摘要
对可见光自然图像人造目标检测技术研究进行综述,在人造目标检测技术的问题描述基础上,基于人造目标与自然背景存在的特性差异,将现有技术按照基于几何特征、分形特征、概率模型、水平集、聚类等类别进行了讨论,并借鉴ATR(自动目标识别)技术评价标准分析了各个算法的优缺点。最后探讨了人造目标检测技术存在的问题和进一步的研究方向。
This paper summarized several man-made objects detection methods.Based on features extraction from man-made object and natural background,followed by presentation of issues on man-made object detection technique,discussed the current algorithms,including based on geometry feature,fractal model,probabilistic model,level set,clustering.Analyzed the advantages and defects of each method detailedly based on ATR（automatic targets recognition）detection performance.Finally,pointed out the present issues and possible further research direction.
出处
《计算机应用研究》
CSCD
北大核心
2010年第7期2430-2434,共5页
Application Research of Computers
关键词
人造目标检测
几何特征
分形
概率模型
水平集
聚类
性能评估
man-made object detection
geometry feature
fractal model
probabilistic model
level set
clustering
performance evaluation
