基于逻辑回归的匹配滤波器设计方法

Match Filter Design Method based on Logistic Regression

摘要:

针对遥感相机常规采样图像的目标检测问题,提出了基于逻辑回归的匹配滤波器设计方法;针对不同的图像目标特征,充分分析常规采样后图像上目标点的特征,选择全面且适合的学习样本；考虑到所获得的图像内混入杂波和噪声,因此,采用最大化信杂噪比作为逻辑回归算法的性能指标,实现匹配滤波模板的迭代优化;对比加入匹配滤波器前后目标临域内的信杂噪比,并将优化的匹配滤波器与文献中的最优滤波器相比,试验证明与文献中的滤波器相比,优化的匹配滤波器在目标检测方面更加有效。

Abstract:

A match filter design method based on logistic regression is present to solve the target detection for rotating camera images with single sampling. According to the characters of the targets in the image, analyze the target character with single sampling adequately, and chose the learning sample covering all the characters. Considering the clutter and noise are confused into the image, the Single Cluster Noise Ratio (SCNR) is introduced to evaluate the effectiveness of the match filter optimized. The optimization condition is the maximum of the SCNR, and the iteration optimization is utilized to obtain the module of the match filter. The simulation is completed to compare the performances of the match filter optimized using the method proposed in this paper and in the literature. The experiment indicates that it is more valid for the match filter optimized.