专利说明书附图中附图标记识别应用研究

为了能够自动智能获得专利说明书附图中的附图标记,提出了一种专利附图标记识别算法.根据专利说明书附图中附图标记的特征,集合预处理、灰度转换、二值化、边缘检测等图像识别算法,开发了灰度阈值、矩形框边容错值、矩形框高容错值、标示、提取融合、降噪、识别等新算法,准确完整地识别出专利附图中附图标记.与普通图像数字识别算法相比,附...

Reference Signs In Patent Description Drawings Detection

In order to automatically and intelligently obtain the reference signs in the patent specification drawings, a patent reference sign recognition algorithm is proposed. According to the characteristics of the reference signs in the patent specification drawings, use image recognition algorithms such as preprocessing, gray conversion, binarization, edge detection, etc., and develope new algorithms such as gray threshold, rectangular frame edge tolerance value, rectangular frame high tolerance value,mark, extraction fusion, noise reduction, recognition, accurately and completely identify the reference signs in the patent drawings. Compared with the common image digital recognition algorithm, the accuracy of the reference signs recognition algorithm is improved by more than 30%, and the noise interference is very little. The common form defects of drawings in patent specifications are analyzed. The recognition algorithm provides theoretical basis and technical support for patent drawing examination and patent search intelligent systems. At the same time, the algorithm can mark the position of reference signs, which is convenient for examiners and pre-examiners to check and identify condition.