基于稀疏表示和深度学习的图像识别算法研究  

Research on Image Recognition Algorithm Based on Sparse Representation and Depth Learning

作　　者：吕焦盛 LYU Jiaosheng(College of Information Engineering,Zhengzhou University of Industrial Technology,Zhengzhou 451100,China)

机构地区：[1]郑州工业应用技术学院信息工程学院,河南郑州451100

出　　处：《新乡学院学报》2018年第9期31-34,共4页Journal of Xinxiang University

基　　金：河南省科技攻关计划项目(0721002210032)

摘　　要：针对传统的图像稀疏表示方法只注重提高特征的提取率而忽略了图像的多尺度信息,在对图像稀疏表示的过程中易受噪声干扰,及系统鲁棒性不高等问题,提出了多尺度稀疏表示方法。针对图像底层像素特征不稳定,易受环境影响,且不能反映语义信息等问题,提出利用深度网络具有的高层特征抽取这一能力,将多尺度稀疏表示和深度学习网络结合起来,构建完善的学习框架。实验结果表明,该算法可以提高图像的识别性能,同时增强系统的鲁棒性。Aiming at the traditional image sparse representation method, it only focuses on improving the extraction rate of featuresand neglects the multi-scale information of the image, and in the process of image sparse representation susceptible to noiseinterference, system robustness is not high, the proposed multi-scale sparse representation. At the same time, because the underlyingpixel features of the image are unstable and sensitive to the environment, and can not reflect the semantic information, this paperproposesa new learning framework that uses deep networks to extract high -level features and combines multi -scale sparserepresentation with deep learning networks. Experiments show that the algorithm can improve the recognition performance of theimage and enhance the robustness of the system.

关 键 词：特征提取 稀疏表示 深度学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]