基于模糊神经网络的视频图像目标精准识别研究  

Research on Accurate Recognition of Video Image Target Based on Fuzzy Neural Network

作　　者：王学忠[1] WANG Xuezhong(School of Electronic&Electrical Engineering,Anhui Sanlian University,Hefei Anhui 230601,China)

机构地区：[1]安徽三联学院电子电气工程学院,安徽合肥230601

出　　处：《信息与电脑》2022年第16期188-190,共3页Information & Computer

基　　金：安徽高校自然科学研究项目“基于深度学习的暴力检测及CNN人脸图像识别算法研究”(项目编号:KJ2021A1174)。

摘　　要：由于对特征的分析结果存在误差,导致对视频图像目标识别的精度较低,因此提出基于模糊神经网络的视频图像目标精准识别方法。首先,利用Relief算法选择目标特征,根据特征与同类样本和不同类别样本之间距离的关系,对特征进行个性化赋权;其次,将权重最高的m个特征作为模糊神经网络识别的特征。最后,在计算模糊化层各个节点的高斯隶属度函数参数后,采用误差反向传播算法计算模糊神经网络各层级之间的梯度,并反向逐层计算特征的累积误差,实现对视频图像目标的识别。测试结果表明,设计方法的识别精度最大值可以达到96.44%,最小值达到88.62%,具有良好的识别效果。Due to the errors in the analysis results of the features,the accuracy of the target recognition in the video image is low.Therefore,a research on the accurate recognition of the video image target based on the Fuzzy Neural Network is proposed.Firstly,the Relief algorithm is used to select the target feature.According to the relationship between the feature and the distance between the same sample and different categories of samples.Secondly,the features are individually weighted,and the m features with the highest weight are used as the features identified by the Fuzzy Neural Network.Finally,after calculating the Gaussian membership function parameters of each node of the fuzzification layer,the error back propagation algorithm is used to design the gradient between each level of the Fuzzy Neural Network,and the cumulative error of the feature is calculated in reverse layer by layer,so as to realize the accuracy of the video image target identify.The test results show that the maximum recognition accuracy of the design method can reach 96.44%,and the minimum value can reach 88.62%,which has a good recognition effect.

关 键 词：模糊神经网络(FNN) 视频图像目标 RELIEF算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]