
the first completion unit is used for acquiring option features in the option areas, clustering the option areas according to the option features to acquire a plurality of area clusters, acquiring and comparing option frames of the same area clusters, acquiring missed identification options, and completing the frames of the missed identification options;
in the main topic area, a first execution module comprises:
the calculation unit is used for acquiring the area frame of the subjective question area and calculating the frame interval between two adjacent subjective question areas;
the judging unit is used for judging whether the frame interval is larger than an interval threshold value;
and the second completion unit is used for obtaining and completing the frame of the missed identification area if the frame distance is larger than the distance threshold value, wherein the missed identification area exists between two adjacent subjective question areas.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the answer sheet target area detection method according to any one of claims 1-7.
10. A data processing apparatus comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the answer sheet target area detection method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111323174.8A CN113762274B (en) | 2021-11-10 | 2021-11-10 | Answer sheet target area detection method, system, storage medium and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111323174.8A CN113762274B (en) | 2021-11-10 | 2021-11-10 | Answer sheet target area detection method, system, storage medium and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113762274A true CN113762274A (en) | 2021-12-07 |
| CN113762274B CN113762274B (en) | 2022-02-15 |
Family
ID=78784824
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111323174.8A Active CN113762274B (en) | 2021-11-10 | 2021-11-10 | Answer sheet target area detection method, system, storage medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113762274B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114639108A (en) * | 2022-05-19 | 2022-06-17 | 江西风向标智能科技有限公司 | Appraising mark identification method, system, storage medium and equipment of subjective question |
| CN114863095A (en) * | 2022-03-25 | 2022-08-05 | 电子科技大学 | Answer sheet image segmentation method based on color conversion |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105989347A (en) * | 2015-02-28 | 2016-10-05 | 科大讯飞股份有限公司 | Intelligent marking method and system of objective questions |
| CN108171297A (en) * | 2018-01-24 | 2018-06-15 | 谢德刚 | A kind of answer card identification method and device |
| CN108388892A (en) * | 2018-05-04 | 2018-08-10 | 苏州大学 | Paper automated processing system based on OpenCV and method |
| CN108416345A (en) * | 2018-02-08 | 2018-08-17 | 海南云江科技有限公司 | A kind of answering card area recognizing method and computing device |
| CN108764074A (en) * | 2018-05-14 | 2018-11-06 | 山东师范大学 | Subjective item intelligently reading method, system and storage medium based on deep learning |
| CN108805760A (en) * | 2017-05-06 | 2018-11-13 | 南京多邦软件有限公司 | A kind of intelligently reading system |
| CN110516208A (en) * | 2019-08-12 | 2019-11-29 | 深圳智能思创科技有限公司 | A kind of system and method extracted for PDF document table |
| CN110689013A (en) * | 2019-10-10 | 2020-01-14 | 北京课程帮科技有限公司 | Automatic marking method and system based on feature recognition |
| CN110929562A (en) * | 2019-10-12 | 2020-03-27 | 杭州电子科技大学 | Answer sheet identification method based on improved Hough transformation |
| US10878270B1 (en) * | 2018-06-26 | 2020-12-29 | Amazon Technologies, Inc. | Keypoint-based multi-label word segmentation and localization |
| CN112270261A (en) * | 2020-10-28 | 2021-01-26 | 广州华多网络科技有限公司 | Segmentation method and device for question stem and answer mixture and storage medium |
| CN113283431A (en) * | 2021-07-26 | 2021-08-20 | 江西风向标教育科技有限公司 | Intelligent method and system integrating deep learning and logic judgment |
| CN113469147A (en) * | 2021-09-02 | 2021-10-01 | 北京世纪好未来教育科技有限公司 | Answer sheet identification method and device, electronic equipment and storage medium |
-
2021
- 2021-11-10 CN CN202111323174.8A patent/CN113762274B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105989347A (en) * | 2015-02-28 | 2016-10-05 | 科大讯飞股份有限公司 | Intelligent marking method and system of objective questions |
| CN108805760A (en) * | 2017-05-06 | 2018-11-13 | 南京多邦软件有限公司 | A kind of intelligently reading system |
| CN108171297A (en) * | 2018-01-24 | 2018-06-15 | 谢德刚 | A kind of answer card identification method and device |
| CN108416345A (en) * | 2018-02-08 | 2018-08-17 | 海南云江科技有限公司 | A kind of answering card area recognizing method and computing device |
| CN108388892A (en) * | 2018-05-04 | 2018-08-10 | 苏州大学 | Paper automated processing system based on OpenCV and method |
| CN108764074A (en) * | 2018-05-14 | 2018-11-06 | 山东师范大学 | Subjective item intelligently reading method, system and storage medium based on deep learning |
| US10878270B1 (en) * | 2018-06-26 | 2020-12-29 | Amazon Technologies, Inc. | Keypoint-based multi-label word segmentation and localization |
| CN110516208A (en) * | 2019-08-12 | 2019-11-29 | 深圳智能思创科技有限公司 | A kind of system and method extracted for PDF document table |
| CN110689013A (en) * | 2019-10-10 | 2020-01-14 | 北京课程帮科技有限公司 | Automatic marking method and system based on feature recognition |
| CN110929562A (en) * | 2019-10-12 | 2020-03-27 | 杭州电子科技大学 | Answer sheet identification method based on improved Hough transformation |
| CN112270261A (en) * | 2020-10-28 | 2021-01-26 | 广州华多网络科技有限公司 | Segmentation method and device for question stem and answer mixture and storage medium |
| CN113283431A (en) * | 2021-07-26 | 2021-08-20 | 江西风向标教育科技有限公司 | Intelligent method and system integrating deep learning and logic judgment |
| CN113469147A (en) * | 2021-09-02 | 2021-10-01 | 北京世纪好未来教育科技有限公司 | Answer sheet identification method and device, electronic equipment and storage medium |
Non-Patent Citations (6)
| Title |
|---|
| HAIYING ZHANG 等: "Robust Detection Method of Small Targets in Sea-Clutter via Improved Fast Clustering Segmentation", 《2016 8TH INTERNATIONAL CONFERENCE ON INTELLIGENT HUMAN-MACHINE SYSTEMS AND CYBERNETICS》 * |
| HUI WEI 等: "Contour segment grouping for object detection", 《JOURNAL OF VISUAL COMMUNICATION AND IMAGE RESPRESENTATION》 * |
| 孙建芳 等: "扫描阅卷系统中模板定制和图像聚类方法的研究", 《计算技术与自动化》 * |
| 孙琳 等: "一种适用于移动设备在线阅卷的答题卡自动识别算法", 《计算机测量与控制》 * |
| 方慧琴 等: "阅卷系统中的答题区域快速智能分割算法研究", 《现代电子技术》 * |
| 要曙丽 等: "一种答题卡客观题识别算法", 《图学学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114863095A (en) * | 2022-03-25 | 2022-08-05 | 电子科技大学 | Answer sheet image segmentation method based on color conversion |
| CN114863095B (en) * | 2022-03-25 | 2023-11-28 | 电子科技大学 | Answer sheet image segmentation method based on color conversion |
| CN114639108A (en) * | 2022-05-19 | 2022-06-17 | 江西风向标智能科技有限公司 | Appraising mark identification method, system, storage medium and equipment of subjective question |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113762274B (en) | 2022-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110647829A (en) | Bill text recognition method and system | |
| CN113762274B (en) | Answer sheet target area detection method, system, storage medium and equipment | |
| CN111597908A (en) | Test paper correcting method and test paper correcting device | |
| CN111291629A (en) | Method and device for recognizing text in image, computer equipment and computer storage medium | |
| CN105046200B (en) | Electronic paper marking method based on straight line detection | |
| CN110582783B (en) | Training device, image recognition device, training method, and computer-readable information storage medium | |
| CN111626284A (en) | Method and device for removing handwritten fonts, electronic equipment and storage medium | |
| CN110866529A (en) | Character recognition method, character recognition device, electronic equipment and storage medium | |
| CN112507758A (en) | Answer sheet character string identification method, answer sheet character string identification device, terminal and computer storage medium | |
| WO2021232670A1 (en) | Pcb component identification method and device | |
| CN112446259A (en) | Image processing method, device, terminal and computer readable storage medium | |
| CN116798036B (en) | Method and device for identifying and checking answer sheet objective question identification result | |
| CN115393861B (en) | Method for accurately segmenting handwritten text | |
| CN110135225A (en) | Sample mask method and computer storage medium | |
| CN112347997A (en) | Test question detection and identification method and device, electronic equipment and medium | |
| CN111144270B (en) | Neural network-based handwritten text integrity evaluation method and evaluation device | |
| CN111144425B (en) | Method and device for detecting shot screen picture, electronic equipment and storage medium | |
| CN106033534B (en) | Electronic paper marking method based on straight line detection | |
| CN111008594A (en) | Error correction evaluation method, related equipment and readable storage medium | |
| CN114882204A (en) | Automatic ship name recognition method | |
| CN108921006B (en) | Method for establishing handwritten signature image authenticity identification model and authenticity identification method | |
| CN110490056A (en) | The method and apparatus that image comprising formula is handled | |
| Aliev et al. | Algorithm for choosing the best frame in a video stream in the task of identity document recognition | |
| JP7293658B2 (en) | Information processing device, information processing method and program | |
| CN112613367A (en) | Bill information text box acquisition method, system, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 330000 Building 1, maiyuan Road, Nanchang Economic and Technological Development Zone, Jiangxi Province Patentee after: Jiangxi wind vane Intelligent Technology Co.,Ltd. Address before: 330000 Building 1, maiyuan Road, Nanchang Economic and Technological Development Zone, Jiangxi Province Patentee before: JIANGXI VANEDUCATION TECHNOLOGY Inc. |
|
| CP01 | Change in the name or title of a patent holder |