局部Radon变换技术及在毛杆折痕识别中的应用

Feather quill crease recognition based on local Radon transform

针对羽毛杆折痕识别问题,提出一种新的特征提取方法。为了消除Radon变换对缩放平移敏感问题,采用改进的Radon变换提取目标区域的不变矩,并引入局部投影技术消除羽毛杆生理纹理干扰。通过改变尺度因子获得矩不变量矩阵,并采用奇异值分解(SVD)获得特征不变量用于分类识别。实验结果表明该方法具有较强的鲁棒性,具有较高的折痕识别率。

Concerning the recognition of feather quill crease feature, a new feature extraction method was proposed. In order to solve the scaling and translation sensitivity of Radon transform, an improved Radon transform was used to extract moment invariants of target region and local projection technology was adopted to eliminate interference physiological texture of feather quill. Obtaining invariants matrix by changing the scale factor, Singular Value Decomposition （SVD） was provided here to obtain feature invariant for classification and recognition. The results show that this suggested method has higher robustness and higher recognition rate compared with other algorithms.