基于分段轮廓平滑的目标识别

Object Recognition Based on Smoothing Segmented-contour

提出了一种基于分段轮廓平滑的目标识别算法.首先通过曲率将轮廓划分为特征区域和非特征区域;然后在不同区域内分别采用不同方差的高斯函数进行轮廓平滑;最后采用基于仿射不变矩的目标识别算法对平滑后的目标轮廓进行识别.结果表明,该算法不仅取得了更好的轮廓平滑效果,而且在强噪音条件下能够显著提高识别准确率.

An object recognition algorithm based on segmented-smoothing contour is proposed.The contour is divided into two kinds of zones by curvature：feature zones and non-feature zones.Then,the different kinds of zones are smoothed with Gaussian functions with different variance.Finally,the object recognition algorithm based on affine moment-invariant is applied to recognize the smoothed contours.The simulation result shows the better contour-smoothing results and the significant improvement to recognition rate.