An OpenCV-based method for detecting label information of building drawings

• Main Authors: ZHU QUANYIN, FENG WANLI, YAN YUNYANG, PAN SHUXIN, XU KANG, GAO SHANGBING, LIU TAO
• Format: Patent
• Language: Chinese; English
• Published: 14.12.2018
• Subjects: CALCULATING; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses an OpenCV-based method for detecting label information of building drawings. The invention extracts label information in building drawings effectively by optimizing and combining partial algorithms provided by OpenCV and OCR character recognition. Firstly, the pdf format is converted to png format supported by OpenCV, then grayscale, binarization and Canny edge detection are performed, and then the connected region is formed by dilatation, and the label area is determined according to the contour feature. Finally, the label information is recognized by OCR. The method of the invention effectively realizes the information detection of the label area in the building drawings, realizes the pre-examination work of the building drawings, and improves the examination efficiency of the building drawings. 本发明公开了种基于OpenCV的建筑图纸标签信息检测方法，本发明通过对OpenCV提供的部分算法的进行优化与组合，结合OCR文字识别，有效的提取了建筑图纸中的标签信息。首先对pdf格式的建筑图纸文件转换为OpenCV支持的png格式，接着进行灰度化、二值化和Canny边缘检测，然后通过膨胀，形成图像上的连通域，并根据轮廓特征确定标签区域，最后使用OCR进行标签信息识别。本发明方法