基于结构聚集度的海面目标特征描述及匹配方法

Feature Description and Matching Method of On-sea Target Based on Structure Aggregation

在对海面目标进行光电图像跟踪时,由于跟踪技术的局限性,跟瞄点往往会出现对目标的"失跟"现象,为了判断这一情况,通常需要对前后帧跟瞄区域的目标图像特征进行对比判别。但红外图像基本不存在颜色及纹理特征,而仅使用形状特征对于持续运动中的复杂目标图像又难以取得良好的效果。为解决这一问题,提出了一种基于目标结构聚集度的特征描述及匹配方法。首先对目标图像进行均匀矩形划分,在此基础上建立基于矩形块填充比的邻域矩阵作为图像特征,计算相邻帧图像该特征的欧氏距离差,并加入面积差参数进行特征匹配。对多种不同情况下的实测及模拟目标进行了计算机仿真实验,结果表明这种特征能够有效判断"失跟"情况,并且对于光学及红外目标旋转具有较好的不变性。

In the optical and infrared image tracking of the On-sea target,for the limitations of tracking technology,the aiming point will always lose target.In order to judge the situation,usually the target image characteristics of front and back-frame should be compared and discriminated.However,the color and texture features is not existed in the infrared image,and only using the shape features to the complex moving target image is difficult to achieve good results.To solve the problem,a feature description and matching method based on target structure aggregation is presented.Firstly,the target image is divided into uniform rectangular.Then the neighborhood matrix based on rectangular block filling ratio is employed for comparison feature.The Euclidean distance is used for feature matching which is added by area difference parameters.The computer simulations are made for measured and tested targets under a variety of different cases.The results show that such features can effectively determine the loss track situation,and have good rotation invariance for optical and infrared target.