TWI239200B - System for capturing and identifying image, method for identifying images and method for acquring additional remarks about test paper - Google Patents

Abstract

A system for capturing and identifying an image includes an image-capturing unit and an image-identifying unit. The image-capturing unit is adapted to capture a single image by a single photo. The image-identifying unit is electrically connected with the image-capturing unit, and is adapted to identify the single image and provides related assistant learning information desired by a reader, based on the identifying result. The captured image is identified, including the steps of removing the interference with light with estimation of parameters of a light curve, dividing characters, normalizing characters and comparing features. The system can be used for capturing the information about a question of a test paper. A question of the test paper has a mark-code image. After captured, the mark-code image can be identified to attain mark-code characters. Based on the mark-code characters, related information is provided.

Description

1239200 Case No. 93100029 Amendment V. Description of the Invention (1) [Technical Field to which the Invention belongs] The present invention relates to a document acquisition recognition system and method, and more particularly to a method that can capture a single document image and can Document acquisition recognition system and method for recognizing single document image. [Previous technology] No matter the past learners or today's learners, most people still mainly read the contents of physical plane books. This is because the traditional plane books are physical and the eyes look more comfortable. Reading flat books is more in line with the natural reading habits of human beings, and easy to think. Although traditional print books have many advantages, they also have a serious disadvantage. Traditional paper print books are static. Their content is fixed after leaving the printing factory and cannot be added. When the learner cannot see it somewhere When understood, traditional print books will not provide further explanation in time to assist learners. In view of the limitation of the traditional print books, it can not provide readers with relevant information learning quickly, and can only be limited to the text on the paper when reading, which is also likely to cause poor learning results. The breadth will be inadequate. Therefore, many manufacturers have developed some related e-learning products, but the functions and user convenience are not ideal. Next, we will discuss the use of e-learning products developed by several manufacturers and their shortcomings. 1. As far as E Β 00 Κ 0 2 of Taiwan Ji Ding Technology Co., Ltd. and Leap P ad e-book products of Baoer Technology Co., Ltd. are concerned, it is a printed paper card or a thin flat book It is placed on a plastic hardware platform of the same size as the book. The user holds a stylus to hold the picture on the paper card, so that

11948twf1.ptc Page 9 1239200 _Case No. 93100029_Year Month__ V. Description of the invention (2) The plastic hardware platform detects the position of the pen tip with pressure-sensitive Fang Wu (or electromagnetic change), thereby playing out the commentary voice. The disadvantages are: 1. The plastic hardware platform must be carried at all times to read, which is inconvenient to carry. 2. Books must be placed horizontally on a table to operate. Do not pick them up for comfortable reading. 3. The compilation of books is very limited. The length and width of the book are fixed. The size of the book must match the size of the plastic hardware platform. It can only be used for specific books. 4. These products can only play compressed voice. After decompression, not only the sound quality is poor, but also the capacity is very small, which cannot provide multi-media effects. Generally, it is only suitable for children. 2. As far as Bar Code Reeader is concerned, it prints the bar code on a flat book. The user uses a bar code reading pen to scan the bar code and identify the representative from the bar code. Storage location of the auxiliary teaching material file, so as to play the auxiliary teaching material. The disadvantages are as follows: 1. Learners must use their wrists to read the correct entire barcode with a consistent horizontal direction and stable scanning speed. In order to increase the power, these products will require the user to move the pen. If the user's posture is poor, it will cause frequent error rate and produce erroneous information. Therefore, it must be scanned repeatedly, which is easy to cause frustration and is not ideal for operation. 2. The book should be printed with many huge and long rectangular bar codes. These bar code symbols are not only meaningless to the user, but also destroy the aesthetic and texture of the design on the pages of the book, only causing discomfort.

11948twf1.ptc Page 10 1239200 Amendment For toddlers with a lot of comments, print a lot of long solutions between the lines. Lay flat on the table and scan for ‘cause reading. The C-Pen optical scanning pen of the company's C-Pen optical scanning pen printed on a flat book continues the user's pen-scanning process to identify the scanned image as a check-word, document editing, and press the pen body flat on the book using the wrist Sentence a whole line of sentences. A certain gesture for improving the writing skills of the case No. 93100029 V. Description of the invention (3) Suitable for children's books with less annotations. On books, it is difficult to write a narrow square bar code in a book article to provide the necessary notes for the learner. 3. Generally, this type of product requires books to have better operating effects. They cannot be picked up. An optical scanning pen is used to scan the characters. It is to connect the characters that pass through at a speed of dozens of times per second. From the multiple connected characters obtained, enter these characters into the computer, fill out the forms, etc. application. The disadvantages are as follows: 1. The user must first hold the pen and hang it down (the square pen tip cannot leave the paper surface), and the direction and stable speed can scan the correct power. The C-Pen optical scanning pen requires a posture. If the user's posture is poor, it will cause frequent error rate and produce wrong messages, so it must be scanned repeatedly, which is easy to cause frustration and is not ideal for operation. 2. Since the user must align his square pen pen vertically to the first character of the place to be scanned, and then run the pen to scan, the C-Pen optical scanning pen is mainly designed to scan a whole Long lines of strings (to save users from having to type one by one), not suitable for scanning a word or word in a dense book. 4. The Meng Tian Scanning Pen of Taiwan Meng Tian Technology Co., Ltd. and Wiz C 〇 m

11948twf1.ptc Page 11 1239200 Case No. 93100029 Date Amendment V. Description of Invention (4)

Technologies' Super Pen, both use the CIS linear roller optics of ordinary desktop scanners to obtain images. It uses an optical scanning pen to scan text printed on flat books. Multiple linear images are continuously obtained from the & scanning process. These linear image groups i Z are used to identify the scanned text, and these texts are entered into computer search words, document editing, and form filling applications. The operation method is the same as that of the C-Pen optical scanning pen of C-Technology, Sweden, and the disadvantages and inconveniences are similar. 5. Chatp⑼ (chat pen) developed by Sweden's Ericsson company needs to be matched with a special paper printed with a special pattern such as the return point n L tt _ off point moment pattern to capture and identify the user's direct writing on this special paper The manuscript and input to the computer 0 Chat Pen must be applied in a dedicated secret for the inconvenience. It can only be operated on the paper, which is added by Xu. As mentioned above, the current type input scanner on the market is performing document fetching S = = f academic system. The pen-line environment used has no character rotation and characters. ; When operating, must have high quality recognition rate in good light. Therefore, under the circumstance of photographing and oblique oblique, the significant impact can be achieved, which also causes the user ^ = potential to have a lot of inconvenience to the quality of the scan recognition rate. When using these products, there is a lot of content. [Summary of the Invention] In view of this, the ~~ identification method of the present invention uses a digital camera ... is to provide a document capture image, the user does not need to move the pen to sweep the mouth, = Zhang capture Retrieving documents in a way improves the chance of successful identification. It is worth, 1 i to simplify the operating procedures, and it is conceivable that the

1239200 _Case No. 93100029_ 年月 日 _ 修 __ Five. Description of the invention (5) The digital camera optical pen does not require books to be captured on the desktop to capture images, and users can lean on the sofa, when When you pick up a flat book and read it comfortably, you can use the digital camera optical pen to capture text, numbers, or codes on the book, which is very convenient for operation. Another object of the present invention is to provide a method for document image recognition. The digital camera optical pen can be used with Bluetooth wireless technology to allow users to sit and lie anywhere in the home to read books at any time. Get a variety of high-day quality electronic media commentary to make reading a flat book a delightful and exquisite experience. Another object of the present invention is to provide a method for recognizing a document, which uses a digital camera optical pen that can be used with a computer. The digital camera optical pen can capture an annotated code graphic image on a book, and analyze and compare the image and text image to identify the information contained in it (such as page number and / or sentence code). Relevant files corresponding to this information (such as page numbers and / or sentence codes) are retrieved from the database device and used for other purposes; or one of the auxiliary files corresponding to this information (such as page numbers and / or sentence codes) is assisted The textbook file is played out to help the user to understand the content of the physical plane book marked by the comment code more effectively to achieve the purpose of assisted learning. In addition, another object of the present invention is to provide a new personal examination practice environment. The document acquisition method and digital camera optical pen of the present invention can be used with various computer learning aids to apply answers on specially designed paper test papers so that students' paper answers and learning aids Interactive multimedia textbooks and simulated test questions to achieve personalized test practice

11948twfl.ptc Page 13 1239200 _Case No. 93100029_ Year Month Day __ 5. Description of the invention (6) Learning environment. In this application design, it is mainly to strengthen the mechanism developed by students to obtain relevant information in a timely and relevant manner during the exam. In view of the limitation of the length of traditional paper-based test questions, it is impossible to provide relevant information learning of the testees, and students can only be limited to the text on the paper during the test, which may also lead to poor learning results and understanding of the problem. The depth and breadth are insufficient, and there is a gap in the learning of knowledge. At the same time, in the traditional paper test, the answer information established by the learner cannot be automatically recorded in real time and analyzed to further give the learner appropriate assistance. Therefore, the system design of the new personal examination practice environment is proposed. Its main technical features are: 1. It combines the advantages of paper-based test questions and electronic format files to achieve a wealth of information that cannot be provided by paper-based test questions. 2. This new type of test is dynamic and can be explored in more depth on the part of the test that is inadequately practiced and understood, and prepare the next personalized test paper based on the status of the tester's response. 3. Since the new quiz application uses the digital camera optical pen to link to the database to extract the relevant test question information, the establishment of the database will be multifaceted. 4. Teachers and parents can also use this application to obtain the answer information and analysis established by the learners on the learning aids to further give the learners appropriate assistance. In order to achieve the above object of the present invention, the present invention proposes a document acquisition recognition system, which is suitable for the operation of acquiring and recognizing a document. The document acquisition recognition

11948twf1.ptc Page 14 1239200 _Case No. 93100029_Year Month Day__ V. Description of the Invention (7) The system includes at least an image capture unit and an image recognition unit. The image capture unit is suitable for capturing a single image of a document, and the image recognition unit is electrically connected to the image capture unit. The image recognition unit is suitable for identifying a single image and can provide relevant information according to the recognition result. According to a preferred embodiment of the present invention, the information of a single image can be transmitted to a computer host using wireless communication or wired communication, for example. The image capturing unit is, for example, a digital camera, and the document is, for example, an examination paper, a book, a print medium, or a graphic image on an information hardware screen. The method for identifying an image of the present invention includes the following steps: 1) performing the steps of ray surface parameter estimation and removing light effects on the captured image to obtain a modified image; 2. the character cutting step of the modified image To obtain a character image; 3. a step of character normalizing the character image to obtain a normalized character image; 4. a step of feature comparison of the normalized character image. It is worth noting that, when the captured image includes identifiable characters, before the step of feature comparison, the step of linearly transforming the normalized character image by linear discrimination analysis is performed; and when the captured image includes When the image includes special symbols, before performing the feature comparison step, the normalized character image must be converted into a feature code, where these special symbols conform to the definition of Hadamard Code. The above-mentioned steps for estimating the light surface parameters and removing the light effects on the captured image include: 1. calculating the six surface parameters of the quadric surface using the minimum mean square error; 2. reconstructing a standard quadric surface using these surface parameters 3. Subtract the standard quadric surface from the image to get a corrected image; 4. Use equalization techniques to improve the contrast of the corrected image.

11948twf1.ptc Page 15 1239200 _ Case No. 93100029_ Year Month ___ V. Description of the invention (8) The above steps of character normalizing a character image to obtain a normalized character image include: 1. From the image Calculate a common variation matrix at the coordinates of each point on the point; 2 Calculate multiple eigenvalues and multiple original eigenvectors of the common variation matrix; 3. Divide the eigenvector by the corresponding eigenvalue to obtain multiple Normalize the feature vector; 4. Coordinate transform the image with the normalized feature vector to form a feature image; 5. Calculate a rotation angle by the complex momentum method; 6. Generate a rotation matrix by the rotation angle, and use rotation The matrix rotates the feature image to a standard angle to obtain a normalized character image. In order to make the above-mentioned objects, features, and advantages of the present invention more comprehensible, a preferred embodiment is given below and described in detail with the accompanying drawings as follows: [Embodiment] Figures 1A and 1B It is a schematic diagram showing a document image recognition system according to a preferred embodiment of the present invention. Please refer to FIG. 1A and FIG. 1B. The present invention proposes a document image recognition system. The digital camera optical pen 1 1 0 adopted by the present invention can be used with, for example, a computer host 1 2 0, so as to achieve a flat book. The purpose of assisted learning. The digital camera optical pen 1 1 0 includes an image capture unit 1 12 and an image recognition unit 1 1 4 as shown in FIG. 1A, and the image recognition unit 1 1 4 can also be installed on the computer host 1 2 0 Inside, as shown in Figure 1B. The image capture unit 1 12 is used to capture the static image of the annotation code printed on a document 130, where the document 130 is, for example, a text image on an examination paper, a book, a flat media or an information hardware screen . The image recognition unit 1 1 4 is electrically connected to the image capture unit 1 1 2 and performs an image processing program,

11948twfl.ptc Page 16 1239200

Case No. 93100029 _ I 丨 V. Description of the invention (9) A graphic analysis program and text recognition program. Please refer to FIG. 1A. The image capturing unit 112 is, for example, _M ', which uses the single-photographing method to first capture the digital phase 2 digital image recognition unit in the camera optical pen U0. After the Baier image, the page number image can be identified, and the page number characters can be output to the host computer by means of wired transmission. The hot and 'spring transmitter' is used as the image capture unit 1 1 2 For example, using a single sheet to extract the Zhao Xiangfang, ΐ a ΐ, and after receiving the sentence code image i, the image recognition unit 114 can recognize the sentence code image, and can use the output method of the Yi 3 image to output The sentence code characters are stored in the host computer. , Wheel or wired or, please refer to Figure 丨 B. After capturing the page number image by the photographic method of image capture, it is convenient to output the page number image to the host computer by single-line transmission using the helmet. 2 The image recognition unit i 14 in the host 120 receives the page number image, recognizes the page number image on the computer, and outputs a page number character. You can use 1 1 2 to capture a receiver by taking a single photo. The image capture can use wireless or wired transmission σ code @, at this time host 12 " is located on the computer host 12〇 image U maps to the computer: After the sentence code image, this sentence can be recognized; out, = material 廑 ΐ ϋ: characters and sentence code characters, τ can be found by the computer host 1-20 Corresponding auxiliary teaching material files are played out, and U Bei users can more effectively understand the marked place ^ = f to assist auxiliary learning purposes. Shellfish Thousands of Facebook Straw Content

11948twf1.ptc Page 17 1239200 _Case No. 93100029_ Year Month Day__ V. Description of the Invention (10) As mentioned above, the present invention proposes a method for document image recognition. First, a document must be provided, and the document has at least one image. , Then use a digital camera to capture the image of the document in a single capture mode, and then you can identify the image. It is worth noting that the captured images can be identifiable characters, such as Arabic numerals, English letters, punctuation marks, mathematical operation symbols, icons, geometric figures (such as square / rectangle / triangle, etc.) or Chinese characters. . However, the traditional recognizable character set is not sufficiently different between different characters. Therefore, the recognition result may be wrong because the characters are too similar, so you can capture images of special symbols with large differences , To improve the recognition rate, where the special symbol is, for example, in line with the definition of the Hadamard code. Hadamard code is a type of error correction code. For a code of length η, the Hamming D istance between any two codes is n / 2, that is, between any two codes. The degree of difference is the largest. Samples with large differences (P a 11 e r η) can reduce the probability of misjudgment, and the Hadamard code can be expressed in the form of two-dimensional samples to improve the system recognition rate. For more details about the Hadamard code, please refer to G w 〇- J ο ng Yu, C hu η-S hie η L u, Hong-Yuan Mark Liao, n A Message-based cocktail watermark i ng system ", in Pattern Recognition, vol. 36, pp. 9 5 7-9 6 8, 2003. All the contents disclosed in the above information can be used as reference materials in this case. For a correspondence example between the recognizable character set and the special symbol character set, refer to FIG. 2. Please refer to FIG. 3, which illustrates a flowchart of image recognition for recognizable characters according to the present invention. Captured as a single frame with a digital camera

11948twf1.ptc Page 18 1239200 Case No. 93100029 Revised May 15th, Invention Description (11) After the identifiable image (S 2 0 1) of the document, the light surface parameter estimation and the removal of light effects can be performed on the image. Steps to obtain a corrected image (S 2 0 2). Then, a character cutting step may be performed on the modified image to obtain a character image (S 2 0 3). Next, a character normalization step may be performed on the character image to obtain a normalized character image (s 2 0 4). Then, a step of linearly transforming the normalized character image can be performed by linear discrimination analysis (S205). Then, a feature comparison step (S 2 0 6) may be performed on the normalized character image after linear conversion. Figure ′ shows its edge for a special symbol character according to the present invention. After using a digital camera to capture the f special symbol image of the file f in a single capture (S301), it is possible to estimate the number of lines and surfaces and remove the effects of light ^ = image ⑻. … Next, the corrected image can be taken; line: take: = step to obtain a character image (S303). The step of performing character normalization on the y of the vice-knife to obtain a positive y = the character image (S 3 0 4). Next, the Zhengdan Yuanzi Yuan image (S 3 0 5). Then, a step of converting the image to the feature code and the feature comparison can be performed (S 3 0 6). The normalized character image of I and I codes is performed on the discriminated character discs ^ In the course, it will be performed on the image 2:: τ Yuan's image recognition has identified the irradiation side 4 which restricts light, so it can be used when There is not a strict background that cannot be corrected due to insufficient contrast, 2 may cause the text to be cut to the right, which can correct the difference in light sources.

11948twf1.ptc

1239200 _Case No. 93100029_ 年月 日 __ V. The effect of the description of the invention (12) will help the text to be cut correctly. In the case of a point light source, the brightness change projected on a plane is a quadric surface, which can be described using six parameters. Least Mean Square Error is used to estimate the six surface parameters that have the smallest difference between the surface model and the image, and then subtract the influence of the quadric surface from the image to obtain a sheet that is less affected by light. Affected image. Please refer to FIG. 5, which illustrates a flowchart of estimating a light surface parameter and removing a light effect on an image according to the present invention. First, the six surface parameters of the quadric surface are calculated based on the minimum mean square error for the input image (S401); then, a standard quadric surface is reconstructed using these surface parameters (S 4 0 2); Remove the standard quadric surface to obtain a corrected image (S 4 0 3); then, use the equalization technique to improve the contrast of the corrected image (S 4 0 4). In this way, an image that is less affected by the light source can be obtained. In the above image recognition process for identifiable characters and special symbol characters, the character normalization step is performed. The reason is that the image is affected by the distance between the digital camera optical pen and the paper, the angle and tilt of the image, or the difference in the size of the characters on the paper. Or scaling issues. Through the character normalization process, characters can be converted into characters with the same orientation, size, and aspect ratio, which helps to improve the recognition rate. Please refer to FIG. 6, which shows a flowchart of the character normalization step according to the present invention. And character normalize the character image for normalization

11948twfl.ptc Page 20 1239200 _Case No. 93100029_ Year Month Revision_ V. Description of the invention (13) The steps of character image include: 1. Calculate a total variation matrix from the coordinates of each point on the image (S 5 0 1); 2. Calculate multiple eigenvalues of the covariance matrix and multiple original eigenvectors (S 5 0 2); 3. Divide each eigenvector by the corresponding eigenvalue to obtain multiple normals Eigenvectors (S 5 0 3); 4 · coordinate transformation of the image with normalized eigenvectors to form a characteristic image (S 5 0 4); 5 · calculate a rotation angle (S 5 0 by complex momentum method) 5); 6 · Generate a rotation matrix at the rotation angle, and rotate the feature image into a standard angle by the rotation matrix to obtain a normalized character image (S 506); 7. Finally, the character image can be Adjust to standard size and standard rotation angle (S 5 0 7). For details on character normalization, please refer to Soo-Chang Pei and Chao-Nan Lin, MImage normalization for pattern recognition " in Image and Vision Computing, vo 1.13, no. 10, pp. 7 1 1-7 2 3 , 1 9 9 5; Dinggang Shen and

Horace H. S. lp, " Generalized affine invariant image normalization, 丨 丨 in IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 19, no. 5,

May, 1997. All the contents disclosed in the above information can be used as reference materials in this case. In the above-mentioned image recognition process for recognizable characters, a linear conversion analysis step is used to perform linear conversion. By performing a linear conversion step on the feature vectors, the differences in the characteristics of the same type of letters can be made smaller, and different types of letters can be different. The greater the difference in the characteristics of the letters, the higher the recognition rate of the system. Refer to the details of the linear transformation step by linear discrimination analysis for reference

11948twfl.ptc Page 21 1239200 Case No. 93100029 Amendment V. Description of Invention (14)

Li-Fen Chen, Hong-Yuan Mark Liao, M i ng-Ta t Ko, J a -Chen Lin, and Gwo-Jong Yu, MA new LDA-based face recognition system which can solve the small sample size problem ", in Pattern Recognition, vol · 33, pp. 1 7 1 3-1 7 2 6, 2 0 0 0 〇 In terms of the application of the present invention, for example, the above-mentioned document retrieval system can be used in combination with examination papers, so that students can easily To obtain information related to the test form to increase the depth and breadth of students' learning. Compared to — • 1 · 1 1 ....... 1% A-like test paper, students can only know the answer to the test question, if you want to know other, detailed explanation, you must spend a lot of extra time It can only be found in the relevant reference check ^, which will reduce the learning efficiency of students. 'The purpose of this document is to more clearly illustrate the use of the document acquisition identification system of the present invention as a test-assisted learning textbook'. Refer to FIG. 7, which illustrates the use of the document acquisition identification system of the present invention to obtain examination paper information. First provide a test paper (S601), when the student encounters a picture of the question that will not be. You can use the optical pen to aim at the question number and / or option f. ^ ′ Then capture the image of the question number ^ D ′ through a digital camera in a single operation (S 602). Then use the aforementioned image item and / or option image to obtain the title character and / 5 5 T J f image code character (S 6 0 3). According to the title character and /. ≪ note to gradually familiarize with this type of character, you can use the computer host database to find 3 ^ Item find # of the ticket ticket code textbook file, and play it out to assist the use ^ The principle of the ancient project subsidy, or even if needed, supplemented with other similar practical questions related to η this test question, to facilitate students to ^ ° column *

11948twf1.ptc 1239200 _Case No. 93100029_Year Month and Day__ V. Description of the invention (15) to achieve the purpose of assisted learning (S 604); At the same time, the user uses the digital camera optical pen in the paper test The selected answers can also be sent to the host computer in real time, and the host computer will automatically record the answer information created by the learner and analyze it to further give the learner appropriate assistance. To sum up, the present invention has at least the following advantages: 1. The document image capturing identification method of the present invention uses a digital camera to capture the image of a document in a single capture mode, and the user does not have to run a pen to scan, which can simplify the operation Procedures, and can increase the probability of successful identification. It is worth noting that the digital camera optical pen used in the present invention does not require books to be captured on the desktop to capture images, and users can lean on the sofa and pick up a flat book in a comfortable place. When reading, the digital camera optical pen can capture the text, numbers or codes on the book, which is very convenient for operation. 2. The document image capturing identification method of the present invention, which uses a digital camera optical pen, can be used with Bluetooth wireless technology, allowing users to sit and lie anywhere in the home to read books at will, and obtain a variety of attitudes at any time High-quality electronic media commentary makes reading a flat book a pleasing and exquisite enjoyment. 3. The document acquisition method of the present invention, which uses a digital camera optical pen, can be used with a computer. The digital camera optical pen can capture a text image of the annotation code on a book, and analyze and compare the graphic image to identify the page number and sentence code contained in it, and then retrieve it from a database device Corresponding page number and sentence code

11948twf1.ptc Page 23 1239200 _ Case No. 93100029_ Year Month Day _ Amendment _ V. Description of the Invention (16) and play it out to help users more effectively understand the content of the physical plane book marked by the comment code, To achieve the purpose of assisted learning. 4 · The document image acquisition identification method of the present invention can be applied to obtaining information of examination papers, wherein the examination papers have at least one label code, and the image acquisition identification system of the present invention can capture images of the marked points of the examination papers and identify Annotate the image of the code, and output a code character, and then provide related auxiliary learning information based on the code character. The invention enables students' paper answers to interact with multimedia teaching materials and simulated test questions on learning aids, so as to realize a new type of personalized test practice environment and creatively improve the traditional paper-based test methods. Although the present invention has been disclosed as above with a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make some changes and retouch without departing from the spirit and scope of the present invention. The scope of protection of the invention shall be determined by the scope of the attached patent application.

11948twf1.ptc Page 24 1239200 _ Case No. 93100029 __ Brief Description of Drawings Figures 1A and 1B show the unintended intention of the image acquisition identification system according to a preferred embodiment of the present invention. Figure 2 shows an example of the correspondence between the recognizable characters and the special symbol character set. FIG. 3 is a flowchart of image recognition for recognizable characters according to the present invention. FIG. 4 shows a flowchart of image recognition for special symbol characters according to the present invention. FIG. 5 shows a flowchart of estimating a light surface parameter and removing a light influence on an image according to the present invention. Fig. 6 shows the flow of the character normalization step according to the present invention. Fig. 7 shows the flow chart of obtaining the examination paper information using the document acquisition system according to the present invention. [Illustration of graphic labeling] 110 digital camera optical pen 112 image capture unit 114 image recognition unit 120 computer host 130 files

11948twf1.pt c p. 25

Claims (1)

1239200 _Case No. 93100029_ 年月 日 __ VI. Scope of patent application 1. A document acquisition recognition system, which is suitable for image recognition of a document. The document acquisition recognition system includes at least: an image capture A unit that captures a single image of the document in a single photograph; and an image recognition unit that is electrically connected to the image capture unit, the image recognition unit is adapted to recognize the single image and according to the recognition result Provide related information, wherein the method for identifying the single image by the image recognition unit includes: performing light surface parameter estimation and removing light effects on the single image to obtain a corrected image; and performing character cutting on the corrected image to obtain a character Image; character normalize the character image to obtain a normalized character image; and perform feature comparison on the normalized character image. 2. The document image recognition system according to item 1 of the scope of patent application, wherein the image capture unit and the image recognition unit are arranged in an optical pen. 3. The document image recognition system as described in item 2 of the scope of patent application, wherein one of the two methods is wireless communication and wired communication, and the information of the single image is transmitted from the image recognition unit to a computer. Host. 4. The document acquisition system according to item 1 of the scope of patent application, wherein the image acquisition unit is arranged in an optical pen, and the image recognition unit is arranged in a computer host. 5. Document acquisition system as described in item 4 of the scope of patent application, 11948twf1.ptc Page 26 1239200 _ Case No. 93100029_ Year Month Date _iMz_ VI. The scope of patent application Among them are wireless communication and wired communication, one of which is to extract the information of a single image from the image The unit is transmitted to the image recognition unit. 6. The document acquisition identification system described in item 1 of the scope of patent application, wherein the image acquisition unit is a digital camera. 7 · The document acquisition system as described in item 1 of the scope of patent application, wherein the document is a paper examination paper. 8. The document acquisition system according to item 1 of the scope of patent application, wherein the document is a paper book. 9 The document acquisition system as described in item 1 of the scope of patent application, wherein the document is a print medium. 10. The document acquisition system as described in item 1 of the scope of patent application, wherein the document is a graphic image on an information hardware screen. 1 1. A recognition method suitable for recognizing an image captured from a document, the recognition method includes at least: performing a light surface parameter estimation and removing a light effect on the image to obtain a modified image; and the modified image Performing character cutting to obtain a character image; performing character normalization on the character image to obtain a normalized character image; and performing feature comparison on the normalized character image. 12. The identification method as described in item 11 of the scope of patent application, wherein when the image includes at least one identifiable character, a linear discrimination analysis method is used before performing a feature comparison on the normalized character image. The normalized character 11948twf1.ptc Page 27 1239200 _Case No. 93100029_Year Month_Charm_ VI. Patent Application Scope The image is linearly converted. 1 3 · The identification method as described in item 11 of the scope of patent application, wherein when the image includes at least one special symbol, before performing the feature comparison on the normalized character image, the normalized character image is also Into a feature code. 14. The identification method as described in item 13 of the scope of the patent application, wherein the special symbol conforms to the definition of a Hadamard code. 15. The identification method as described in item 11 of the scope of patent application, wherein the steps of estimating light surface parameters and removing light effects for the image include: calculating a plurality of surface parameters of a quadratic surface with a minimum mean square error; The surface parameters reconstruct a standard quadric surface; and subtracting the standard quadric surface from the image to obtain the modified image. 16. The identification method as described in item 15 of the scope of patent application, wherein after obtaining the modified image, it further includes using an equalization technique to improve the contrast of the modified image. 1 7. The identification method according to item 11 of the scope of patent application, wherein the character normalization of the character image to obtain the normalized character image includes: the coordinate position of each point on the image Calculate a common variation matrix; calculate multiple eigenvalues and original feature orientations of the common variation matrix, divide the feature vectors by the corresponding feature values, and obtain multiple normalized features respectively Vector; coordinate transformation of the image with the normalized feature vectors to form a 11948twf1.ptc Page 28 1239200 _ Case No. 93100029_year month_iMz_- 6. When applying for a patent with at least one identifiable character, before performing a feature comparison on the normalized character image, a linear discrimination analysis is also performed. Method to linearly transform the normalized character image. 2 1. The method for obtaining examination paper information by using a document acquisition recognition system as described in item 19 of the scope of patent application, wherein when the image of the markup code includes at least one special symbol, the normalized character image is processed. Before the feature comparison, the normalized character image is also converted into a feature code. 2 2. The method for obtaining examination paper information by using a document acquisition system as described in item 21 of the scope of patent application, wherein the special symbol conforms to the definition of a Hadamard code. 2 3. The method for obtaining examination paper information using a document acquisition recognition system as described in item 19 of the scope of patent application, wherein light surface parameter estimation and light effect removal are performed on the labeled image to obtain the corrected image. The steps include: calculating a plurality of surface parameters of a quadric surface with a minimum mean square error; reconstructing a standard quadric surface with the surface parameters; and subtracting the standard quadric surface from the image to obtain the modified image. 2 4. The method for obtaining examination paper information using a document acquisition system as described in item 23 of the scope of patent application, wherein after obtaining the corrected image, the method further includes: using an equalization technique to improve the contrast of the corrected image. 25. The method for obtaining examination paper information using a document acquisition system as described in item 19 of the scope of patent application, wherein the character normalization of the character image to obtain the normalized character image includes: 11948twf1.ptc Page 30 1239200 _Case No. 93100029_Year_Month__ Sixth, the scope of patent application is calculated from the coordinate position of each point on the image to calculate a common variation matrix; calculate the multiple eigenvalues and multiple of the common variation matrix Original feature vectors; dividing the feature vectors by corresponding feature values to obtain a plurality of normalized feature vectors respectively; performing coordinate transformation on the image with the normalized feature vectors to form a feature image Calculate a rotation angle using the complex momentum method; and generate a rotation matrix with the rotation angle, and rotate the feature image to a standard angle by using the rotation matrix to obtain the normalized character image. 26. The method for obtaining examination paper information using a document acquisition and identification system as described in item 18 of the scope of patent application, wherein the document acquisition and identification system includes: an image acquisition unit adapted to acquire the tag code An image; and an image recognition unit, which is electrically connected to the image capture unit. The image recognition unit is adapted to recognize the image of the label and provide related information according to the recognition result. 2 7. The method for obtaining examination paper information using a document acquisition recognition system as described in item 26 of the scope of patent application, wherein the image acquisition unit and the image recognition unit are arranged in an optical pen. 28. The method for obtaining examination paper information using the document acquisition system as described in item 27 of the scope of patent application, wherein the method uses wireless communication and wired communication to select one of the two methods to obtain the information of the single image. Distinguish from this image 11948twf1.ptc Page 31 1239200 _ Case No. 93100029_year month_iMz_ VI. Patent Application Scope The identification unit is transmitted to a computer host. 29. The method for obtaining examination paper information using a document image recognition system as described in item 26 of the scope of patent application, wherein the image capture unit is configured in an optical pen and the image recognition unit is configured in a Computer host. 30. The method for obtaining examination paper information using a document acquisition system as described in item 29 of the scope of patent application, wherein wireless communication and wired communication are used to select the information of the single image. Transmitting from the image capturing unit to the image recognition unit. 31. The method for obtaining examination paper information using a document acquisition system as described in item 26 of the scope of patent application, wherein the image acquisition unit is a digital camera. 11948twf1.ptc Page 32