为了解决核最近特征线与特征平面分类器在计算大数据样本量与高维数时工作量较大的问题,根据局部最近邻准则,提出针对这2种分类器的改进策略,使其不仅能够降低失效的可能性,而且在保证相近识别率的条件下,提高算法的实时性能,利用3类不同飞机实测距离像回波数据对其进行测试,实验结果表明,该改进策略是有效可行的。
In order to solve larger workload problems when Kernel Nearest Feature Line(KNFL) and Kernel Nearest Feature Plane(KNFP) classifiers compute large data size and high dimensionality, an improved strategy for these two classifiers is proposed according to locally nearest neighbor rule, which reduces the disabled possibility, and promotes the algorithms real-time performances under condition of similar recognition rate. It is tested by using the real range profiles of three types of aircrafts. Experimental results show this strategy is effective and feasible.