基于多元Logistic回归分析的匹配算法性能评估  

Performance Evaluation of Image Registration Algorithm Based on Multiple Logistic Regression Model

作　　者：王洋[1] 徐刚锋[2] 左东广[1] 

机构地区：[1]第二炮兵工程学院101教研室,陕西西安710025 [2]第二炮兵装备研究院,北京100085

出　　处：《红外技术》2008年第10期563-566,共4页Infrared Technology

摘　　要：在实际应用中,景像匹配算法的性能受旋转、缩放、平移等多种因素影响。选取适当的图像性能指标和算法性能指标作为输入和输出,利用多元Logistic回归模型建立两者之间的联系,通过回归分析评估匹配算法性能优劣。该方法能够综合多种影响因素评估算法性能,并确定各因素与算法性能指标的相关程度,为算法应用及改进提供依据。通过多元回归分析对两种匹配算法的性能进行了定量评估与比较,验证了评估模型的有效性。In practice, the performance of image registration algorithms is impacted by many kinds of factors, such as translation, rotation, and scale, etc. The relation between image metrics and algorithm performance metrics is established using multiple Logistic regression model, and regression analysis is used to evaluate algorithms performance. By this method, Algorithms performance can be evaluated with complexly considering multi-factors, and the correlation between every image metric and every algorithm performance metric is calculated. That is used to select or use algorithms. Quantitative evaluation and comparison are carried out for two algorithms performance by multiple regression analysis, and the feasibility has been obtained.

关 键 词：匹配算法 性能评估 多元Logistic回归模型 

分 类 号：TN911.73[电子电信—通信与信息系统]