舰船目标图像不变特征及识别仿真研究  

Research on Invariant Features of Ship Target Image and Its Recognition

作　　者：季超[1] 杨晓东[1] 

机构地区：[1]海军潜艇学院,山东青岛266042

出　　处：《计算机仿真》2016年第3期5-8,89,共5页Computer Simulation

基　　金：海军装备部预研项目(4010103010301-05)

摘　　要：在舰船目标图像优化识别中,传统不变矩高阶矩易受噪声干扰,存在信息冗余,且特征量匹配也仅局限于计算图像与模板特征量间的距离度量或相关度量。为解决上述问题,提出了一种采用无冗余复数不变矩和多项式曲线拟合的舰船目标图像识别算法,选取复数不变矩作为描述图像识别的不变特征,以图像的复数不变矩作为数据点,结合多项式曲线拟合理论,计算图像复数不变矩的多项式曲线拟合系数,通过拟合系数间的相似性度量对图像目标进行描述,并应用于实际的舰船图像并对其有效性进行了仿真分析。仿真结果表明,所提方法对舰船目标图像不变特征描述具有较好稳定性,为舰船目标图像识别提供了参考。An image recognition method for ship target is proposed, which combines non-redundant complex mo- ments with polynomial curve fitting. Complex invariants are selected as the features for image recognition. Combined with the theory of polynomial curve fitting, the fitting coefficients of the complex invariants are calculated, and the similarity measure between the fitting coefficients is used to describe the image target, finally it is applied in real ship images. The experimental results show that the proposed method which is used to describe the invariant features of the ship target image has good stability and effective meaning in image target recognition.

关 键 词：舰船目标图像识别 复数矩 多项式曲线拟合 不变特征 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]