【摘要】 字符识别目前广泛用于文献资料的检索等场合。目前大多数字符识别使用摄像头采集图像,预处理后,进行字符校正,字符分割,字符识别。传统算法主要存在运行速度慢,耗时长,内存占用多等问题。同时,用摄像头采集图像,其本身也有一定的缺陷,例如对光线敏感等。因此,文中提出使用一种改进过的FPA字符细化算法结合SWT文本检测算法对文本进行预处理,使用HOG特征向量提取以及SVM分类等算法进行字符识别。同时,给出一种新型的基于CIS(接触式图像传感器)图像采集和处理系统,CIS自带光源,具有一些摄像头所不及的优点。文中方案经过测试,识别率较高,运行速率较快,内存占用率和处理时间较传统的字符识别方式都有一定的改善和提高。更多还原

【Abstract】 Character recognition is widely used in information retrieval and so on. At present,traditional character recognition algorithms use cameras for image acquisition. After proprocessing,it conducts character correction,character segmentation and character recognition. The main existing problems of the traditional algorithms are slow running speed,long time consuming,big memory occupancy and so on.Also,using camera to capture images shas a certain defect itself,such as light sensitive,etc. Therefore,this paper proposes an improved FPA character thinning algorithm combined with SWT text detection algorithm to preprocess the image,using HOG feature vector extraction and SVM classification algorithm for character recognition. At the same time,a new type of image acquisition and processing system based on CIS( contact image sensor) is presented,which has the advantages of CIS light source over cameras.Being tested,the recognition rate is higher,while the running rate and the memory occupancy rate are improved with the processing time less comparing to the traditional character recognition methods.更多还原