重庆理工大学学报（自然科学版）Issue(3)：65-70,6.DOI:10.3969/j.issn.1674-8425(z).2015.03.013
图像多特征融合的障碍物检测
Obstacle Detection Based on Multi-Feature Fusion
摘要
Abstract
Aiming to solving the problem that when the obstacles and ground graphic in outdoor envi-ronment can not be distinguished very obviously,there will be a misjudgment and missing detection phenomenon for the outdoor environment when detecting obstacle based on color or luminance informa-tion,we adopt the method of multiple feature fusion to detect obstacles. LBP algorithm was used in extracting image texture feature. Meanwhile,adaptive canny algorithm was used in extracting obstacle edge information. Texture feature and edge information were merged with linear weighted method,and then updated the background information with Gaussian model method,and at last judged the obstacle by comparing overall differences between two adjacent frames. Compared with traditional three frame method,average background modeling method,the experimental results of Gaussian mixture model method show that Gaussian mixture model achieves a better effect in obstacle detecting when being contrasted to using a single feature.关键词
障碍物检测/LBP 算法/Canny 算法/线性加权/高斯混合模型Key words
obstacle detection/LBP algorithm/Canny algorithm/linear weight/Gaussian mixture model分类
计算机与自动化引用本文复制引用
张建勋,汪波,侯之旭,靳冲..图像多特征融合的障碍物检测[J].重庆理工大学学报（自然科学版）,2015,(3):65-70,6.基金项目
国家自然科学基金资助项目（61173184）；重庆理工大学研究生创新基金资助课题 （）
