Image processing method

Abstract

The embodiment of the invention provides an image processing method, and the method comprises the steps: determining an answer image of an answer sheet, and enabling the answer image to comprise first answer information with a score and second answer information without a score; based on answer position information corresponding to the answer sheet, determining a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information from the answer image; identifying a score of the first answer information from the first answer area, and performing score evaluation on the second answer information in the second answer area by using a score evaluation model to obtain a score of the second answer information; and determining a target score corresponding to the answer sheet based on the score of the first answer information and the score of the second answer information.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of artificial intelligence technology, and in particular to an image processing method. One or more embodiments of this specification also relate to a computing device, a computer-readable storage medium, and a computer program product. Background Art

[0002] With the continuous development of computer technology, the method of using computer equipment to recognize scores on answer sheets has also been widely used.

[0003] Currently, in the process of using computer equipment to recognize scores on answer sheets, the score corresponding to the answer sheet cannot be accurately identified due to the large amount and complex types of answer information contained in the answer sheet. Therefore, how to accurately identify the score corresponding to the answer sheet has become an urgent problem to be solved. Summary of the invention

[0004] In view of this, an embodiment of the present specification provides an image processing method. One or more embodiments of the present specification also relate to an image processing apparatus, a computing device, a computer-readable storage medium and a computer program product to solve the technical defects existing in the prior art.

[0005] According to a first aspect of an embodiment of this specification, there is provided an image processing method, including:

[0006] Determining an answer image of the answer sheet, wherein the answer image includes first answer information with a score and second answer information without a score;

[0007] Based on the answer position information corresponding to the answer sheet, determining a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information from the answer image;

[0008] Identifying the score of the first answer information from the first answer area, and performing score evaluation on the second answer information in the second answer area using a score evaluation model to obtain the score of the second answer information;

[0009] Based on the score of the first answer information and the score of the second answer information, a target score corresponding to the answer sheet is determined.

[0010] According to a second aspect of the embodiments of this specification, there is provided an image processing apparatus, including:

[0011] An image determination module is configured to determine an answer image of an answer sheet, wherein the answer image includes first answer information with a score and second answer information without a score;

[0012] an area determination module, configured to determine, from the answer image, a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information based on the answer position information corresponding to the answer sheet;

[0013] a first score determination module configured to identify the score of the first answer information from the first answer area, and to perform score evaluation on the second answer information in the second answer area using a score evaluation model to obtain the score of the second answer information;

[0014] The second score determination module is configured to determine the target score corresponding to the answer sheet based on the score of the first answer information and the score of the second answer information.

[0015] According to a third aspect of an embodiment of this specification, a computing device is provided, including:

[0016] Memory and processor;

[0017] The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the steps of the above-mentioned image processing method are implemented.

[0018] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores a computer program / instruction, and when the computer program / instruction is executed by a processor, the steps of the above-mentioned image processing method are implemented.

[0019] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, comprising a computer program / instruction, which implements the steps of the above-mentioned image processing method when executed by a processor.

[0020] One or more embodiments of this specification provide an image processing method. In the process of score recognition of an answer sheet, the first answer area corresponding to the first answer information and the second answer area corresponding to the second answer information can be determined from the answer image according to the answer position information corresponding to the answer sheet, so that the first answer area and the second answer area can accurately classify the answer information with a large number and complex types. In addition, the score of the first answer information is identified from the first answer area, and the score evaluation model is used to evaluate the score of the second answer information in the second answer area to obtain the score of the second answer information; different score determination methods are used in a targeted manner according to different types of answer information, so as to accurately obtain the score of each answer information. Finally, based on the score of the first answer information and the score of the second answer information, the target score corresponding to the answer sheet is accurately obtained, avoiding the problem that the score corresponding to the answer sheet cannot be accurately identified due to the large number and complex types of answer information contained in the answer sheet. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is an application schematic diagram of an image processing method provided by an embodiment of this specification;

[0022] Figure 2 is a flow chart of an image processing method provided by an embodiment of this specification;

[0023] Figure 3 It is a schematic diagram of an answer graph, an answer sheet graph, and matching information in an image processing method provided in one embodiment of this specification;

[0024] Figure 4 is a schematic diagram of a question answering style in an image processing method provided by an embodiment of this specification;

[0025] Figure 5 is a schematic diagram of a processing process of an image processing method provided by an embodiment of this specification;

[0026] Figure 6 is a structural schematic diagram of an image processing device provided by an embodiment of this specification;

[0027] Figure 7 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0028] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0029] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0030] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0031] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0032] First, the terms involved in one or more embodiments of this specification are explained.

[0033] Target detection model: It is an algorithm used in the field of computer vision to identify and locate the location and category of one or more targets (objects) in an image. Target detection models can accomplish two tasks at the same time: 1) Identify the targets present in the image and determine what they are (for example, people, cars, cats, etc.); 2) Locate the exact location of these targets in the image, usually by drawing bounding boxes (bbox).

[0034] YOLO (You On ly Look Once): is an object detection algorithm. Its core idea is to solve the object detection task as a regression problem. Unlike traditional object detection methods, the YOLO algorithm predicts the probabilities of multiple bounding boxes and their categories in a single network at the same time, thereby achieving fast and efficient object detection. The main advantage of the YOLO algorithm is its speed, which is particularly suitable for real-time object detection applications.

[0035] IoU (Intersection over Union): refers to the intersection ratio. IoU is an indicator that measures the degree of overlap between two bounding boxes and is widely used in the field of target detection. It is the ratio of the intersection area of ​​two bounding boxes to the union area, that is, IoU = intersection area / union area. Among them, the variants of IoU include but are not limited to: 1. GIoU (Generated IoU): GIoU not only measures the overlapping area, but also measures the distance between the two boxes by considering the area of ​​the minimum circumscribed rectangular box. The value range of GIoU is [-1,1]. When the two images completely overlap, IoU = GIoU = 1, and when the two images do not intersect, IoU = 0, GIoU = -1. 2. DIoU (Division IoU): DIoU takes into account the distance between the center points of the two boxes and is suitable for processing non-overlapping and distant bounding boxes. 3.CIoU (Complete IoU): CIoU is one of the most complex IoU variants. It takes into account the distance between the center points of the two boxes, the similarity of the shapes / sizes of the two boxes, and balances the two through the weight term α.

[0036] Matching information: refers to information that is consistent with the layout of the answer sheet. The matching information is not limited to a carrier and can be a picture or a special data structure. Different boxes are used to mark key areas in the matching information to assist in the collection of homework information and answer information.

[0037] Answer sheet collection: The paper answer sheet is converted into an electronic image through a high-speed scanner or by taking a photo and uploading it; and a series of algorithms are used to detect, identify, and correct the answer information in the answer sheet image.

[0038] Answer sheet filling area: The area on the paper answer sheet used for answering objective questions (multiple-choice questions, true-or-false questions, etc.), generally filled in with a 2B pencil.

[0039] Answer sheet score trace: The paper answer sheet contains the teacher’s scoring information.

[0040] Score box: A box used to designate where the teacher will leave marks within the box.

[0041] OCR (Optical Character Recognition) refers to the process in which an electronic device (such as a scanner or digital camera) examines characters printed on paper, determines their shape by detecting dark and light patterns, and then uses character recognition methods to translate the shape into computer text.

[0042] With the continuous development of computer technology, the method of using computer equipment to identify the scores of answer sheets has also been widely used. At present, in the process of using computer equipment to identify the scores of answer sheets, due to the large amount and complex types of answer information contained in the answer sheet, it is impossible to accurately identify the score corresponding to the answer sheet.

[0043] For example, in examination scenarios such as the junior high school entrance examination and the college entrance examination, students can answer questions on answer sheets; in the answer sheet solutions used in the junior high school entrance examination and the college entrance examination, each answer sheet is printed on special paper, and can be scanned or collected by designated machines and graded by designated systems.

[0044] However, the above answer sheet solution has certain problems. The printing cost and collection cost of answer sheets for traditional exams are very high. Answer sheets are made of hard cardboard, and the printing cost of this special cardboard is high. In addition, answer sheet collection requires scanning with a special machine, and after collection, teachers also need to correct them on a special system. The entire answer and scoring process is: students need to answer -> scan -> teachers correct in the system -> obtain students' test scores, which makes the answer sheet collection procedure complicated.

[0045] In order to allow students to adapt to answer sheet exams in advance, a lower-cost method is needed to support exam practice in answer sheet scenarios; and automatically collect the scores of each question in the test paper (in order to conduct question-level analysis and statistical display according to students and classes).

[0046] Based on this, an image processing method is provided in this specification. One or more embodiments of this specification also relate to an image processing apparatus, a computing device, a computer-readable storage medium and a computer program product, which are described in detail one by one in the following embodiments.

[0047] See also Figure 1 , Figure 1 is an application diagram of an image processing method provided by an embodiment of this specification. The image processing method provided by the embodiment of this application can be applied to Figure 1 The application scenarios shown are not limited to these. Figure 1 In the application scenario shown, the image processing method of this specification can be deployed in a server 10, and the server 10 can be connected to one or more client devices 20 through a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. The client device 20 here can include but is not limited to: a smart phone, a tablet computer, a laptop computer, a PDA, a personal computer, a smart home device, a car device, etc. The client device 20 can interact with the user through a graphical user interface to implement the method provided in the embodiment of this specification.

[0048] In an embodiment of the present specification, the system composed of the client device 20 and the server 10 can execute the following steps: the client device 20 executes sending the answer image of the answer sheet to the server 10; the server 10 executes the steps of segmenting the answer image using matching information, identifying the scores of objective questions and subjective questions, and determining the scores of the answer sheet; wherein, segmenting the answer image using matching information means: segmenting the answer area according to the matching information; the answer area is divided into two types: one is the filling area for objective questions, and the filling area of ​​each question needs to be intercepted to obtain a screenshot of the objective question, and the other is the answer area for the answer question, and the answer screenshot corresponding to the answer question is obtained by intercepting the answer area.

[0049] Among them, the score recognition of objective questions and subjective questions refers to: using the model to score each question in the objective question screenshot, and obtaining all the score traces in the answer question screenshot, so as to determine the score of the answer question. Among them, determining the score of the answer sheet refers to: adding the score of the objective question and the score of the subjective question to obtain the score of the answer sheet (i.e., the target score); the score of the answer sheet can be sent to the client device for display later.

[0050] See also Figure 2 , Figure 2 A flowchart of an image processing method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0051] Step 202: Determine an answer image of the answer sheet, wherein the answer image includes first answer information with a score and second answer information without a score.

[0052] Among them, the answer sheet can be understood as an answer sheet used for examinations, or an answer sheet used for homework writing; the answer sheet includes the first answer information and the second answer information written by the student, and the score of the first answer information.

[0053] The first answer information may be a type of answer information, and the first answer information may be text-type answer information; for example, the first answer information may be answer information (such as fill-in-the-blank questions, answer questions, etc.) for answering questions (such as fill-in-the-blank questions, answer characters written for answering questions, etc.).

[0054] The second answer information may be another type of answer information, such as fill-in type answer information or symbol information (such as symbols such as √ and ×); for example, the second answer information may be answer information for objective questions (such as multiple-choice questions and true-or-false questions).

[0055] The first answer information with a score refers to the score obtained by evaluating the first answer information in the answer sheet, and the score of the first answer information is written on the answer sheet.

[0056] The answer image can be understood as an image obtained by performing an answer sheet capture operation on the answer sheet.

[0057] In the answer sheet test practice scenario, students' answers are divided into two categories: objective questions (multiple-choice questions) and composition questions (fill-in-the-blank questions, answer questions, etc.).

[0058] Specifically, the method can be applied to the server and receive the answer image of the answer sheet sent by the client (ie, the client device); the server can then perform score statistics on the answer image to obtain the target score of the answer sheet.

[0059] Taking the application of the image processing method provided in this specification in an examination scenario as an example, the image processing method is explained; when this method is applied in an examination scenario, in the client question combination service of the examination scenario (selecting questions in the online question bank to generate an assignment list), a test paper document file (such as a word, PDF file) of a test paper image (containing only test questions) can be generated, and an answer sheet image (containing only the blanks that need to be answered, the style can refer to the answer sheets of the middle school entrance examination and the college entrance examination) will also be generated, as well as matching information consistent with the layout of the answer sheet image. Different boxes are used to mark key areas in the matching information to assist in the collection of assignment information and answer information.

[0060] For examples of user combination answer graphs, answer sheet graphs, and matching information, see Figure 3 , Figure 3 It is a schematic diagram of an answer chart, an answer sheet chart, and matching information in an image processing method provided in one embodiment of this specification. Figure 3 It can be seen that this method can render the answer sheet image based on the test paper layout ( Figure 3 (b in the figure); generate an answer sheet image based on the questions and order selected by the user when setting the questions. Figure 3 Figure a in the figure is the answer figure. Based on figure a, it can be seen that corresponding filling areas can be generated for objective questions (multiple-choice questions, true-or-false questions, etc.); corresponding answer areas can be generated for other answer questions, and each answer area contains a score box; the score box of this method is generated on the right side or upper right corner of each question.

[0061] Rendering matching information based on test paper layout (visualization see Figure 3 c); when the answer sheet is generated, the corresponding matching information is generated. Corresponding solid color boxes will be generated at the locations of the filling area, answer area, and score box. It should be noted that the matching information is used to mark the location and area of ​​each element (filling or answer area) in the answer sheet, and is not limited to the form.

[0062] The answer sheet image and matching information are generated 1:1, and the superimposed display effect is as follows Figure 3 In addition, the answer sheet image and the corresponding matching information are stored in association in the database, and can be used to obtain relevant information of the answer image.

[0063] After completing the question setting, the user can print the test paper document file or test paper image as an answer sheet for use; the students can fill in the answers and the teacher can score the answer sheet, and then the answer image is captured from the answer sheet after the answer is completed, and the answer image (i.e., the answer image) is uploaded to the server through the client device (the answer image can be in A3 or A4 paper format).

[0064] In one or more embodiments provided in this specification, determining the answer sheet image of the answer sheet includes:

[0065] Receiving the answer image to be processed of the answer sheet sent by the client, wherein the answer image to be processed is obtained by capturing the image of the answer sheet after answering;

[0066] The image direction of the answer image to be processed is adjusted using a preset image direction, and / or the image of the answer image to be processed is segmented using a preset image size to obtain the answer sheet image;

[0067] The client may be a client device; the answer image to be processed may be understood as an answer image that needs to be preprocessed. For example, the answer image to be processed may be an answer image with incorrect text direction and incorrect image size.

[0068] Among them, the preset image direction can be set according to the actual application scenario, for example, the preset image direction can be the positive direction of the text; the preset image size can be set according to the actual application scenario, for example, the preset image size can be the size of A3 or A4 paper.

[0069] Using the above example, this method can convert and split the answer sheet; the answer sheet style is divided into two types, namely A4 single column and A3 multi-column (double column and triple column). For the answer sheet in multi-column format, this method can split it into single column and process it uniformly. Figure 4 , Figure 4 is a schematic diagram of a question answering style in an image processing method provided in an embodiment of this specification. Figure 4 As you can see, the answer graph can have different formats, for example, Figure 4 There are single-column answer chart, double-column answer chart, and three-column answer chart.

[0070] In actual applications, the A4 answer sheet may be upside down when scanning, so it needs to be straightened. Specifically, the scenario in which schools use score collection is mainly for exams, and they are accustomed to using A3-style test papers (consisting of two A4 pages (i.e. double-column format) or A3 paper divided into three columns). Due to the size of the high-speed scanner, the A3-style test paper enters the high-speed scanner with the short edge when scanning, so it needs to be straightened before collection. In order to improve the collection effect and reuse the A4-style job processing flow, the A3-style test paper needs to be divided into a style similar to A4 according to the number of columns. The specific process is as follows:

[0071] 1. Orientation: Orient the answer image to the direction where the text is positive (i.e. the preset image direction). The following three strategies can be used.

[0072] Strategy 1: Use the marks in the question sheet to determine the direction of the question sheet by detecting and identifying the marks; Strategy 2: Use OCR to obtain the text direction and determine the direction based on the text direction; Strategy 3: Set the scanning direction when scanning with a high-speed scanner, and automatically process and straighten it after scanning.

[0073] 2. Cutting judgment: Because there are both A3 and A4 papers in the input, A3 paper needs to be cut but A4 paper does not, so a judgment logic needs to be added. The following two strategies can be used.

[0074] Strategy 1: Judging by the paper size, A3 paper is larger than A4 paper. Strategy 2: Judging by the paper proportions, after the conversion, A3 paper is wider than it is long, while A4 paper is longer than it is wide.

[0075] 3. Cutting: Cut the A3 paper into two A4 paper. Then use the following general process to process the A4 paper. The following two strategies can be adopted.

[0076] Page cutting strategy 1: Double columns are directly cut into 2 pages in the middle, and three columns are directly divided and cut into 3 pages; Page cutting strategy 2: Assisted cutting through marking point detection.

[0077] Step 204: Based on the answer position information corresponding to the answer sheet, determine a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information from the answer image.

[0078] The answer position information may be understood as information indicating the position of the answer area of ​​each question in the answer sheet. For example, the answer position information may be coordinate information, matching information, and the like.

[0079] The first answer area may be an image area corresponding to the first answer information; and the second answer area may be an image area corresponding to the second answer information.

[0080] In one or more embodiments provided in this specification, determining, from the answer image, a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information based on the answer position information corresponding to the answer sheet includes steps 1 and 2:

[0081] Step 1: Determine the initial answer position information corresponding to the answer sheet, and adjust the initial answer position information based on the answer sheet image corresponding to the answer sheet and the answer image to obtain the target answer position information.

[0082] The initial answer position information can be understood as the answer position information that has a positional correspondence or a layout correspondence with the answer sheet image, for example, matching information that is consistent with the layout of the answer sheet image. The target answer position information can be understood as the answer position information that has a positional correspondence or a layout correspondence with the answer image, for example, matching information that is consistent with the layout of the answer image.

[0083] Specifically, this method takes into account that the answer image corresponding to the printed answer sheet may have typesetting errors, which may cause the answer image to not match the answer position information. Therefore, the initial answer position information corresponding to the answer sheet can be determined, and the initial answer position information can be adjusted based on the answer sheet image and the answer image corresponding to the answer sheet to obtain the target answer position information that matches the answer image.

[0084] In one or more embodiments provided in this specification, the determining the initial answer position information corresponding to the answer sheet includes:

[0085] receiving an answer position information identifier corresponding to the answer image and sent by a client, and acquiring the initial answer position information from an information storage unit based on the answer position information identifier; or

[0086] The answer position information mark is identified from the answer image, and based on the answer position information mark, the initial answer position information is acquired from the information storage unit.

[0087] The answer location information identifier can be understood as information used to uniquely identify an answer location information, such as the number, name, storage address, name of the assignment or answer sheet, id of the assignment or answer sheet, etc. corresponding to the answer location information. The information storage unit can be understood as a unit used to store the answer location information, such as a database, a local disk, a cloud storage service, a server, etc.

[0088] Using the above example, this method needs to obtain the answer sheet image (answer sheet image) and matching information before determining the first answer area and the second answer area; and there are several ways to obtain the answer sheet image and matching information in the grading process:

[0089] 1. The client specifies when submitting an assignment or answer graph: The associated information of the answer graph, answer sheet graph and matching information can be directly specified by the user when submitting an assignment or answer graph (through the assignment name, assignment id, etc.).

[0090] 2. Obtain through the answer chart: Print the associated information or the encrypted and transcoded associated information in the answer sheet chart. The information (assuming it is a string) can be printed directly, or included in the answer chart in a non-explicit manner such as a QR code, barcode, or feature image (for example, a QR code), and the associated information (i.e., the answer position information identifier) ​​can be obtained by identifying the information on the answer chart.

[0091] Based on the above embodiments, it can be seen that the method determines the answer position information identifier in a variety of ways, thereby flexibly determining the initial answer position information based on the answer position information identifier, thereby improving the applicability of the method.

[0092] In one or more embodiments provided in this specification, adjusting the initial answer position information based on the answer sheet image corresponding to the answer sheet and the answer image to obtain the target answer position information includes:

[0093] Performing feature extraction on the answer sheet image to obtain multiple answer sheet area features of the answer sheet image, and performing feature extraction on the answer image to obtain multiple answer area features of the answer image;

[0094] Calculating the regional similarity between the target answer sheet region feature and the target answer region feature, wherein the target answer sheet region feature is any one of the multiple answer sheet region features, and the target answer region feature is any one of the multiple answer region features;

[0095] Based on the region similarity, a plurality of similar region feature pairs are determined from the plurality of answer sheet region features and the plurality of answer region features, wherein a similar region feature pair includes one answer sheet region feature and one answer region feature;

[0096] Based on the multiple similar region feature pairs, calculating a mapping matrix between the answer sheet image and the answer image;

[0097] Based on the mapping matrix and the answer image, the initial answer position information is adjusted to obtain the target answer position information.

[0098] Among them, the answer sheet area features can be understood as features representing local area information of the answer sheet image, for example, features of pixels in the answer sheet image; or features of a local sub-image (a local image of 2×2 pixels) in the answer sheet image.

[0099] The answer area feature can be understood as a feature representing the local area information of the answer image, for example, a feature of a pixel in the answer image; or a feature of a local sub-image (a local image of 2×2 pixels) in the answer image.

[0100] The regional similarity can be understood as a numerical value indicating the similarity between the feature of an answer sheet region and the feature of an answer area, for example, cosine similarity, distance similarity, etc.

[0101] The similar area feature pair can be understood as a feature pair consisting of similar answer sheet area features and answer area features.

[0102] The mapping matrix can be understood as a matrix used to represent the position offset state between the answer image and the initial answer position information. For example, the mapping matrix can be an affine transformation matrix.

[0103] Continuing with the above example, before segmenting the answer graph based on matching information, this method needs to calculate the mapping matrix between the answer graph and the answer sheet graph.

[0104] Considering that in order to save costs, you may be accustomed to printing the question sheet and then photocopying it. At this time, the question sheet content may be displaced or tilted, resulting in the position of the questions in the answer diagram and the answer sheet diagram not being one-to-one corresponding. Therefore, the matching information and the answer diagram cannot be directly overlapped. In order to solve this problem, the answer sheet diagram is needed as an intermediate reference to locate the matching information. The process is as follows:

[0105] 1. Calculate the features of the answer graph: This method uses HOG features to extract the region of a specific response in the answer graph and obtain the regional features of the region (i.e., the answer region features).

[0106] 2. Calculate the features of the answer sheet image: Use the same feature extractor to extract the regional features of the answer sheet image (i.e., the answer sheet regional features).

[0107] 3. Compare the extracted regional features of the answer sheet image with the region of the answer graph, select several points with the highest similarity (i.e., regional features) to calculate the affine transformation matrix. The affine transformation matrix can transform the answer sheet image to perfectly overlap with the answer graph.

[0108] 4. According to the mapping matrix, the matching information and the answer sheet are superimposed: the calculated affine transformation matrix is ​​used to transform the matching information. Since the size, proportion and layout of the matching information and the answer sheet are consistent, the transformed matching information can perfectly overlap with the answer sheet. The effect of perfect overlap is shown in Figure 3 Figure c in .

[0109] Based on the above embodiments, it can be known that the initial answer position information is adjusted by calculating the mapping matrix to obtain the target answer position information that matches the answer image, so as to facilitate the subsequent accurate determination of the first answer area and the second answer area based on the target answer position information.

[0110] Step 2: Based on the target answer position information, determine from the answer image a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information.

[0111] Specifically, the method may perform image segmentation on the answer image based on the target answer position information to obtain a first answer area image corresponding to the first answer information and a second answer area image corresponding to the second answer information.

[0112] In one or more embodiments provided in this specification, the first answer area is a first answer area image, and the second answer area is a second answer area image;

[0113] The determining, based on the target answer position information, from the answer image a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information, comprises:

[0114] Based on the target answer position information, identifying from the answer image a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information;

[0115] The answer image is segmented based on the first answer area and the second answer area to obtain the first answer area image and the second answer area image.

[0116] Among them, the first answer area image can be understood as a sub-image obtained by segmenting the answer image and corresponding to the first answer information; the second answer area image can be understood as a sub-image obtained by segmenting the answer image and corresponding to the second answer information.

[0117] Continuing with the above example, this method can segment the answer area based on the matching information: specifically, the answer area in the answer image is segmented according to the superimposed rectangular frame area. There are two types of answer areas: one is the filling area (i.e., the second answer area) of the objective question (i.e., the second answer information). It is necessary to capture the filling area of ​​each question to obtain an objective question screenshot (i.e., the second answer area image). The objective question screenshot can be subsequently input into the model, and the model will identify the filling result. The other is the answer area (i.e., the first answer information area) of the answer question (i.e., the first answer information). There is a score box area in each area, which is used to specify the score mark position and assist in attributing the score mark; by capturing the answer area, a screenshot of the answer question corresponding to the answer question (i.e., the first answer area image) is obtained.

[0118] Based on the above embodiments, it can be seen that the method obtains the first answer area image and the second answer area image by performing image segmentation on the answer image, thereby accurately classifying a large number of complex answer information, thereby facilitating subsequent accurate score recognition.

[0119] Step 206: Identify the score of the first answer information from the first answer area, and use the score evaluation model to perform score evaluation on the second answer information in the second answer area to obtain the score of the second answer information.

[0120] Among them, the score evaluation model can be understood as a model used to perform score evaluation on the second answer information. For example, the score evaluation model can be a deep learning model, a target detection model, etc.

[0121] Specifically, the method provided in this specification can use OCR technology to identify the score of the first answer information from the first answer area; and use a score evaluation model to perform score evaluation on the second answer information in the second answer area to obtain the score of the second answer information.

[0122] In one or more embodiments provided in this specification, the first answer area is a first answer area image;

[0123] The step of identifying the score of the first answer information from the first answer area includes steps 1 to 3:

[0124] Step 1: Input the first answer area image into a score recognition model, in which the score recognition model identifies the score value in the first answer area image and the score type of the score value.

[0125] Among them, the score recognition model can be understood as a model for recognizing the score of the first answer information in the first answer area image. For example, the score recognition model can be a target detection model, a YOLO model, a NasNet model, etc., and no specific restrictions are made here.

[0126] The score type may be understood as information used to indicate the type of the score value. For example, the score type may be a plus point type or a minus point type.

[0127] Specifically, the method may input the answer screenshot into the YOLO model for score recognition, thereby obtaining the score (ie, score value) and score type contained in the answer screenshot.

[0128] In one or more embodiments provided in this specification, in the score recognition model, identifying the score value and the score type of the score value in the first answer area image includes:

[0129] In the score recognition model, determining at least one score position box in the first answer area image;

[0130] Perform score recognition on the at least one score position frame to obtain the score value and the score type of the score value.

[0131] The score position frame may be understood as a position frame used to frame the score value in the first answer area image, such as a boundary frame.

[0132] Using the above example, this method can detect and identify score traces. Specifically, this method uses the YOLO training detection and recognition model to identify the answer screenshots, and obtains the bbox (i.e., score position box) and trace content of all score traces in the answer image. Because of the teacher's grading habits, it is necessary to identify the specific score (i.e., score value) in the bbox, and also to identify whether it is a scoring type or a deduction type (i.e., score type). Specifically, the following steps are included:

[0133] 1. Identification of scoring or deduction: Due to deduction, there is a minus sign [-] before the score; for scoring, there is a plus sign [+] or nothing before the score. Therefore, it is necessary to identify the plus sign and minus sign separately.

[0134] The score type corresponding to the score value is determined.

[0135] 2. Specific score identification: Identify the specific value of the score (i.e. the score numerical value).

[0136] Based on the above embodiments, it can be seen that the method can use the score recognition model to identify the score value and the score type of the score value in the first answer area image, so that the score calculation can be performed accurately later, avoiding the problem of inaccurate score calculation due to the teacher's grading habits.

[0137] Step 2: Identify the first answer information corresponding to the score value from the first answer area image.

[0138] In one or more embodiments provided in this specification, the score value is multiple, and the first answer information is multiple;

[0139] The step of identifying the first answer information corresponding to the score value from the first answer area image includes:

[0140] Determine a plurality of score position frames corresponding to a plurality of score values, and identify score writing areas corresponding to a plurality of first answer information and score writing area frames corresponding to the score writing areas from the first answer area image;

[0141] Calculating the position distance between the target score position frame and the target score writing area frame, wherein the target score writing area frame is any one of a plurality of score writing area frames, and the target score position frame is any one of the plurality of score position frames;

[0142] Based on the position distance, determining an associated score position frame for the target score writing area frame from the multiple score position frames, wherein one score writing area frame corresponds to one associated score position frame, and each score writing area frame corresponds to a different associated score position frame;

[0143] The score value corresponding to the associated score position box is determined as the score value of the first answer information corresponding to the target score writing area box.

[0144] Among them, the score writing area can be understood as the area in the answer sheet used to write scores, for example, the score writing area can be a score frame area; the score writing area frame can be understood as a position frame used to frame the score writing area in the first answer area image, such as a boundary box.

[0145] The associated score position box can be understood as a score position box that is close to the target score writing area box. For example, the associated score position box can be a score position box whose position distance to the target score writing area box is greater than or equal to a preset distance threshold; or, the score position box among multiple score position boxes whose position distance to the target score writing area box is the smallest.

[0146] Using the above example, in the process of determining the score of each answer question, this method needs to perform score trace attribution; specifically, it means attributing the score trace according to the relationship value between the score trace bbox (i.e., the score position box) and the score box bbox (i.e., the score writing area box); the specific execution steps are as follows:

[0147] 1. Relationship value calculation:

[0148] The relationship value is used to measure how close the relationship between two bboxes is. There are several ways to calculate it:

[0149] ① Calculate the distance between two bboxes (i.e., position distance), which can be the distance between corner points, the closest distance, the distance between centers, etc.

[0150] ② Calculate the IoU or various IoU variants of the bbox.

[0151] 2. Determine the validity: This method sets a threshold, and only the score traces with relationship values ​​above the threshold are considered valid.

[0152] 3. A score box (i.e., score writing area) is only allowed to have one score trace. If there are multiple valid score traces around the score box, you need to select one of them as the final result: first, sort them according to the relationship value between the score trace bbox and the score box bbox; then, take the score with a higher relationship value as the trace result of the score box (i.e., the associated score position box).

[0153] 4. Because each question has a corresponding score box, and each question has a different score definition, it is not supported that a score mark is attributed to different score boxes at the same time. When assigning scores, from top to bottom, a mark is given to the score mark that has been assigned to the score box, indicating that the score box will not participate in the attribution of the remaining question scores.

[0154] It should be noted that this method sets a fallback logic for questions that do not leave a score mark, and specifies a default score for such cases. The fallback logic used by this method is: no score + no answer = 0 points, no score + answer = full score. Whether the question has an answer = whether there is a handwritten OCR detection result within the question box.

[0155] Based on the contents of the above embodiments, it can be known that the method accurately identifies the corresponding score for each first answer information through the position distance, thereby ensuring the accuracy of the target score.

[0156] Step three: Determine the score of the first answer information based on the score value and the score type.

[0157] In one or more embodiments provided in this specification, the score type is a deduction type or a bonus type;

[0158] The determining the score of the first answer information based on the score value and the score type includes:

[0159] In the case where the score type is the deduction type, determining a score threshold corresponding to the first answer information, and subtracting the score threshold from the score value to obtain the score of the first answer information; or

[0160] When the score type is the bonus type, the score value is used as the score of the first answer information.

[0161] Among them, the score threshold can be the question score corresponding to the first answer information, such as 5 points, 10 points, etc.

[0162] Using the above example, this method takes into account the teacher's scoring habits and has different processing methods for scoring or deducting points. The specific steps include the following:

[0163] 1. When the score type is a deduction type, determine the score corresponding to the answer question (for example, 10 points), and subtract the score from the score value (for example, 4 points) to obtain the score of the answer question (for example, 6 points).

[0164] 2. When the score type is a bonus type, the identified score value (eg, 4 points) is used as the score of the answer question (eg, 4 points).

[0165] Based on the contents of the above embodiments, it can be seen that the score of the first answer information is accurately calculated in consideration of different situations of adding points and deducting points, thereby avoiding the problem of inaccurate score calculation due to the teacher's grading habits.

[0166] In one or more embodiments provided in this specification, the second answer area is a second answer area image;

[0167] The using the score evaluation model to perform score evaluation on the second answer information in the second answer area to obtain the score of the second answer information includes:

[0168] The second answer area image is input into the score evaluation model, and the second answer information in the second answer area image is identified in the score evaluation model.

[0169] Based on the second answer information and the standard answer corresponding to the second answer information, a score for the second answer information is determined.

[0170] Continuing with the above example, this method can input the objective question screenshot into the deep learning model (i.e., the score evaluation model) for score evaluation. During the processing of the deep learning model, first, the objective question answer information (i.e., the second answer information) written in the objective question screenshot can be identified. Specifically, this method can detect and identify the filling area; by obtaining the filling area position (i.e., the second answer information area), the filling area of ​​each question is intercepted, and the rectangular frame screenshot obtained by interception (i.e., the second answer area image) is identified by the model (based on deep learning model training), and the filling result type of each question (i.e., the second answer information) is obtained.

[0171] Secondly, the standard answer corresponding to the objective question answer information is determined, that is, the standard answer to the objective question. Specifically, the method obtains the answer to the fill-in-the-blank question by matching the information.

[0172] Finally, a score evaluation is performed based on the objective question answer information and the standard answer, thereby determining the score corresponding to the objective question answer information. Specifically, the detected filling result type and the answer are matched by characters to obtain the score of the filling question answer.

[0173] It should be noted that, when there are multiple objective question screenshots, the above operation is performed for each objective question screenshot, so as to determine the score of the objective question answer information in each objective question screenshot.

[0174] Based on the contents of the above embodiments, it can be known that the method adopts different score determination methods according to different types of answer information, so as to accurately obtain the score of each answer information.

[0175] In one or more embodiments provided in this specification, the second answer information is the answer information corresponding to the filling-in question;

[0176] The determining the score of the second answer information based on the second answer information and the standard answer corresponding to the second answer information includes:

[0177] Determine the filling question corresponding to the second answer information, and determine the standard answer corresponding to the filling question;

[0178] The second answer information is matched with the standard answer for consistency, and a score of the second answer information is determined based on the matching result.

[0179] The score threshold may be the question score corresponding to the second answer information, such as 2 points, 3 points, etc. The matching result is a result indicating whether the second answer information is consistent with the standard answer.

[0180] Continuing with the above example, this method can determine the objective question corresponding to the filled-in answer information (i.e., the second answer information), and determine the standard answer and question score corresponding to the objective question; then, the filled-in answer is matched with the standard answer for consistency to determine whether the filled-in answer is consistent with the standard answer, and the score of the filled-in answer information is calculated based on the matching result and the question score; thereby achieving accurate identification of the score of the filled-in question.

[0181] In one or more embodiments provided in this specification, determining the score of the second answer information according to the matching result includes:

[0182] If it is determined according to the matching result that the second answer information is consistent with the standard answer, the score threshold of the filling question is determined to be the score of the second answer information; or

[0183] When it is determined according to the matching result that the second answer information is inconsistent with the standard answer, a preset error score is determined as the score of the second answer information.

[0184] Among them, the preset error score can be understood as the score assigned to the second answer information when it is determined that the second answer information contains an answer error, such as 0 points, 1 point, etc.

[0185] Using the above example, the detected filling result and the answer are matched by characters to obtain the score of the filling question. In this process, the score judgment logic is: for multiple-choice questions and true-or-false questions, if the matching results are consistent, full marks (i.e., score threshold) are given, and if the matching results are inconsistent, zero marks (i.e., preset error score) are given.

[0186] In one or more embodiments provided in this specification, the second answer information includes at least one answer selection information, and the standard answer includes a plurality of standard answer information;

[0187] The step of determining the score of the second answer information according to the matching result includes:

[0188] In the case where it is determined according to the matching result that the at least one answer selection information is a subset of the plurality of standard answer information, calculating the score of the second answer information according to the number of the at least one answer selection information and the score threshold; or

[0189] When it is determined, based on the matching result, that the at least one answer selection information is an intersection of the plurality of standard answer information, a preset error score is determined as the score of the second answer information.

[0190] Among them, at least one answer selection information can be understood as at least one selection information for a multiple-choice question in the process of answering the multiple-choice question; multiple standard answer information can be understood as multiple correct options for the multiple-choice question.

[0191] In addition, when the method determines, based on the matching result, that the at least one answer selection information is a subset of the multiple standard answer information, the score of the second answer information is calculated based on the number of the at least one answer selection information and the score threshold for filling in the question.

[0192] Using the above example, the detected filling result and the answer are matched by characters to obtain the score of the filling question. In this process, the score judgment logic of the multiple-choice question is: if the matching results are consistent, full marks are given; if the matching results are inconsistent, it is determined whether some correct options are missing (for example, C is missing from the correct options A, B, C) in the two options A and B selected during the answering process (that is, at least one answer selection information); if so, it is determined to be half correct, and half of the score of the filling question (that is, the preset correct score) is used as the final score; if there is an incorrect result, zero points are given.

[0193] In one or more embodiments provided in this specification, after identifying the score of the first answer information from the first answer area, the method further includes:

[0194] When it is determined that the score of the first answer information is inconsistent with the score range of the first answer information, replacing the score of the first answer information with the target score to obtain the replaced score of the first answer information; or

[0195] When it is determined that the score of the first answer information is inconsistent with the score range of the first answer information, a score verification request is sent to the client, and a score replacement request sent by the client for the first answer information is received. The score of the first answer information is replaced with the target score carried in the score replacement request to obtain the score of the replaced first answer information.

[0196] The score range may be understood as the score range of the first answer information, for example, a numerical range of 0-5.

[0197] The target score may be understood as a score re-assigned to the first answer information, wherein the target score may be zero or the score threshold.

[0198] The score verification request can be understood as a request for verifying whether the score of the first answer information is abnormal; the score replacement request can be understood as a request for replacing the score of the first answer information; the target score carried by the score replacement request can be the score with which the score of the first answer information needs to be replaced, such as 1 point, 2 points, etc.

[0199] Continuing with the above example, this method will perform the operation of identifying the upper limit of the score during the score statistics process; when the teacher's score exceeds the question limit, it needs to be processed using the following two strategies.

[0200] Strategy 1: When the score or deduction exceeds the score range of the question (i.e. the score range), return full marks or zero (i.e. the target score); Strategy 2: Return the score verification request of the grading exception to the client, and ask the teacher to confirm manually; When receiving the score replacement request sent by the client, replace the score of the answer question with the target score carried in the score replacement request.

[0201] Based on the above embodiments, it can be seen that this method ensures the accuracy of the score of the first answer information by verifying whether the score of the first answer information is consistent with the score range of the first answer information.

[0202] Step 208: Based on the score of the first answer information and the score of the second answer information, determine the target score corresponding to the answer sheet.

[0203] Specifically, the score of the first answer information and the score of the second answer information are added together to obtain the target score corresponding to the answer sheet.

[0204] In one or more embodiments provided in this specification, after determining the target score corresponding to the answer sheet based on the score of the first answer information and the score of the second answer information, the method further includes:

[0205] The target score is sent to a target score acquisition unit.

[0206] Among them, the target score acquisition unit can be understood as a device that needs to obtain the target score (such as an application, a server, a service system), or a score storage device (database, disk) that stores the target score, or a score display device (such as a client, a smart terminal) that displays the target score.

[0207] Continuing with the above example, when this method obtains the final score corresponding to the answer sheet (i.e., the target score), the final score can be sent to the client for display to the user, or the final score can be sent to the database for storage, thereby achieving flexible processing of the target score.

[0208] One or more embodiments of this specification provide an image processing method. In the process of score recognition of an answer sheet, the first answer area corresponding to the first answer information and the second answer area corresponding to the second answer information can be determined from the answer image according to the answer position information corresponding to the answer sheet, so that the first answer area and the second answer area can accurately classify the answer information with a large number and complex types. In addition, the score of the first answer information is identified from the first answer area, and the score evaluation model is used to evaluate the score of the second answer information in the second answer area to obtain the score of the second answer information; different score determination methods are used in a targeted manner according to different types of answer information, so as to accurately obtain the score of each answer information. Finally, based on the score of the first answer information and the score of the second answer information, the target score corresponding to the answer sheet is accurately obtained, avoiding the problem that the score corresponding to the answer sheet cannot be accurately identified due to the large number and complex types of answer information contained in the answer sheet.

[0209] The following combination Figure 5 , taking the application of the image processing method provided in this specification in an examination scenario as an example, the image processing method is further described. Figure 5 A schematic diagram of a processing process of an image processing method provided by an embodiment of the present specification is shown, based on Figure 5 It can be seen that the execution process of this method is as follows:

[0210] Step 502: User group test paper.

[0211] In the client-side question composition service in the examination scenario (selecting questions from the online question bank to generate a homework list), a test paper document file (such as a word or PDF file) of the test paper image (containing only test questions) can be generated.

[0212] Step 504: Generate answer sheet image and matching information.

[0213] (1) Rendering the answer sheet based on the test paper layout:

[0214] Generate an answer sheet based on the questions and order selected by the user when setting the questions ( Figure 3 (b in Figure 1); Figure 3 Figure a in the answer sheet is the answer sheet; the objective questions (multiple choice questions, true or false questions, etc.) generate corresponding fill-in areas (such as Figure 3 The single-choice questions in Figure a); other answer questions generate corresponding answer areas (such as Figure 3 (For example, the second to fourth questions in Figure a of FIG. 1 ), each answer area contains a score box; the score box of this method is generated on the right side or upper right corner of each question.

[0215] (2) Rendering matching information based on test paper layout (visualization see Figure 3 Figure c):

[0216] When the answer sheet is generated, the corresponding matching information is generated. Corresponding solid color boxes will be generated at the locations of the filling area, answer area, and score box. It should be noted that the matching information is used to mark the location and area of ​​each element (filling or answer area) in the answer sheet, and is not limited to the form.

[0217] The answer sheet image and matching information are generated 1:1, and the superimposed display effect is as follows Figure 3 In addition, the answer sheet image and the corresponding matching information are stored in association in the database, and can be used to obtain relevant information of the answer image.

[0218] Step 506: The user prints the answer sheet for use and uploads the answer graph.

[0219] After completing the question setting, the user can print the test paper document file or test paper image as an answer sheet for use; the students can fill in the answers and the teacher can score the answer sheet, and then the answer image is captured from the answer sheet, and the answer image is uploaded to the server through the client device (the answer image can be in A3 or A4 paper format).

[0220] Step 508: The answer chart is converted to regular.

[0221] When scanning, the A4 answer sheet may be upside down, so it needs to be straightened. Specifically, the scenario in which schools use score collection is mainly for exams, and they are accustomed to using A3-style test papers (consisting of two pages of A4 homework, or A3 paper divided into three columns). Due to the size of the high-speed scanner, the A3-style test paper enters the high-speed scanner with the short edge when scanning, so it needs to be straightened before collection. In order to improve the collection effect and reuse the A4-style job processing flow, the A3-style test paper needs to be divided into a style similar to A4 according to the number of columns. The specific process is as follows:

[0222] Straightening means straightening the answer sheet to the direction where the text is normal. Specific strategies include: Strategy 1: Use the marks in the question sheet to determine the direction of the question sheet by detecting and identifying the marks; Strategy 2: Use OCR to obtain the text direction and determine it by the text direction; Strategy 3: Set the scanning direction when scanning with a high-speed scanner, and automatically process the straightening after scanning.

[0223] Step 510: segmentation determination.

[0224] There are two types of answer sheet styles: A4 single column and A3 multi-column (double column and triple column). For multi-column sheets, they are cut into single columns and processed uniformly, that is, they are cut into A4 sheets and processed separately.

[0225] (1) Segmentation judgment:

[0226] Because there are both A3 and A4 papers in the input, A3 paper needs to be cut but A4 paper does not, so a judgment logic needs to be added. The following two strategies can be used.

[0227] Strategy 1: Judging by the paper size, A3 paper is larger than A4 paper. Strategy 2: Judging by the paper proportions, after the conversion, A3 paper is wider than it is long, while A4 paper is longer than it is wide.

[0228] (2) Divide into two or three columns:

[0229] Cut the A3 paper into two A4 pages. Then use the following general process to process the A4 paper. Page cutting strategy 1: Double columns are directly cut into 2 pages in the middle, and three columns are directly cut into 3 pages; Page cutting strategy 2: Assisted cutting through marker point detection.

[0230] Step 512: Obtain the answer sheet image and matching information.

[0231] There are several ways to obtain answer sheet images and matching information during the grading process:

[0232] (1) The client specifies when submitting an assignment or answer graph: The association information between the answer graph, answer sheet graph and matching information can be directly specified by the user when submitting an assignment or answer graph (through the assignment name, assignment ID, etc.).

[0233] (2) Obtaining through the answer chart: Print the associated information or the encrypted and transcoded associated information in the answer sheet. The information (assuming it is a string) can be printed directly, or included in the answer chart in a non-explicit manner such as a QR code, barcode, or feature image (for example, using a QR code), and the associated information can be obtained by identifying the information on the answer chart.

[0234] Step 514: Calculate the mapping matrix between the answer graph and the answer sheet graph.

[0235] In order to save costs, schools are used to printing the test sheet and then photocopying it. At this time, the test sheet content may be displaced or tilted, resulting in the position of the questions in the answer sheet and the answer sheet not being one-to-one corresponding. Therefore, the matching information and the answer sheet cannot be directly overlapped. In order to solve this problem, the answer sheet image is needed as an intermediate reference to locate the matching information. The process is as follows:

[0236] 1. Calculate the features of the answer graph: This method uses HOG features to extract the region of a specific response in the answer graph and obtain the regional features of the region.

[0237] 2. Calculate the features of the answer sheet image: Use the same feature extractor to extract the regional features of the answer sheet image.

[0238] 3. Compare the extracted regional features of the answer sheet image with the region of the answer graph, select several points with the highest similarity (i.e., regional features) to calculate the affine transformation matrix. The affine transformation matrix can transform the answer sheet image to perfectly overlap with the answer graph.

[0239] Step 516: According to the mapping matrix, the matching information and the answer graph are superimposed.

[0240] The calculated affine transformation matrix is ​​used to transform the matching information. Since the matching information and the answer sheet are consistent in size, proportion, and layout, the transformed matching information can perfectly overlap with the answer sheet. The effect of perfect overlap can be seen in Figure 3 Figure c in .

[0241] Step 518: Segment the answer area based on the matching information.

[0242] The answer area is segmented according to the superimposed rectangular frame area. There are two types of answer areas: one is the filling area of ​​the objective question. The filling area of ​​each question needs to be captured to obtain the objective question screenshot, and the model will then identify the filling results.

[0243] The other is the answer area for answering questions. Each area has a score box area for specifying the score mark position and assisting in attributing the score mark. By intercepting the answer area, a screenshot of the answer question corresponding to the answer question is obtained.

[0244] Step 520: Detection and identification of filled areas.

[0245] The filling area position obtained through the above process is used to intercept the filling area of ​​each question, and the rectangular frame obtained by interception is recognized by the model (based on deep learning model training) to obtain the filling result type of each question.

[0246] The answer to the fill-in-the-blank question is obtained by matching the information, and the score of the fill-in-the-blank question is obtained by matching the detected fill-in-the-blank result type and the answer character by character.

[0247] In the process of obtaining scores for filling in the blanks and answering the questions, the score judgment logic for multiple-choice questions is: if the matching results are consistent, full marks are given; if the matching results are inconsistent, zero marks are given.

[0248] In the process of obtaining scores for filling in the blanks, the score judgment logic for multiple-choice questions is:

[0249] If the matching results are consistent, full marks will be awarded; if the matching results are inconsistent, it will be determined whether some correct options are missing among the two options A and B selected during the answering process (for example, C is missing from the correct options A, B, C); if so, it will be determined as half correct, and half of the score of the question will be taken as the final score; if there is an incorrect result, zero marks will be awarded.

[0250] Step 522: Score trace detection and identification.

[0251] The method used to implement score trace detection and recognition is as follows:

[0252] (1) Based on the YOLO training detection and recognition model, the bbox and trace content of all score traces in the answer graph are obtained. Because the teacher is used to asking questions, it is necessary not only to identify the specific score, but also to identify whether it is a credit or a deduction.

[0253] (2) Identification of scoring or deduction: In the case of deduction, there is a minus sign [-] before the score; in the case of scoring, there is a plus sign [+] or nothing before the score. Therefore, it is necessary to identify the plus sign and minus sign separately to determine the score type corresponding to the score value.

[0254] (3) Specific score identification: Identify the specific value of the score.

[0255] (4) Score cap: When the teacher's score exceeds the question limit, strategic handling is required.

[0256] Strategy 1: When the score or deduction exceeds the question score range, return to full score or zero.

[0257] Strategy 2: Return the score verification request of the abnormal correction to the client, and ask the teacher to manually confirm; when receiving the score replacement request sent by the client, replace the score of the answer question with the target score carried in the score replacement request.

[0258] (5) Abnormal situation handling: The teacher's correction traces or the student's answers may be misidentified and attributed to the score box. When the recognition result of the trace attributed to the score box is not a number, it is necessary to prompt that the result is abnormal. For example, it returns abnormal correction and asks the teacher to manually confirm.

[0259] Step 524: fractional trace attribution: the fractional trace is attributed according to the relationship value between the fractional trace bbox and the fractional frame bbox.

[0260] (1) Relationship value calculation: The relationship value is used to measure how close the relationship between two bboxes is. There are several ways to calculate it: ① The distance between the two bboxes, which can be the corner distance, the closest distance, the distance between the centers, etc.; ② The IoU of the bbox and various IoU variants (this method can use IoU).

[0261] (2) Determine the validity: set a threshold, and only the score traces with relationship values ​​above the threshold are considered valid.

[0262] (3) A score box is only allowed to have one score trace. If there are multiple valid score traces around the score box, one of them needs to be selected as the final result: first, sort them according to the relationship value between the score trace bbox and the score box bbox; then take the score with the higher relationship value as the trace result of the score box.

[0263] (4) Because each question has a corresponding score box, and each question has a different score definition, it is not possible to simultaneously assign a score mark to different score boxes. When assigning scores, from top to bottom, a mark is given to the score mark that has been assigned to a score box, indicating that the score box will not participate in the assignment of scores for the remaining questions.

[0264] (5) For questions that do not leave a score mark, set a fallback logic to specify a default score for this type of situation. The fallback logic used in this method is: no score + no answer = 0 points, no score + answer = full score. Whether the question has an answer = whether there is a handwritten OCR detection result within the question frame.

[0265] Step 526: Summarize the score collection results.

[0266] The score collection results are summarized, and the subsequent processes process the relevant information and return it to the user.

[0267] Based on the above steps, the image processing method of this specification provides a method for collecting answer sheets. In order to accurately collect the scores of objective questions and essay questions, and to maximize the automated collection while ensuring the accuracy of the scores of each question, this method directly collects the students' filling results for objective questions; a score box is set for essay questions to collect the teacher's scoring traces. Considering that the answer sheet exam scene is usually A3 paper (divided into two columns A4 or three columns), this method supports the judgment and segmentation of paper types, and uniformly collects the results after converting different paper types into A4 paper types.

[0268] This method automatically collects the answer results of the fill-in-the-blank questions and specifies a score box for each answer question to assist in the score mark collection. After the teacher marks the answer sheet, the student's test score can be directly collected through the system. At the same time, in order to improve the accuracy of fill-in-the-blank questions and score collection, this method uses matching information to assist detection and recognition.

[0269] Based on the above content, it can be seen that the technical effects achieved by this method include but are not limited to:

[0270] 1. Low printing cost: ordinary A3 paper can be printed.

[0271] 2. The collection procedure is simple: after students answer the questions on the answer sheet and the teacher marks and grades them, they can directly take a photo and upload it or upload it via a high-definition scanner.

[0272] 3. Filling result processing: The filling results are processed by deep learning models, which are more robust and more accurate than traditional methods.

[0273] 4. Leave a mark on the score to prevent interference: Leave a mark in the score box to eliminate interference from other information as much as possible.

[0274] 5. Convenient attribution of score traces: Each question has its own score box as the score trace area, and the traces in the area are the traces of the question. No need to set up complex attribution logic.

[0275] 6. Neat pages: The teacher’s marks are concentrated in the designated area, which looks clean and tidy.

[0276] 7. Robustness: Printers in different schools are different. The color of the fill-in area, the edge of the score box, and the background color of the printed homework are different: it may be too dark to affect the recognition of the score mark, or too light to affect the detection of the score box. In order to improve the robustness of detecting and recognizing the score box, this method uses matching information to mark the correction area, so that it can be effectively recalled regardless of whether the specified area is printed too dark or too light.

[0277] In summary, this method provides an answer sheet style, including a filling area that can automatically collect results and an answer area that needs to be collected after the teacher scores and leaves a mark; it realizes the automatic collection of answer results on the answer sheet; by setting a score frame to specify the teacher's score mark area, the score mark in the score frame is accurately detected and recognized; and matching information is used to improve the detection and recognition effects of the filling area and the score frame.

[0278] Corresponding to the above method embodiment, this specification also provides an image processing device embodiment, Figure 6 FIG. 2 shows a schematic diagram of the structure of an image processing device provided by an embodiment of the present specification. Figure 6 As shown, the device comprises:

[0279] The image determination module 602 is configured to determine an answer image of the answer sheet, wherein the answer image includes first answer information with a score and second answer information without a score;

[0280] The area determination module 604 is configured to determine, from the answer image, a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information based on the answer position information corresponding to the answer sheet;

[0281] A first score determination module 606 is configured to identify the score of the first answer information from the first answer area, and use a score evaluation model to perform a score evaluation on the second answer information in the second answer area to obtain a score of the second answer information;

[0282] The second score determination module 608 is configured to determine the target score corresponding to the answer sheet based on the score of the first answer information and the score of the second answer information.

[0283] Optionally, the second answer area is a second answer area image;

[0284] The first score determination module 606 is further configured to:

[0285] inputting the second answer area image into the score evaluation model, and identifying the second answer information in the second answer area image in the score evaluation model;

[0286] Based on the second answer information and the standard answer corresponding to the second answer information, a score for the second answer information is determined.

[0287] Optionally, the second answer information is answer information corresponding to the filling-in question;

[0288] The first score determination module 606 is further configured to:

[0289] Determine the filling question corresponding to the second answer information, and determine the standard answer corresponding to the filling question;

[0290] The second answer information is matched with the standard answer for consistency, and a score of the second answer information is determined based on the matching result.

[0291] Optionally, the first score determination module 606 is further configured to:

[0292] If it is determined according to the matching result that the second answer information is consistent with the standard answer, the score threshold of the filling question is determined to be the score of the second answer information; or

[0293] When it is determined according to the matching result that the second answer information is inconsistent with the standard answer, a preset error score is determined as the score of the second answer information.

[0294] Optionally, the second answer information includes at least one answer selection information, and the standard answer includes a plurality of standard answer information;

[0295] The first score determination module 606 is further configured to:

[0296] In the case where it is determined according to the matching result that the at least one answer selection information is the intersection of the plurality of standard answer information, a preset error score is determined as the score of the second answer information; or

[0297] When it is determined, based on the matching result, that the at least one answer selection information is an intersection of the plurality of standard answer information, a preset error score is determined as the score of the second answer information.

[0298] Optionally, the first answer area is a first answer area image;

[0299] The first score determination module 606 is further configured to:

[0300] Inputting the first answer area image into a score recognition model, wherein the score recognition model recognizes the score value and the score type of the score value in the first answer area image;

[0301] identifying the first answer information corresponding to the score value from the first answer area image;

[0302] Based on the score value and the score type, a score for the first answer information is determined.

[0303] Optionally, the first score determination module 606 is further configured to:

[0304] In the score recognition model, determining at least one score position box in the first answer area image;

[0305] Perform score recognition on the at least one score position frame to obtain the score value and the score type of the score value.

[0306] Optionally, there are multiple score values ​​and multiple first answer information;

[0307] The first score determination module 606 is further configured to:

[0308] Determine a plurality of score position frames corresponding to a plurality of score values, and identify score writing areas corresponding to a plurality of first answer information and score writing area frames corresponding to the score writing areas from the first answer area image;

[0309] Calculating the position distance between the target score position frame and the target score writing area frame, wherein the target score writing area frame is any one of a plurality of score writing area frames, and the target score position frame is any one of the plurality of score position frames;

[0310] Based on the position distance, determining an associated score position frame for the target score writing area frame from the multiple score position frames, wherein one score writing area frame corresponds to one associated score position frame, and each score writing area frame corresponds to a different associated score position frame;

[0311] The score value corresponding to the associated score position box is determined as the score value of the first answer information corresponding to the target score writing area box.

[0312] Optionally, the score type is a deduction type or a bonus type;

[0313] The first score determination module 606 is further configured to:

[0314] In the case where the score type is the deduction type, determining a score threshold corresponding to the first answer information, and subtracting the score threshold from the score value to obtain the score of the first answer information; or

[0315] When the score type is the bonus type, the score value is used as the score of the first answer information.

[0316] Optionally, the area determination module 604 is further configured to:

[0317] Determine initial answer position information corresponding to the answer sheet, and adjust the initial answer position information based on the answer sheet image corresponding to the answer sheet and the answer image to obtain target answer position information;

[0318] Based on the target answer position information, a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information are determined from the answer image.

[0319] Optionally, the area determination module 604 is further configured to:

[0320] receiving an answer position information identifier corresponding to the answer image and sent by a client, and acquiring the initial answer position information from an information storage unit based on the answer position information identifier; or

[0321] The answer position information mark is identified from the answer image, and based on the answer position information mark, the initial answer position information is acquired from the information storage unit.

[0322] Optionally, the area determination module 604 is further configured to:

[0323] Performing feature extraction on the answer sheet image to obtain multiple answer sheet area features of the answer sheet image, and performing feature extraction on the answer image to obtain multiple answer area features of the answer image;

[0324] Calculating the regional similarity between the target answer sheet region feature and the target answer region feature, wherein the target answer sheet region feature is any one of the multiple answer sheet region features, and the target answer region feature is any one of the multiple answer region features;

[0325] Based on the region similarity, a plurality of similar region feature pairs are determined from the plurality of answer sheet region features and the plurality of answer region features, wherein a similar region feature pair includes one answer sheet region feature and one answer region feature;

[0326] Based on the multiple similar region feature pairs, calculating a mapping matrix between the answer sheet image and the answer image;

[0327] Based on the mapping matrix and the answer image, the initial answer position information is adjusted to obtain the target answer position information.

[0328] Optionally, the image processing apparatus further includes a score replacement module configured to:

[0329] When it is determined that the score of the first answer information is inconsistent with the score range of the first answer information, replacing the score of the first answer information with the target score to obtain the replaced score of the first answer information; or

[0330] When it is determined that the score of the first answer information is inconsistent with the score range of the first answer information, a score verification request is sent to the client, and a score replacement request sent by the client for the first answer information is received. The score of the first answer information is replaced with the target score carried in the score replacement request to obtain the score of the replaced first answer information.

[0331] Optionally, the image determination module 602 is further configured to:

[0332] Receiving the answer image to be processed of the answer sheet sent by the client, wherein the answer image to be processed is obtained by capturing the image of the answer sheet after answering;

[0333] The image direction of the answer image to be processed is adjusted using a preset image direction, and / or the image of the answer image to be processed is segmented using a preset image size to obtain the answer sheet image;

[0334] Optionally, the image processing device further includes a score sending module configured to:

[0335] The target score is sent to a target score acquisition unit.

[0336] One or more embodiments of the present specification provide an image processing device, which can determine the first answer area corresponding to the first answer information and the second answer area corresponding to the second answer information from the answer image according to the answer position information corresponding to the answer sheet, so as to accurately classify the large number of complex types of answer information through the first answer area and the second answer area. In addition, the score of the first answer information is identified from the first answer area, and the score evaluation model is used to evaluate the score of the second answer information in the second answer area to obtain the score of the second answer information; different score determination methods are used in a targeted manner according to different types of answer information, so as to accurately obtain the score of each answer information. Finally, based on the score of the first answer information and the score of the second answer information, the target score corresponding to the answer sheet is accurately obtained, avoiding the problem that the score corresponding to the answer sheet cannot be accurately identified due to the large number of answer information contained in the answer sheet and the complex types.

[0337] The above is a schematic scheme of an image processing device of this embodiment. It should be noted that the technical scheme of the image processing device and the technical scheme of the above-mentioned image processing method belong to the same concept, and the details not described in detail in the technical scheme of the image processing device can be referred to the description of the technical scheme of the above-mentioned image processing method.

[0338] Figure 7 The block diagram of a computing device 700 according to an embodiment of the present specification is shown. The components of the computing device 700 include but are not limited to a memory 710 and a processor 720. The processor 720 is connected to the memory 710 via a bus 730, and the database 750 is used to store data.

[0339] The computing device 700 also includes an access device 740 that enables the computing device 700 to communicate via one or more networks 760. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 740 may include one or more of any type of network interface, wired or wireless (e.g., a network interface card (NIC)), such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a World Wide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, and a Near Field Communication (NFC).

[0340] In one embodiment of the present specification, the above components of the computing device 700 and Figure 7 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 7 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0341] The computing device 700 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 700 may also be a mobile or stationary server.

[0342] The processor 720 is used to execute the following computer executable instructions, which implement the steps of the above-mentioned image processing method when executed by the processor.

[0343] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the computing device embodiment, since it is basically similar to the image processing method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the image processing method embodiment.

[0344] An embodiment of the present specification further provides a computer-readable storage medium storing a computer program / instruction, which implements the steps of the above-mentioned image processing method when executed by a processor.

[0345] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the computer-readable storage medium embodiment, since it is basically similar to the image processing method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the image processing method embodiment.

[0346] An embodiment of the present specification further provides a computer program product, including a computer program / instruction, which implements the steps of the above-mentioned image processing method when executed by a processor.

[0347] The above is a schematic solution of a computer program product of this embodiment. It should be noted that the technical solution of the computer program product and the technical solution of the above-mentioned image processing method belong to the same concept, and the details not described in detail in the technical solution of the computer program product can be referred to the description of the technical solution of the above-mentioned image processing method.

[0348] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0349] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0350] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0351] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0352] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. An image processing method, characterized in that: include: Determining an answer image of the answer sheet, wherein the answer image includes first answer information with a score and second answer information without a score; Based on the answer position information corresponding to the answer sheet, determining a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information from the answer image; Identifying the score of the first answer information from the first answer area, and performing score evaluation on the second answer information in the second answer area using a score evaluation model to obtain the score of the second answer information; Based on the score of the first answer information and the score of the second answer information, a target score corresponding to the answer sheet is determined.

2. The image processing method according to claim 1, characterized in that: The second answer area is a second answer area image; The using the score evaluation model to perform score evaluation on the second answer information in the second answer area to obtain the score of the second answer information includes: inputting the second answer area image into the score evaluation model, and identifying the second answer information in the second answer area image in the score evaluation model; Based on the second answer information and the standard answer corresponding to the second answer information, a score for the second answer information is determined.

3. The image processing method according to claim 2, characterized in that: The second answer information is the answer information corresponding to the filling-in question; The determining the score of the second answer information based on the second answer information and the standard answer corresponding to the second answer information includes: Determine the filling question corresponding to the second answer information, and determine the standard answer corresponding to the filling question; The second answer information is matched with the standard answer for consistency, and a score of the second answer information is determined based on the matching result.

4. The image processing method according to claim 3, characterized in that: The step of determining the score of the second answer information according to the matching result includes: If it is determined according to the matching result that the second answer information is consistent with the standard answer, the score threshold of the filling question is determined to be the score of the second answer information; or When it is determined according to the matching result that the second answer information is inconsistent with the standard answer, a preset error score is determined as the score of the second answer information.

5. The image processing method according to claim 3, characterized in that: The second answer information includes at least one answer selection information, and the standard answer includes a plurality of standard answer information; The step of determining the score of the second answer information according to the matching result includes: In the case where it is determined according to the matching result that the at least one answer selection information is an intersection of the plurality of standard answer information, a preset error score is determined as the score of the second answer information; or When it is determined, based on the matching result, that the at least one answer selection information is an intersection of the plurality of standard answer information, a preset error score is determined as the score of the second answer information.

6. The image processing method according to any one of claims 1 to 5, characterized in that: The first answering area is a first answering area image; The step of identifying the score of the first answer information from the first answer area includes: Inputting the first answer area image into a score recognition model, wherein the score recognition model recognizes the score value and the score type of the score value in the first answer area image; identifying the first answer information corresponding to the score value from the first answer area image; Based on the score value and the score type, a score for the first answer information is determined.

7. The image processing method according to claim 6, characterized in that: In the score recognition model, identifying the score value in the first answer area image and the score type of the score value includes: In the score recognition model, determining at least one score position box in the first answer area image; Perform score recognition on the at least one score position frame to obtain the score value and the score type of the score value.

8. The image processing method according to claim 6, characterized in that: There are multiple score values ​​and multiple first answer information; The step of identifying the first answer information corresponding to the score value from the first answer area image includes: Determine a plurality of score position frames corresponding to a plurality of score values, and identify score writing areas corresponding to a plurality of first answer information and score writing area frames corresponding to the score writing areas from the first answer area image; Calculating the position distance between the target score position frame and the target score writing area frame, wherein the target score writing area frame is any one of a plurality of score writing area frames, and the target score position frame is any one of the plurality of score position frames; Based on the position distance, determining an associated score position frame for the target score writing area frame from the multiple score position frames, wherein one score writing area frame corresponds to one associated score position frame, and each score writing area frame corresponds to a different associated score position frame; The score value corresponding to the associated score position box is determined as the score value of the first answer information corresponding to the target score writing area box.

9. The image processing method according to claim 6, characterized in that: The score type is a deduction type or a bonus type; The determining the score of the first answer information based on the score value and the score type includes: In the case where the score type is the deduction type, determining a score threshold corresponding to the first answer information, and subtracting the score threshold from the score value to obtain a score for the first answer information; or When the score type is the bonus type, the score value is used as the score of the first answer information.

10. The image processing method according to any one of claims 1 to 5, characterized in that: The determining, based on the answer position information corresponding to the answer sheet, a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information from the answer image includes: Determine initial answer position information corresponding to the answer sheet, and adjust the initial answer position information based on the answer sheet image corresponding to the answer sheet and the answer image to obtain target answer position information; Based on the target answer position information, a first answer area corresponding to the first answer information and a second answer area corresponding to the second answer information are determined from the answer image.

11. The image processing method according to claim 10, characterized in that: The determining of the initial answer position information corresponding to the answer sheet includes: receiving an answer position information identifier corresponding to the answer image and sent by a client, and acquiring the initial answer position information from an information storage unit based on the answer position information identifier; or The answer position information mark is identified from the answer image, and based on the answer position information mark, the initial answer position information is acquired from the information storage unit.

12. The image processing method according to claim 10, characterized in that: The adjusting the initial answer position information based on the answer sheet image corresponding to the answer sheet and the answer image to obtain the target answer position information includes: Performing feature extraction on the answer sheet image to obtain multiple answer sheet area features of the answer sheet image, and performing feature extraction on the answer image to obtain multiple answer area features of the answer image; Calculating the regional similarity between the target answer sheet region feature and the target answer region feature, wherein the target answer sheet region feature is any one of the multiple answer sheet region features, and the target answer region feature is any one of the multiple answer region features; Based on the region similarity, a plurality of similar region feature pairs are determined from the plurality of answer sheet region features and the plurality of answer region features, wherein a similar region feature pair includes one answer sheet region feature and one answer region feature; Based on the multiple similar region feature pairs, calculating a mapping matrix between the answer sheet image and the answer image; Based on the mapping matrix and the answer image, the initial answer position information is adjusted to obtain the target answer position information.

13. The image processing method according to any one of claims 1 to 5, characterized in that: After identifying the score of the first answer information from the first answer area, the method further includes: When it is determined that the score of the first answer information is inconsistent with the score range of the first answer information, replacing the score of the first answer information with the target score to obtain the replaced score of the first answer information; or When it is determined that the score of the first answer information is inconsistent with the score range of the first answer information, a score verification request is sent to the client, and a score replacement request sent by the client for the first answer information is received. The score of the first answer information is replaced with the target score carried in the score replacement request to obtain the score of the replaced first answer information.

14. The image processing method according to any one of claims 1 to 5, characterized in that: The step of determining the answer sheet image of the answer sheet includes: Receiving the answer image to be processed of the answer sheet sent by the client, wherein the answer image to be processed is obtained by capturing the image of the answer sheet after answering; The image direction of the answer image to be processed is adjusted using a preset image direction, and / or the image of the answer image to be processed is segmented using a preset image size to obtain the answer sheet image; After determining the target score corresponding to the answer sheet based on the score of the first answer information and the score of the second answer information, the method further includes: The target score is sent to a target score acquisition unit.

15. A computing device, characterized in that: include: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer program / instructions are executed by the processor, the steps of the method described in any one of claims 1 to 14 are implemented.

16. A computer-readable storage medium, characterized in that: It stores a computer program / instruction, which implements the steps of the method described in any one of claims 1 to 14 when executed by a processor.

17. A computer program product, characterized in that The method comprises a computer program / instruction which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 14.

Images