LTFigure 直接下面的 LTChar + 标题位置错误。

作者: joaquimcampos创建于 2022年8月24日更新于 2026年9月12日
标签type: bugcomponent: converterstatus: needs solution

There are two issues when running pdfminer on the following pdf: issue_100.pdf.

  1. Text extraction puts the first-page header at the end of the page.
  2. No matter what laparams one uses (e.g. increasing a lot word_margin), the header is not recognized as a single text block, nor are the characters connected into words. In fact, the layout analysis shows that the header appears as several individual LTChar instances directly below an LTFigure instance without any TextContainer in between. (This might lead to 1. in some way). I don't think this is supposed to happen. Might be related to #763 and #526. To reproduce, run the following code on the pdf issue_100.pdf. You can experiment with different laparams.

内容来源: pdfminer/pdfminer.six