大数据环境下微小图像特征智能识别仿真研究

Research on the Simulation of Micro Image Feature Recognition in Large Data Environment

对微小图像特征的智能识别,能够有效提高对微小图像本质特征的识别率,对各个领域中的微小图像研究均具有重要意义。对大数据特征的识别,需要将微小图像相同尺度不同方向的特征进行融合,并对其进行预处理,完成对微小图像特征的智能识别。传统方法对微小图像进行滤波去噪处理获得平滑图像,获取微小图像光照不变量数据的特征,但忽略了先对其进行预处理,导致识别精度偏低。提出基于克隆选择的大数据环境下微小图像特征智能识别方法。上述方法先对微小图像进行Gabor变换,按照两个融合规则将微小图像相同尺度不同方向的特征进行融合,得到微小图像5个尺度上的融合特征,采用克隆选择理论对微小图像融合特征进行预处理,形成微小图像的抗体库,结合K近邻判别理论完成待识别微小图像特征智能识别。实验结果表明,所提方法能够有效提升微小图像特征平均识别率,且鲁棒性较好。

This paper proposes a method for feature intelligent recognition of tiny image under big data environment. Firstly,the paper carried out Gabor conversion for tiny image and fused feature of tiny image with same scale and different direction according to two fusion rules ,then obtained fusion feature at five scales of tiny images. Moreo- ver,we used clone selection to carry out pretreatment for the fusion feature and generated antibody library of the tiny image. Finally,integrated with K neighbor judgment,the feature intelligent recognition was completed. Experimental results show that the method can improve average rate of recognition of tiny image feature effectively and has better robustness.