基于视通路非对称STDP机制的边缘提取  

Edge Extraction Based on Asymmetric STDP of Visual Pathway

作　　者：房涛 方琳灵 刘晓磊 刘艳 潘树文 FANG Tao;FANG Linling;LIU Xiaolei;LIU Yan;PAN Shuwen(College of Electronics and Information,Hangzhou Dianzi University,Hangzhou Zhejiang,310018,China;Department of Rich Communication Suite,Research and Development Center,China Mobile(Hangzhou)Information Technology Co.,Ltd.,Hangzhou Zhejiang 311121,China;School of Electronics and Information Engineering,Harbin Institute of Technology,Harbin Heilongjiang 150080,China;School of Information&Electrical Engineering,Zhejiang University City College,Hangzhou Zhejiang 310015,China)

机构地区：[1]杭州电子科技大学电子信息学院,浙江杭州310018 [2]中国移动杭州研发中心融合通信系统部,浙江杭州311121 [3]哈尔滨工业大学电子与信息工程学院,黑龙江哈尔滨150080 [4]浙大城市学院信息与电气工程学院,浙江杭州310015

出　　处：《传感技术学报》2023年第2期280-286,共7页Chinese Journal of Sensors and Actuators

基　　金：教育部产学合作协同育人项目(202102019039);浙大城市学院培育基金项目(J-202223)。

摘　　要：图像边缘的有效检测在图象处理中具有极其重要的作用,对后续图像内容的分析、识别和理解具有至关重要的意义。本文模拟视觉信息处理机制,提出了一种基于视觉生理机制的非对称脉冲时间依赖可塑性(Spike-Timing-Dependent Plasticity,STDP)图像边缘检测的新方法。首先,通过Gabor滤波器模拟视觉生理朝向特性对原始图像预处理,并将朝向特征融合重构出图像的初级边缘特征信息;其次,构建基于非对称STDP机制的由动态突触组成的神经元网络对初级边缘特征信息进行进一步加工,获得脉冲响应图像,并通过高斯滤波器对脉冲响应图像进行滤波;然后利用神经元之间的侧向抑制作用,对滤波后的图像边缘进行精细化处理;最后对结果进行归一化处理,得到待检测实验图像的最终边缘。实验对象为实验室采集的菌落图像和部分随机挑选的公共数据集图像,菌落图像的边缘检测对于后续信息的统计具有重要意义,通过实验结果的定量与定性分析,所提出的方法在不同实验图像上的AUC值和信息熵优于已有的传统方法,并保留更多的图像边缘细节信息。所提出的边缘检测方法,根据视觉机制避免了黑箱模拟,为低级图像特征处理提供了新的思路。The effective detection of image edges plays an extremely important role in image processing,and is of great significance to the analysis,recognition and understanding of subsequent image content.The visual information processing mechanism is simulated,and a new method for image edge detection is proposed based on asymmetrical Spike-Timing-Dependent Plasticity(STDP)according to the physiological mechanism of vision.Firstly,the original image is preprocessed by simulating the visual physiological orientation characteristics through the Gabor filter,and the orientation features are fused to reconstruct the primary edge feature information of the image.Secondly,a neuron network composed of dynamic synapses based on the asymmetric STDP mechanism is constructed to further process the primary edge feature information to obtain an impulse response image,and the impulse response image is filtered by a Gaussian filter.The edge of the filtered image is then refined by using the lateral inhibition between neurons.Finally,the results are normalized to obtain the final edge of the experimental image to be detected.The experimental objects are the colony images collected in the laboratory and some randomly selected public dataset images.The edge detection of colony images is of great significance for the statistics of subsequent information.Through the quantitative and qualitative analysis of the experimental results,the AUC value and information entropy of the proposed method are better than the existing traditional methods on different experimental images,which can retain more image edge details.The edge detection method proposed avoids the black box simulation according to the visual mechanism,and provides a new idea for low-level image feature processing.

关 键 词：边缘提取 非对称STDP 脉冲响应 动态突触 侧向抑制 

分 类 号：TN911.73[电子电信—通信与信息系统] TP183[电子电信—信息与通信工程]