针对传统光谱匹配算子在“异物同谱”现象下对光谱精细化差异分辨能力较弱的情况,提出基于位置向量统计(PVS)的光谱匹配算子,同时提出通过匹配算子融合来提高目标识别的方法。PVS算子是在光谱吸收特征中吸收深度的一个延伸,算子首先利用位置向量对光谱曲线进行放大,然后利用投票统计的方法进行地物划分。实验结果表明,在检测概率为70%的情况下,PVS算子的虚警率在两个数据集上平均降低了1.73个百分点和4.77个百分点;同时在算子融合识别中,在检测概率为75.43%的情况下,融合PVS算子的虚警率在两个数据集上平均能够分别降低2.35个百分点和8.26个百分点。

In view of the weak resolution ability of traditional spectral matching operators in the phenomenon of “foreign objects in the same spectrum”, a spectral matching operator based on position vector measurement (PVS) is proposed, and a method of improving target recognition by fusion of matching operators is proposed. PVS operator is an extension of the absorption depth in the spectral absorption feature. The operator first uses the position vector to amplify the spectral curve, and then uses the method of voting statistics to divide the ground features. The experimental results show that when the detection probability is 70%, the false alarm rate of PVS operator is reduced by 1.73 percentage points and 4.77 percentage points on average in the two datasets. At the same time, in the case of the detection probability of 75.43%, the false alarm rate of the fused PVS operator can be reduced by 2.35 percentage points and 8.26 percentage points on average on the two datasets, respectively.