基于多特征融合的运动物体识别方法研究纪筱鹏.pdfVIP

4 1 3 V o l14 1 N o13 2 0 0 9 3 JOURNA L OF HAR BIN INST ITUTE OF TECHNOLOGY M ar. 2009 1 1 1 2 纪筱鹏, 魏志强 , 冯业伟 , 王 鹏 1. , 266 100, jx iaopeng@ ouc. edu. cn; 2. , 266000) : 为了实现运动目标的实时检测, 本文基于多特征融合, 提出了 一种运动物体分级识别 法. 首先通 过对运动目标的分割, 分析得到各 个物体的形状矩特征面积大小及运动速度. 对上述特征, 采用支持向量机 法对其矩特征进行快速的一级分类, 然后对一级分类结果及各物体的面积速度等信源根据其对分类的重 要程度, 采用模糊积分技术进行多特征融合分类. 实验表明, 本 法具有较强的学习能力及较高的识别精度, 可有效地应用于道路监控等系统. : 多特征融合; 支持向量机; 模糊积分; 矩特征; 目标分类与识别 : TN9 1 : A : 0367 - 6234 2009) 03- 0169- 0 3 M oving object recogn ition based on m u lt-i feature fu sion 1 1 1 2 JI X iao-peng , W E I Zh -i q iang , FENG Y e-w ei , WANG P eng 1. Com puter Science D epartm ent, O cean U n ive rs ity o f Ch ina, Q ingdao 266100, China, jx iaopeng@ ouc. edu. cn; 2. Q ingdao M unic ipa l B ranch Shandong P rov inc ia l Branch Corporation CH INA T ELECOM, Q ingdao 266000, Ch ina) Abs tract: To realize the rea-l tmi e detection of mov ing targets, a h ierarchy of mi age recognition based on m ulti-feature form ov ing objects is proposed in th is paper. F irstly, the shape m om ent features, size and veloc-i ty of each ob ject are obta ined through the segm entation of m oving ob jects. Secondly, suppo rt vector m achines are u sed to app ly the first c lassif ication to above features. T hen fuzzy integral is used to app ly the m u lti-feature fu sion c