计算机与数字工程2023，Vol.51Issue(10)：2405-2412,8.DOI:10.3969/j.issn.1672-9722.2023.10.036
改进HOG特征的支持向量机工程车辆识别方法
Support Vector Machine Engineering Vehicle Recognition Method Based on Improved HOG Feature
摘要
Abstract
Urban traffic safety is becoming more and more important,and the demand for engineering transportation vehicle monitoring is increasing day by day.In order to improve the recognition accuracy of urban road intelligent monitoring system,an en-gineering vehicle recognition method based on improved gradient direction histogram(HOG)multilevel support vector machine(SVM)is proposed in this paper.Firstly,in order to avoid the mixed Gaussian background model(GMM)error caused by the rapid change between frames,the digital image quality judgment factor is designed to eliminate the video frames with low signal-to-noise ratio.At the same time,the adhered vehicle contour in the foreground is segmented based on the convex hull algorithm to suppress the interference of different viewing angles and light shadows on feature extraction.Secondly,according to the hog feature,the diag-onal gradient detail is fused to enhance the ability of wheel profile description.Pre screening based on block gradient direction vari-ance to eliminate low signal-to-noise score blocks and enhance feature robustness.Local binary pattern(LBP)operators with differ-ent weight radius scales are fused,and texture attributes are added to further enhance feature resolution.Finally,a multi-level SVM classifier is designed according to the different characteristic attributes of vehicles to effectively improve the classification accu-racy of target engineering vehicles.The experimental results show that the recognition accuracy and recall rate of engineering vehi-cles in this paper are 92%and 90%,which can better complete the task of road engineering vehicle recognition.关键词
图像分类/前景提取/特征提取/特征降维/支持向量机Key words
image classification/foreground extraction/feature extraction/feature dimensionality reduction/support vector machine分类
信息技术与安全科学引用本文复制引用
吴逸,吴静静..改进HOG特征的支持向量机工程车辆识别方法[J].计算机与数字工程,2023,51(10):2405-2412,8.基金项目
国家自然科学基金项目(编号:61873246,62072416)资助. （编号:61873246,62072416）
