分层视觉特征感知在目标识别中的应用  

Application of Hierarchical Visual Perception in Target Recognition

作　　者：片兆宇[1] 史天予 袁德鹏[2] 胡玉兰[1] 王栋[1] Pian Zhaoyu;Shi Tianyu;Yuan Depeng;Hu Yulan;Wang Dong(School of Information Science and Engineering, Shenyang Ligong University, Shenyang 110159;School of Computer Science and Engineering, Northeastern University, Shenyang 110819)

机构地区：[1]沈阳理工大学信息科学与工程学院,沈阳110159 [2]东北大学计算机科学与工程学院,沈阳110819

出　　处：《计算机辅助设计与图形学学报》2017年第6期1093-1102,共10页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61373089);辽宁省教育厅科学技术项目(4801004yb02)

摘　　要：为了模拟视觉通路的特征抽象与自学习能力,在视神经信息分层处理机制的基础上提出一种特征感知模型.在改进卷积神经网络框架的基础上,首先依据视网膜中神经元的方向选择性、空间局部性以及神经元间的侧抑制性,在初级视觉特征处理中构建一种视网膜拓扑映射;然后在中级视觉特征处理中引入生物神经的稀疏表达法,构建神经突触激活函数,解决了神经计算中常见的过拟合问题;最后提出模拟腹部通路信息传递的具有计算感知不变性的分层视觉特征感知计算模型.应用不同规模数据集进行测试的结果表明,该模型对大规模的目标识别问题具有较好的识别效果,目标识别的平均准确率可达85%.To reproduce the ability of feature abstraction and self-learning of visual pathways,a feature perceptionmodel is proposed based on the hierarchical processing mechanism of optic nerve.On the basis of the improvedconvolution neural network,the model firstly established a retinal topological mapping in the primaryvisual feature processing according to the directional selectivity,the spatial locality and the lateral inhibition ofneurons in retina;and then,provided a neural synaptic activation function to solve the problem of over-fitting inneural computation,by introducing the sparse representation of the biological nerve in the secondary visual featureprocessing;finally,the hierarchical visual perception computing model with computing perceptual invariancewas proposed by reproducing information transmission of the ventral stream.The test results applying differentscale data sets show that the proposed model has sound recognition effect on large scale target recognition problems,and the average recognition accuracy can reach85%.

关 键 词：分层感知 卷积神经网络 激活函数 目标识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]