改良HOG结合SVM的快速障碍识别  

Fast Obstacle Recognition Combining the Improved HOG with SVM

作　　者：赵柏山[1] 荣子航 ZHAO Baishan;RONG Zihang(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学信息科学与工程学院,沈阳110870

出　　处：《微处理机》2023年第5期54-56,共3页Microprocessors

摘　　要：为提升车辆驾驶的安全性,通过主成分分析降维来改良方向梯度直方图,并结合支持向量机分类,提出一种快速障碍识别方法。该方法在普通HOG特征仅提取横竖两方向梯度分量的基础上,补充了主、副对角线两个方向上像素的灰度信息,使HOG特征具有更丰富的边缘信息,同时针对HOG特征维数高而造成识别速度偏慢的问题,使用PCA来降维,以最大限度保证降维后的识别成功率以及识别速率提升。实验结果表明,本方法相较于改良前的HOG特征结合SVM的方法更好地提升了识别成功率并且有效地降低了识别时间。Aiming at improving the safety of vehicle driving,the direction gradient histogram is improved by dimensionality reduction of principal component analysis,and a fast obstacle identification method is proposed by combining the improved HOG with support vector machine classification.Different from extracting only the gradient components in horizontal and vertical directions in ordinary HOG features,the method adds the gray information of pixels in the main-diagonal and sub-diagonal directions to the HOG,which makes the HOG features have more edge information,while aiming at the problem of slow recognition speed caused by the high feature dimension of HOG,PCA is used to reduce the dimension,so as to ensure the recognition success rate and im-prove the recognition speed to the greatest extent.The experimental results show that the method can improve the recognition success rate and effectively reduce the recognition time compared with the method using ordinary HOG features and SVM.

关 键 词：障碍识别 支持向量机 方向梯度直方图 主成分分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]