【作者】 杨智勇；

【导师】 周琪云； 周定康；

【作者基本信息】 江西师范大学 ， 计算机应用技术， 2004， 硕士

【摘要】 OCR(Optical Character Recognition，光学文本识别)技术作为基本的模式识别技术，在计算机输入系统、智能交通系统和安防系统等领域都获得了广泛的应用。根据应用领域的不同，可分为通用的OCR技术及复杂背景下的OCR技术两大类，前者主要应用于将文字材料自动识别录入到计算机系统中；后者则应用于复杂的工业环境中，如智能交通系统中的汽车牌照识别、集装箱编号识别、火车车皮编码识别等领域。复杂背景下的OCR技术涉及的图象处理与模式识别技术较通用的OCR技术更为复杂，是文本识别技术的研究前沿之一。 本文在结合LPR(License Plate Recognition，车牌识别)技术进行实验的基础上，对复杂背景下的文本提取技术进行了研究，提出了基于PCNN(Pulse-CoupledNeural Networks，脉冲耦合神经网络)的边缘检测新方法；同时提出了基于形态学运算的断裂噪声过滤技术，有效实现文本字符特征的提取。 PCNN由Eckhorn根据猫大脑皮层中的视觉神经元同步脉冲(Synchronous Burst)现象所提出。由于具有一系列良好的特性，PCNN在图象处理、模式识别等领域获得了广泛的应用。本文提出的基于PCNN的边缘检测方法可直接对复杂环境下所采集的灰度图象进行边缘提取，并在此基础上实现对目标文本定位。 而在对包含目标文本的图象进行灰度拉伸和阈值分割等预处理以获得有有效的字符特征时，经常会造成目标文本的笔画发生断裂，采用常规的形态学开运算与闭运算对图象进行处理则会产生更严重的笔画断裂甚至缺失。本文利用形态学闭运算的变形对含目标文本的图象进行增强处理，可有效消除文本的断裂噪声。 实验结果表明，采用本文提出的技术方法可以有效检测复杂背景下的文本边缘和消除断裂噪声，增强目标文本的字符特征。更多还原

【Abstract】 As a basic Patten Recognition technology, OCR (Optical Character Recognition) technology is widely used in areas such as Computer Input System, Intelligent Transportation System, Security System etc. According to different applications, OCR technology can be divided into General OCR technology and Complicated-Background OCR technology, while the former is mainly used to automatically recognize and input text material into computer system, the latter is mainly used in complicated industry environment such as License Plate Recognition, Container Serial Number Recognition and Train Wagon Serial Number Recognition in Intelligent Transportation System. Complicated-Background OCR technology uses more complicated image processing techniques and pattern recognition technologies than General OCR technology, and it is also one of the front-line sciences in character recognition field.By experiments on LPR(License Plate Recognition) technology, this paper do some researches on Text Extraction Technology in Complicated-Background OCR technology, and proposes a new edge detect method based on PCNN (Pulse-Coupled Neural Networks), and a break noise filtration technique based on Mathematical Morphologic Process.The PCNN was presented to explain the synchronous burst of the neurons in the cat visual cortex by Eckhorn. For its good properties, it has been widely used in image processing and pattern recognition. This paper proposes a Gray Image Edge Detect Method based on PCNN. This method can be used to detect target characters by fetch edges directly from the gray image that was seized in complicated background.When we do a series of preprocessings to the original image, such as Gray Stretch, Threshold Transform etc, in order to extract their characteristics, they often make the strokes of the target characters broken. But normal Morphologic Opening or Closing process will break the strokes much more seriously. By reforming the Morphologic Closing process, this paper proposes a process to eliminate the breaks of target characters.Experiment result shows that the techniques proposed by this paper can detect the edge of target text and eliminate break noise efficiently, and enhance the characteristics of target text.更多还原