大数据环境下多代表点近邻传播的微小图像特征智能识别与仿真

Simulation of Intelligent Recognition of Small Image Features Based on Multi-representative Neighborhood Propagation in Big Data Environment

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.为了解决上述图像缺陷,提出了一种大数据环境下多代表点近邻传播的微小图像特征智能识别与仿真方法.采用本方法合成当前识别画面时需要将正向映射与逆向映射进行结合应用,同时采用图像识别多代表点近邻传播特性从各个源参考图像中获得弥补空洞所需的像素点数据,以解决对整幅参考图像从深度角度进行比对与智能识别.最后,通过实验研究分析表明:文中所提的智能识别方法可以有效提高微小图像特征智能识别的准确率,加强识别算法的鲁棒性.

In traditional methods,small images are filtered and denoised to obtain smooth images,and the features of invariable data of small images are obtained. However,the pretreatment of them is neglected,leading to low recognition accuracy. In order to solve the above image defects,a simulation method for small image feature recognition based on multi representative neighborhood propagation in big data environment is proposed in this paper. When we use this method to synthesize the current recognition picture,we need to combine forward mapping with reverse mapping. At the same time,the multi-point neighbor propagation characteristics of image recognition are used to obtain the pixel data needed to make up the hole from each source reference image,so as to solve the comparison and intelligent recognition of the whole reference image from the depth angle. Finally,the experimental analysis shows that the intelligent recognition method proposed in this paper can effectively improve the accuracy of intelligent recognition of small image features and enhance the robustness of the recognition algorithm.