基于区域检测算法与不变矩特征的飞机特征提取方法研究

摘要

为了克服现有算法处理结果易受影像获取时光照、视角变化的影响,提出一种结合Harris-Laplace特征区域检测算法与梯度仿射不变矩特征描述子的飞机自动识别方法.首先,采用Harris-Laplace算子检测图像中飞机目标特征区域,在此基础上,采用梯度仿射不变矩,更形象地描述目标特征区域内的仿射变换特性,利用特征区域内的梯度仿射不变子作为飞机目标识别的特征向量用于目标识别.实验结果验证了该算法的稳健性及高效性.%High spatial resolution remote sensing image is meaningful in military and civilian situation.In order to overcome the influence of illumination and perspective,this paper introduces Harris-Laplace and Affine Invariant Fast Local Feature,and proposes a method combining these two operations to recognize the aircraft.Harris-Laplace was used to detect the interest region and Affine Invariant Fast Local Feature was used to describe the interest region.It proved that the technical route this paper offered has a good performance in the recognition of aircraft of high spatial resolution.