基于双近邻模式和最小二乘支持向量机的SAR景象匹配区选择

Selection of Suitable SAR Scene Matching Area Based on Dual-neighbor Pattern and Least Squares Support Vector Machines

通过对景象匹配过程的分析,从模式识别的角度阐述了误匹配产生的原因。从避免误匹配的角度定义了双近邻度、最小距离以反映SAR景象的独特性和匹配的准确性,并结合反映地面景物稳定性的边缘密度,构建反映SAR景象适配性的分类特征向量。基于该分类特征向量,利用最小二乘支持向量机将SAR景象基准图子图划分为匹配正确的子图和匹配错误的子图,并由匹配正确的子图类构成SAR景象适配区。试验结果表明,提出的方法能够有效地规划出所需的SAR景象匹配区。

Selection of suitable matching area is an important element for scene matching navigation system. The reasons for mismatch are expatiated in the view of pattern recognition on the basis of analyzing the process of scene matching. In order to avoid mismatch, dual-neighbor degree is introduced to measure the uniqueness of SAR scene and minimum distance is introduced to estimate the accuracy of scene matching. Combined with edge density which represents the scene stability, the classified feature vector is constructed to reflect the suitability of SAR scene. Based on the classified feature vector, least squares support vector machines （LS-SVM） is employed to divide the sub-images of reference map into two classes. The first class consists of sub-images which can be matched correctly, and the second is composed of mismatched subimages. And suitable matching area is composed of the first class sub-images. An approach of selection suitable SAR scene matching area based on dual-neighbor pattern and LS-SVM is proposed. At the end, experiments are performed and the results show that this approach can effectively single out the suitable SAR scene matching area.