基于生物视觉显著性的车辆车型识别  

Vehicle Type Recognition Based on Biological Vision Salience

作　　者：陈振学[1] 刘成云[1] 常发亮[1] 

机构地区：[1]山东大学控制科学与工程学院,济南250061

出　　处：《计算机科学》2010年第2期207-208,224,共3页Computer Science

基　　金：国家自然科学基金(60975025);教育部博士点专项基金(20090131120039);中国博士后科学基金特别资助(200902563);中国博士后科学基金面上资助(20080441123);山东省博士后创新项目专项资金(200802017)资助

摘　　要：车型的识别问题是典型的目标识别问题,根据生物视觉与模式识别理论,对车辆车型的检测与识别进行了研究,提出了基于最小错误概率的特征显著性车型识别算法。该算法对车型的多个特征进行显著性比较,对较显著的特征赋予较大的权值,然后对多特征的识别结果进行融合处理。实验结果表明该算法对车型的识别具有较高的识别率。Vehicle type recognition is typical target recognition. According to biological vision and pattern recognition theory,the detection and recognition of vehicle type were researched. So, a novel method based on feature salience was presented to recognize vehicle type. Firstly, this algorithm extracts the multi-features of vehicle type and compares their salience. Then, the more salient feature was put larger power. Finally, the salient features were fused to recognize the vehicle type. The experimental results show that the algorithm has better identify rate.

关 键 词：特征选择 生物视觉显著性 最小错误概率 车型识别 特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]