【摘要】 为更有效地探测人体目标,提出了利用红外光进行人体成像及步态识别的新方法。首先将人体散发的红外光转换为电信号,经计算机处理后转换为红外步态图像,然后分割出人体目标并将其规格化叠加处理获取步态特征图,再提取包含整体模型信息的边界矩参数以及包含简化模型信息的骨架特征参数,以此作为步态特征识别参量输入至支持向量机(SVM)进行分类识别。使用自建的红外步态数据库(IRGD)进行实验,正确识别率(PCR)为71～92%,且识别效果受人体携带外物的影响不显著。更多还原

【Abstract】 In order to detect the human body very well,a new method based on infrared(IR) thermal imaging was applied for automatic recognition by gait.The infrared ray from human body was transformed to electric signal.The computer processed the electric signal and produced the gait image based on infrared thermal imaging.Then body silhouette sequences were extracted and added together to obtain the gait feature image(GFI).After this,the parameters involving integral model were extracted by using the theory of invariant moments.The parameters involving simplified model were extracted by using the skeleton theory.These parameters were fitted together for gait recognition.The technique of support vector machines(SVM) was presented for classification.This method was applied to infrared thermal gait database(ITGD) and achieved 71%～92% for the probability of correct recognition(PCR).As such,the PCR was insensitive for loading objects that were taken by person.更多还原