Abstract: An algorithm for extracting scores and personal information from optical mark reader (OMR) cards has been developed, which takes input images from a scanning device, and locates marks based on horizontal and vertical projections. Input images are rotated before processing to ensure satisfactory alignment with the correct orientation. This is achieved by maximizing standard deviation of the vertical projection. The obtained results are then sent to a database for further manipulation. Experiments show that the algorithm is reliable and efficient.

Key words: Image projection, Standard deviation, Object positioning, Image orientation alignment

摘要： 提出了一种基于图像投影的记分卡识别算法，与扫描装置配合可实现高效准确的卡片信息解读，不仅能达到无差错识别，而且容易实现。系统以扫描仪获取的图像为输入，通过横向和纵向投影实现定位和信息提取。通过搜索纵向投影最大标准差所对应的方向实现角度校正，克服了扫描误差引起图像偏斜的问题。

关键词: 图像投影, 标准差, 定位, 图像角度校正

CLC Number: 

• TP391.41