摘要
为了解决车型识别过程中车脸图像特征提取的问题,提出了一种标准化不变矩算法。在对车脸图像进行分割处理、提取出感兴趣区域、采用轮廓跟踪法得到形状轮廓的基础上,对原有的HU不变矩矩算法进行了标准化处理,并与典型的HU不变矩算法结果进行了对比。实验结果表明,该不变矩标准化算法在对图像进行特征提取方面可以得到更高精度的特征值。
A novel standardized moment invariants is presented in order to extract image feature of vehicle face during the process of vehicle model classification and identification. Based on the process of auto-face segmenting,interested region extracting,and the shape contour obtaining by contour tracing technique,the traditional HU moment invariants are normalized,which is in comparison with typical HU moment invariants. Experimental results demonstrate that much higher accuracy in terms of feature vector can be achieved by the introduced model.
出处
《沈阳理工大学学报》
CAS
2016年第6期72-76,共5页
Journal of Shenyang Ligong University
关键词
HU不变矩
标准化
车脸图像
特征提取
HU moment invariants
standardized
vehicle face
extract the image feature
