相干激光雷达距离像与目标表面物理结构特性密切相关,体现目标的本质特征,在目标识别领域引起广泛关注。数据采集过程和采集成本决定了激光雷达不容易采集到大量的图像。在小样本情况下,随着特征维数的增加,识别率可能会下降,即出现休斯现象。为此,把两种特征选择算法——Relief算法和支持向量机回归特征消去(SVM-RFE)算法引入到距离像目标识别。仿真实验结果表明,在3个训练样本时,利用Relief和SVM-RFE算法,可以解决由三组组合矩(Hu矩和仿射矩,仿射矩和Zernike矩以及仿射矩、Hu矩和Zernike矩)引起的休斯现象,并且基于SVM-RFE算法的识别性能略好于Relief算法的识别性能。 The coherent laser radar distance image is closely related to the physical structure of the target surface and reflects the essential characteristics of the target, arousing widespread concern in the field of target recognition. Data acquisition process and acquisition costs determine the laser radar is not easy to capture a large number of images. In the case of small samples, the recognition rate may decrease with the increase of feature dimension, which means Hughes phenomenon. To this end, the two feature selection algorithms - Relief algorithm and support vector machine regression feature elimination (SVM-RFE) algorithm is introduced to the distance to the target recognition. Simulation results show that the three combined moments (Hu moments and affine moments, affine moment and Zernike moments, affine moments, Hu moments and Zernike moments) can be solved using the Relief and SVM-RFE algorithms in three training samples Hughes phenomenon, and the recognition performance based on the SVM-RFE algorithm is slightly better than the Relief algorithm recognition performance.