摘要
目的为解决电气符号的大小、图纸背景的模糊、电气符号的旋转角度等各种干扰因素对计算机识别电气图纸造成的误差问题.方法笔者提出了一种基于HOG的电气符号识别方法.建立电气符号训练集,提取电气符号图像的HOG特征,计算出梯度方向向量个数加权图;使用这些HOG特征和分类信息对支持向量机进行训练;利用支持向量机进行识别.结果HOG算法对电气符号的识别率达到92.5%,与SIFT算法比较,识别效果更为准确.结论所提出的HOG算法克服了外界干扰因素对电气符号识别的影响,提高了识别的准确率,具有良好的检测效果,为将HOG算法应用到其他领域奠定理论基础.
Electrical symbols recognition is one of the hot spots in the field of computer vision and electrical design recently. Various factors will bring errors on computer recognition of electrical drawings, such as size of the electrical symbols, drawings of background fuzzy, rotation of the electrical symbols etc.. There are many electrical symbols recognition methods, but the problems are solved hardly. In order to overcome the external interference factors and improve recognition accuracy, an electrical symbol recognition method based on HOG is presented in the paper. Firstly, the electrical symbol training sets were established, and symbols of HOG feature were extracted. Secondly, support vector machine was trained by mean of these characteristics of HOG and classification information. Finally, symbols were recognized by support vector machines. Con- ducted different experiment of complex background research of electrical symbols and experimental study on the algorithm of the algorithm of SIFT and HOG, the results of experiments show that the proposed algorithm of HOG has good detection after experiment.
出处
《沈阳建筑大学学报（自然科学版）》
CAS
北大核心
2013年第3期571-576,共6页
Journal of Shenyang Jianzhu University：Natural Science
基金
国家自然科学基金项目(61272253)
住房和城乡建设部科技项目(2010-K9-22)
关键词
电气符号识别
HOG特征
支持向量机
梯度直方图
electrical symbol recognition
characteristics of HOG
support vector machine
histogram of gra- dient
