【摘要】 评估算法的性能需要反复大量的测试。人工评估既难以保证评价标准的一致,又无法保证测试的速度和规模,只有自动的性能评估才能解决上述问题。表格结构识别是进一步理解表格内容的前提。在定义通用表格结构描述的基础上,提出了基于表格线的表格结构识别的性能评估方法。自动评估系统能够指出识别结果与基准数据的区别,平均每秒比较38.6个表格,远超过人工评估的效率。更多还原

【Abstract】 It need lots of and repeated tests to evaluate the performance of an algorithm.If performance evaluation per-formed by test engineer,the consistency of evaluation standards can’t be guaranteed,so do the speed and the scale of the evaluation.These problems could be solved by automatic performance evaluation.Table structure recognition is precon-dition to understand table contents.A general table structure description is introduced,and a method of automatic perfor-mance evaluation of table structure recognition based on the lines of table is proposed.The evaluation system can locate discrepancies by comparing recognize results and standards data.The system can process38.6tables per second aver-agely,which is far more than test engineer could.更多还原