摘要
预处理是目标识别过程中重要环节,主要作用可以使目标识别过程中减小计算量并提高识别精度.提出的基于矩阵秩的特性适合于预处理方法.这种基于矩阵秩的特性对目标的平移变换、目标的90°倍数的旋转变换和目标的镜像变换都保持不变,与常见的其他简单特性比较具有明显的优点,这种方法经过实验验证了其有效性,优于其他简单方法,提高了目标识别的效率.
Preprocessing is always an important part of object recognition in digital images, and its role is to deduce data rapidly and improve accuracy for object recognition. A new method is proposed which is based on matrix rank’s feature and is suitable to preprocessing. This feature has scale invariance, multiple of 90~0 invariance, and mirror translation invariance, and it has obvious excellence compared with the other object features. The experiments performed for object recognition have shown that the superiority of the method is better than other simple methods. The method can been used to further improve the performance of the object recognition.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2005年第6期773-775,853,共4页
Journal of Harbin Institute of Technology
基金
国家自然科学基金资助项目(60272073).
关键词
秩
不变特性
目标识别
特征提取
二值图像
rank
invariant feature
object recognition
feature extraction
binary image
