摘要
提出了基于三维视图的模特姿态服饰识别算法技术。首先基于一元高次多项式最小函数二乘拟合方法 ,由单视图通过变形生成三维视图的模特姿态服饰图像 ,然后基于该单视图和生成的三维视图进行模特姿态服饰图像识别。最后根据整个训练数据样本库 ,通过基于三维视图和融合决策的识别算法进行服饰识别 ,实验结果表明本算法比仅仅基于单视图作为训练样本库的模特姿态服饰识别算法的正确率有了很大提高。
The algorithm technique of dress recognition of model posture based on the three dimensions view is put out in this paper. First of all based on the minimal function two multiplying method of polynomial in one element, and dress picture of model posture in three dimensions view formed by single view, then based on view recognition, finally according to sample library of whole training data, the algorithm of recognition is put to use by combining three dimensions view with blending decision. Testing indicates that the ratio of accuracy by using this algorithm is higher than that only based on one dimension view.
出处
《东华大学学报(自然科学版)》
CAS
CSCD
北大核心
2003年第5期42-46,共5页
Journal of Donghua University(Natural Science)
基金
全国教育科学十五规划重点课题 (No:AYA0 1 0 0 34)
国防科技预研项目"评估多媒体数据存取与交换方法的研究"(No:1 5 .8.4)基金资助