System and method for automatic extraction and naming of visual elements
Patent Information
- Authority / Receiving Office
- TW · TW
- Patent Type
- Patents
- Current Assignee / Owner
- DELTA ELECTRONICS INC(CN)
- Filing Date
- 2025-08-14
- Publication Date
- 2026-08-01
Smart Images

Figure TWG2TB001904092_001 
Figure TWG2TB001904092_002 
Figure TWG2TB001904092_003
Abstract
Claims
1. A system for automatically capturing and naming charts, comprising: One processing unit; The system also includes a storage unit storing a computer program containing multiple instructions that, when loaded into the processing unit, cause the processing unit to perform the following steps: acquiring a portable file format (PDF file); converting each page of the PDF file into a page image file; identifying the chart area, chart title area, and chapter title area in each page image file using object detection; for each page image file, extracting the chart title text in the chart title area or the chapter title text in the chapter title area using optical character recognition (OCR); for each page image file, assigning a region type tag to each chart area, chart title area, and chapter title area, and establishing a page structure sequence according to the arrangement order of the chart areas, chart title areas, and chapter title areas in the page image file; analyzing each page structure sequence to match the chart areas and chart title areas; and storing each chart area as a separate image file and naming the image file with the chart title text corresponding to the chart title area matched by that chart area.
2. The system as described in claim 1, wherein the processing unit further performs the following steps to analyze each page structure sequence: dividing each page structure sequence according to a region type marker representing a chapter title area to form a plurality of chapter ranges; within each chapter range, sequentially searching for two adjacent region type markers, one of which represents the chart area and the other represents the chart title area; and determining that the chart area and the chart title area corresponding to the two adjacent region type markers found are a match between each other.
3. The system as described in claim 2, wherein the processing unit further performs the following steps to analyze each page structure sequence: For each page image, determining whether the last section range in the corresponding page structure sequence ends with the region type marker representing the chart area, and whether the first section range of the page structure sequence corresponding to the next page image begins with the region type marker representing the chart area; In response to determining that the last section range ends with the region type marker representing the chart area and the first section range begins with the region type marker representing the chart area, calculating the similarity between the last chart area of the page image and the first chart area of the next page image; and In response that the similarity is higher than a preset threshold, merging the last chart area of the page image and the first chart area of the next page image into the same chart.
4. The system as described in claim 3, wherein the processing unit further performs the following steps to analyze each page structure sequence: In response to the similarity not being higher than the preset threshold, inserting a region type marker representing the chapter title region before the page structure sequence corresponding to the next page image.
5. The system as described in claim 4, wherein the processing unit further performs the following steps to analyze each page structure sequence: in response to determining that the last chart area of the page image has not yet matched any chart title area, and that the page structure sequence corresponding to the next page image begins with the area type marker representing the chart title area, determining that the last chart area of the page image and the first chart title area of the next page image are matched.
6. A method for automatically capturing and naming charts, implemented by a computer system, the method comprising: Obtain a portable file format archive (PDF file); convert each page of the PDF file into a single page view image file; Through object detection, the system identifies the chart area, chart title area, and chapter title area in each page image file. For each page image file, optical character recognition (OCR) is used to extract the chart title text in the chart title area or the chapter title text in the chapter title area. For each page image file, a region type tag is assigned to each of the chart area, chart title area, and chapter title area in the page image file, and a page structure sequence is established according to the arrangement order of the chart area, chart title area, and chapter title area in the page image file. Each page structure sequence is analyzed to match the chart area and chart title area. Each chart area is stored as an image file, and the image file is named with the chart title text corresponding to the chart title area that the chart area matches.
7. The method as described in claim 6, wherein the step of analyzing each page structure sequence to match the chart area and the chart title area further includes: Each page structure sequence is divided according to the area type marker representing the chapter title area to form multiple chapter ranges; Within each chapter, two adjacent area type markers are searched sequentially, one of which represents the chart area and the other represents the chart title area; and the chart area and the chart title area corresponding to the two adjacent area type markers are determined to be a match between them.
8. The method as described in claim 7, wherein the step of analyzing each page structure sequence to match the chart area and the chart title area further includes: For each page image, determine whether the last section of the corresponding page structure sequence ends with the region type marker representing the chart area, and whether the first section of the page structure sequence corresponding to the next page image begins with the region type marker representing the chart area; in response to determining that the last section ends with the region type marker representing the chart area and the first section begins with the region type marker representing the chart area, calculate the similarity between the last chart area of the page image and the first chart area of the next page image; and in response that the similarity is higher than a preset threshold, merge the last chart area of the page image and the first chart area of the next page image into the same chart.
9. The method as described in claim 8, wherein the step of analyzing each page structure sequence to match the chart area and the chart title area further includes: In response to the similarity not being higher than the preset threshold, a region type marker representing the chapter title region is inserted before the page structure sequence corresponding to the next page image.
10. The method as described in claim 9, wherein the step of analyzing each page structure sequence to match the chart area and the chart title area further includes: In response to the determination that the last chart area of the page image file has not yet matched any chart title area, and that the page structure sequence corresponding to the next page image file begins with the area type marker representing the chart title area, the last chart area of the page image file and the first chart title area of the next page image file are determined to match each other.