摘要： 提出了一种新的方法,使用背景-前景分离模型抽取手绘信息,使用启发式规则擦除图像噪音,使用容错机制克服手绘信息随意性造成的识别误差。这种方法不仅保证了识别精度,同时能够精确定位手绘信息填充错误。该方法应用于司机日志图表自动录入与识别系统,通过4组共805页样本集评测,识别精度达到93%,完全超过了实际应用的性能要求。

关键词: 图表识别, 背景-前景分离模型, 启发式规则, 容错机制

Abstract: This paper presents a novel method for driver schedule chart recognition. Background-foreground separation model is used to extract handwritten information, illuminative rules are designed to remove noises in image, and error-toleration mechanism is build to resolve recognition confliction and locate filling error. This method is tested through four test bases containing total 805 pages of real driver log. The accuracy rate of 93% is attained.

中图分类号: 

• TP391.41