首页 | 本学科首页   官方微博 | 高级检索  
     

隐马尔可夫模型下基于SIFT特征的局部遮挡目标识别
引用本文:王惠宇,顾苏杭,吕继东. 隐马尔可夫模型下基于SIFT特征的局部遮挡目标识别[J]. 湖南农业大学学报(自然科学版), 2016, 0(3): 90-93
作者姓名:王惠宇  顾苏杭  吕继东
作者单位:(1.常州轻工职业技术学院, 江苏 常州213164;2.常州大学 信息科学与工程学院, 江苏 常州213164)
摘    要:提出一种利用隐马尔可夫模型建立目标特征匹配库来识别图像中局部遮挡目标的新方法。该方法首先通过SIFT算法提取目标SIFT特征,然后采用隐马尔可夫模型对目标所有的SIFT特征进行训练,得到目标SIFT特征对应的模型输出概率范围,将该概率范围作为目标特征匹配库。在对图像中的目标进行识别时,利用目标特征匹配库可以把目标特征从图像所有特征中识别出来,即使目标遮挡比例为60%时,该方法仍能识别出目标。实验结果表明,新方法可以精准地识别出图像中被遮挡目标,能够很好地解决遮挡情况下的目标识别问题。与现有局部遮挡目标识别算法相比,新方法所取得的目标识别率均有所提高。

关 键 词:隐马尔可夫模型;目标识别;遮挡;SIFT特征;特征匹配

Partially Occluded Object Recognition Based on SIFT Features under Hidden Markov Model
WANG Hui-yu,GU Su-hang,LV Ji-dong. Partially Occluded Object Recognition Based on SIFT Features under Hidden Markov Model[J]. Journal of Hunan Agricultural University, 2016, 0(3): 90-93
Authors:WANG Hui-yu  GU Su-hang  LV Ji-dong
Affiliation:(1.Changzhou Vocational Institute of Light Industry,Changzhou,Jiangsu213164,China;2.Dept.College of Information Science and Engineering,Changzhou University, Changzhou,Jiangsu213164,China)
Abstract:This paper proposed a new method to recognize partially occluded image object via feature matching library by using Hidden Markov Model(HMM). Firstly, the object SIFT features extracted by SIFT algorithm were trained by HMM, and the probability range corresponding to the SIFT features was obtained as the object features matching library. The matching library can identify object SIFT features from all image SIFT features during image object recognition, even when the object occlusion ratio reaches 60%, the new method can still identify the object. The experimental results indicate that the new method can accurately identify partially occluded image object, and it can be a good solution in case of partially occluded object recognition problem. Compared with the existing partial occlusion object recognition algorithm, the new method can improve object recognition rate.
Keywords:HMM   object recognition   occlusion   SIFT feature   feature matching
点击此处可从《湖南农业大学学报(自然科学版)》浏览原始摘要信息
点击此处可从《湖南农业大学学报(自然科学版)》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号