图像目标检测与识别仿真平台设计与实现  

The Design and Realization of Image-based Target Detection and Recognition Simulation Platform

作　　者：王阿敏[1] 高颖[1] 王凤华[1] 郭淑霞[2] 

机构地区：[1]西北工业大学航海学院,西安710072 [2]西北工业大学无人机特种技术重点实验室,西安710065

出　　处：《微处理机》2013年第4期41-45,共5页Microprocessors

基　　金：航天科技创新基金(CASC201102)

摘　　要：研究基于图像的目标识别技术问题。针对精确制导武器系统中传统识别方法的识别率低,不能有效解决配准比对图像局部特征的问题,提出基于图像特征点的提取和识别方法。其算法有两种:一种是传统的由下而上的数据驱动型,即不管目标属于何种类型,一律先对原图像进行一般性分割、标记和特征提取等低层次处理,然后将每个带标记的已分割区域的特征矢量与目标模型相匹配;另一种是由上而下的知识(假设)驱动型。即先对图像中可能存在的特征提取假设,根据假设进行有目的地分割、标记和特征提取,在此基础上与目标模型进行精确匹配。这里采用第二种方法并进行了仿真实验,运行结果表明,基于图像特征点的提取和识别方法可使基于图像的检测和识别达到稳定和可靠,证实了方法的有效性。Theme is on image-based object recognition techniques.For the low identification of the traditional methods in the precision-guided weapon system,which can not solve the problem of the image＇s alignment and local features,the method based on image feature extraction and recognition is proposed.There are two basic target recognition algorithms.One is processed from bottom to top,which is called data-driving method.It begins with such low layer processing as general segmentation,label and feature extraction and judges whether the feature vector extracted from the labeled area is in accordance with the feature vector of the object model.The other is processed from top to bottom,which is called knowledge-driving method.It firstly brings forward a hypothesis on probably existed feature,secondly proceeds with purposeful segmentation,label and feature extraction,then,judges whether the feature vector extracted from the labeled area is in accordance with the feature vector of the object model.This article takes advantage of the second algorithm above and simulation experiments.The results show that the proposed method based on image feature extraction and recognition can make it stable and reliable,and the effectiveness of the method is proved.

关 键 词：边缘检测 图像分割 特征提取 图像匹配 目标识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]