基于DF-Stacking模型的交通标志识别

Traffic Sign Recognition Based on DF-Stacking Model

针对传统机器学习方法抗噪能力欠佳、深度学习方法需依赖大量训练样本等问题,提出基于深度集成森林(DF-Stacking)的交通标志识别方法。采用多粒度扫描结合级联森林提取图像特征,将所得特征输入Stacking集成模块以分类图像。结果表明:DF-Stacking模型在使用少量训练样本时,特征提取精度比深度学习方法有明显提高;模型在高斯噪声、运动模糊等条件下的分类精度均高于单分类器方法,体现出较强的泛化能力。

To address the problems that traditional machine learning methods have poor noise immunity and deep learning methods rely on a large number of training samples,this paper proposes a traffic sign recognition method based on deep forest-Stacking(DF-Stacking).Multi-grained scanning combined with cascade forest is used to extract image features.The obtained features are input to the Stacking ensemble module to classify images.The results show that the DF-Stacking model has significantly improved feature extraction accuracy over the deep learning method when using a small number of training samples.In addition,the classification accuracy of the proposed DF-Stacking model is higher than that of the single classifier method under the conditions of Gaussian noise and motion blur,reflecting the strong generalization ability.