【摘要】 提出了基于模糊测度的目标匹配方法。利用复数小波变换图像,对得到的不同尺度、不同方向子频带的小波系数进行归一化,消除了光照和尺度变化对图像的影响,构造其纹理特征。采用循环移位方法构成匹配特征向量,根据特征分布的模糊隶属度确定各特征分量在相似测度中的权值。复数小波的多尺度、多分辨率等特性有效描述图像能量在不同尺度下、不同方向的分布特征。实验结果表明:当目标在背景中发生相对变化时,应用该算法能够有效匹配目标,其性能优于基于灰度相关匹配的方法。更多还原

【Abstract】 An algorithm of target matching based on fuzzy metric is proposed. The wavelet coefficients with different scales and orientations are normalized by means of complex wavelet transform, which can eliminate effect of illumination and scale change. Texture feature could be constructed from these normalized wavelet coefficients through cycle shift. The weights of different features are determined according to a principle of fuzzy attribute degree. The characteristics of complex wavelet that have properties of multi-scales and multi-resolutions effectively describe the characteristics of energy distribution. The experimental results show that the algorithm can effectively match target while the target changes relatively under background and its performance also supers to the approach based on gray-level region correlation.更多还原