[0032] Example 1:
[0033] Such as figure 1 As shown, the questions involved in the homework book are subjective questions. Each page of the homework book has barcode area 1, student information area 2, subject 3, homework question 4, answer area 6, grading column area 5, and student information area. 2 Including student's name, class and student number, etc.; each subjective question corresponds to a grading column area; the barcode area is set with a bar code, which contains the classification information of the homework, such as chapters, sections, knowledge points, etc.
[0034] Operating this system, such as figure 2 As shown, it includes workbooks, smart terminals, image acquisition devices, image processing devices, and database servers; the image acquisition device is used to collect images on the workbook and save the collected workbook images to the database server; the image acquisition device can be a scan Instrument or camera. The smart terminal can be a mobile phone or a tablet computer or a PC.
[0035] In embodiment 1, the questions involved in the workbook are subjective questions. After the students answer the questions in the answering area, the teacher reviews them first, and then the image acquisition device collects the images of each page of the workbook. The image processing device Recognize the barcode in the workbook collected by the image acquisition device, obtain the work barcode information, locate and segment the image information in the workbook, segment the character image area to be recognized, and perform high-precision recognition of the character image area , Where the character image area to be recognized is the scoring column area, and finally data information such as the scoring result will be obtained, and the data information will be uploaded to the database server for statistics, so as to achieve the vertical (different time period) and horizontal (different individuals) Time) statistical data analysis.
[0036] The database server is used to store the images collected by the image acquisition device, student information, and data processed by the image processing device;
[0037] The intelligent terminal is used to communicate with the database server to display student information, homework information and grading results.
[0038] Such as Figure 5 As shown, the implementation method of this system is as follows:
[0039] 1) After students answer the questions in the answering area, the teacher reviews them first, and then the high-speed scanner scans and collects the images of each page of the workbook.
[0040] 2) The image processing device sequentially performs binarization processing on the scanned images; adopts a method based on the maximum between-class variance to perform image binarization. First, the maximum between-class variance method is used to determine the threshold of the segmented image foreground and background, and then the maximum and minimum gray values of the image are calculated. If the difference between the maximum and minimum gray values exceeds a preset value, the threshold is used for binarization, otherwise the image is binarized to a full background, and the job is marked as a blank job, and no subsequent processing is performed.
[0041] 3) Barcode area positioning: scan the barcode in the image, locate the image direction through the barcode, and perform rotation correction on the tilted image; perform the positioning of the barcode area according to the structure of the barcode and the positioning identification symbols, and calculate the job image based on the positioning information Rotation and tilt direction. Then, using the horizontal and vertical division of the solid line of the score column area, the method based on projection profile analysis is used to accurately locate the tilt direction within a small angle range, and finally the tilt and rotation direction of the image are corrected.
[0042] 4) The location of the student information area; according to the barcode location result and the job structure, the Hough transform line detection method is used to detect areas such as student name, class and student number, and realize the location of the student information area.
[0043] 5) Positioning of the scoring column area; using a method based on line detection to locate and segment the corresponding scoring column image block. First, the Hough transform line detection method is used to detect the vertical segmented solid lines on the left and right sides of the scoring column area to locate the scoring area, and then use the method based on projection contour analysis to locate the horizontal segmentation line of the score column. Finally, according to the position of the horizontal and vertical dividing lines, the image positioning of the scoring area of each topic is realized.
[0044] 6) Segment the image information of the scoring area that has been located, and segment the character blocks to be recognized. Calculate the optimal segmentation path, segment the candidate characters of the character block, and calculate the credibility of the candidate characters; here an over-segmentation strategy is adopted. For each character image block, vertical projection is first performed, and all points whose projection contour value is less than a certain threshold are used as candidate vertical segmentation points to obtain candidate character segmentation blocks. Then, according to the width of the adjacent candidate character segmentation block and the number of stroke crossings scanned in the vertical direction at the vertical segmentation point, the segmentation blocks whose segmentation block width and stroke crossing number are both less than a certain threshold are merged. Finally, the over-segmented candidate segmentation character is obtained. Since handwritten characters are prone to continuous strokes, this method can better adapt to continuous strokes.
[0045] 6) Recognize the segmented characters, and finally get the scoring result and the recognition result of the corresponding test paper information. The text block is divided into non-Chinese character strings (such as student ID, score, etc.) and Chinese character strings (such as name, class), and the recognition processing is performed separately. The recognition process of the text block is as follows: firstly, according to the candidate segmentation characters obtained in the segmentation process of the text block, the candidate character segmentation path diagram is established, and then the optimal segmentation path is calculated by the dynamic programming method to obtain the recognition result of the string. For the recognition of non-Chinese character strings, for each candidate character segmentation block, first calculate the 8-direction gradient feature, and perform LDA dimensionality reduction, and then use the MQDF (ModifiedQuadratic DiscriminantFunction) classifier to classify in the dimensionality reduction feature vector space. Obtain the recognition reliability of each candidate character. The recognition credibility is integrated with geometric information such as the aspect ratio of adjacent candidate character segmentation blocks, and the synthesized recognition credibility is substituted into the optimal segmentation path calculation for text block recognition. For the recognition of Chinese character strings, each candidate character is divided into blocks, and the 8-direction gradient feature is first calculated, and LDA is performed to reduce the dimensionality. In the feature vector space after dimensionality reduction, the MQDF classifier is used to classify, and the recognition credibility of each candidate character is obtained. The recognition credibility and the binary language model information, as well as the width and height of the adjacent candidate character segmentation blocks The geometric information is integrated, and the integrated recognition reliability is substituted into the optimal segmentation path calculation for text line recognition.
