车载遥感高速公路广告影像的文本信息提取研究与应用

Research and Application of Text Information Extraction from Vehicle-mounted Remote Sensing Expressway Billboard Data

针对现有高速公路广告牌人工巡检成本高、效率低且危险性大等问题,提出了一种以车载遥感影像的文本信息识别为技术支持的高速公路广告牌巡检新方法。该方法以车载遥感高速公路广告牌影像为研究数据,建立了一种融合最大稳定极值区域(maximally stable extremal regions,MSER)检测的高精度筛选算法和基于边缘像素点的笔划宽度变换(stroke width transform,SWT)字符识别优化算法,并基于文本信息提取实现广告牌问题的自动化甄别。研究结果表明:倡导的通过车载遥感技术与信息提取所结合的高速公路广告牌智能巡检方法可行且效率高,并且很好地解决了现行作业方式中施工人员的安全问题;提出的广告牌文本信息提取方法准确率高于80%。研究成果为我国高速公路管理及相关部门及时有效地识别、更新、维护高速公路广告牌提供了一种新的原创性技术支撑。

Aiming at the problems of high cost,low efficiency and high risk of manual inspection of existing highway billboards,this paper proposes a new method of highway billboard inspection based on the vehicle-mounted remote sensing platform.Based on UAV highway billboard image data as the research object,it proposes a high-precision screening algorithm that combines the detection of maximum stable extremal regions(MSER)and edge pixel based SWT(stroke width transformation)billboard text extraction method based on character recognition optimization algorithm,and automatic identification of billboard issues based on text information extraction.It also solves the safety problems of construction personnel in the current operation mode.The accuracy of the billboard text information extraction method proposed in this paper is higher than 80%.The research results provide a new and original technical support for expressway management and related departments to identify,update and maintain expressway billboards in a timely and effective manner.