项目名称: 移动云计算模式下的场景文本感知方法研究
项目编号: No.61203259
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 自动化学科
项目作者: 朱远平
作者单位: 天津师范大学
项目金额: 24万元
中文摘要: 智能移动终端的广泛应用催生了人们利用移动终端感知周围场景视觉信息的迫切需求,赋予移动终端文本感知能力是其中重要组成部分,其核心就在于自然场景文字识别技术。针对场景文字中普遍存在退化严重的问题,传统方法不能胜任。本课题拟研究卷积神经网络在灰度字符空间上进行自然场景退化文字的识别,试图避免传统方法基于二值字符识别所带来的二次降质问题,进而提高场景文字识别鲁棒性。同时,为克服移动终端计算能力不足的问题,还将研究适合于移动云计算模式的场景文字识别系统。利用新计算模式的特点,研究新的文本检测和识别方法与流程,降低移动终端计算负荷,并通过面向场景文字识别的数据压缩方法减少数据传输量,最终提高移动终端识别文字的实用性。
中文关键词: 文本感知;场景文字识别;文本检测;移动云计算;
英文摘要: Smart mobile client pushes the demand on text sensing power whose core technique is natural scene text recognition. However, traditional OCR technique is unable to work well on serious degraded characters frequently found in natural scenes. This project will try to recognize degraded characters on grayscale space using convolutional neural network, which try to avoid the second degradation from binarization in traditional character recognition method. Then, recognition robustness will be expected. Meanwhile, to overcome the problem of weak computation power in mobile device, it plans to design character recognition based on mobile cloud computing. Making full use of new computation mode's advantages, new text detection and recognition method are designed to decrease computational load on mobile device. Furthermore, scene text recognition oriented data compression method will be developed to reduce data transmission. By this way, the practicability of mobile character recognition will be enhanced.
英文关键词: Text Sensing;Scene Text Recognition;Text Detection;Mobile Cloud Computing;