摘要
针对多类物体识别中计算量大、识别率低等问题,在现有模拟视觉系统的计算模型基础上,对原模型进行了改进,提出了改进模型.首先,通过有效的算法提取图像中的兴趣点,并以此为中心选择适当尺度的小块作为特征模板,从而提高模板有效性;然后,确立了以固定兴趣点个数的方法来选择兴趣点,从而解决多类物体识别中兴趣点选取的阈值问题.对多类物体分类识别的实验结果表明:改进后的模型比原有模型具有更快的识别速度和更高的识别率.
In order to solve the problems of large amount of computing and low recognition in multiclass object recognition,we proposed a model,which was improved from an existing a computing model of the analog visual systems.First,the interest points were extracted by effetive algorithm,and conidered as center point,by which proper scale small pieces were choose as the characteristic template,so that it can improve the effectiveness of the template.Then,the coordination in all categories target threshold level issues was solved by selecting several number of points.The results show that the improved model can recognize more exactly and quickly than existing one.
出处
《中南民族大学学报(自然科学版)》
CAS
2011年第2期61-66,共6页
Journal of South-Central University for Nationalities:Natural Science Edition
基金
国家自然科学基金资助项目(60972158)
关键词
多类物体识别
计算模型
兴趣点
multiclass object recognition
computing model
interest points