基于Logistic回归模型的景像匹配算法性能评估

Performance Evaluation of Image Registration Algorithm Based on Logistic Regression Model

景像匹配算法性能评估的目的是通过科学客观地评估算法性能,选择满足应用需要的匹配算法及进一步改进算法。将自动目标识别(ATR)算法性能评估中常用的Logistic回归模型应用于景像匹配算法性能的定量评估,进一步深入研究了匹配试验所需的样本量和试验结果的定量分析方法,并通过对两种匹配算法的定量评估试验验证了方法的有效性。

The purpose of the performance evaluation of image registration algorithms is to choose a registration algorithms and improving the algorithms by evaluating their performance scientifically. Logistic regression model generally used in ATR algorithms performance evaluation can also be used in quantitative performance evaluation of registration algorithms. Moreover, the amount of samples needed in registration experiments and quantitativel analysis methods for experimental results are deeply researched. Then the feasibility is proved by evaluation experiment.