基于独立分量分析的形状识别

Shape Recognition Based on Independent Component Analysis

物体的形状识别是模式识别的重要方向之一,广泛应用于图像分析、机器视觉和目标识别等领域。在介绍利用信号的高阶统计信息的独立分量分析方法基础上,提出了基于独立分量分析的形状识别方法。利用独立分量分析算法提取出图像的独立基,根据待识别图像在独立基上投影系数的差别进行分类识别。仿真实验结果表明,该方法对于形状识别有较高的识别率。

Shape recognition of object is one of the important directions in pattern recognition, and is widely used in image analysis, machine vision and target discrimination, etc. The fundamental of independent component analysis （ICA） is briefly introduced which use the higher order statistics information from source signal. A new shape recognition approach based on ICA is proposed. The 1CA algorithm is used to derive the independent primitive sources out of the given series of sample images. The shape can be recognized by the coefficients of the image projecting to the independent sources. Computer simulation results show the method has high recognition rate.