Railway fastener image recognition method based on multi feature fusion

6th Annual International Conference on Material Science and Engineering

Xu, Ruxun^1,2,3 ; Han, Jinyue^1 ; Qi, Wenzhe^1 ; Meng, Jianjun^1,2,3 ; Zhang, Hongqiang^2

Mechatronics TandR Institute, Lanzhou Jiaotong University, Lanzhou

730070, China^1

School of Mechatronic Engineering, Lanzhou Jiaotong University, Lanzhou

730070, China^2

Engineering Technology Center for Informatization of Logistics and Transport Equipment, Lanzhou

730070, China^3

关键词: Ada boost classifiers;    Classification efficiency;    Feature dimensions;    Feature extraction methods;    Multi-feature fusion;    Recognition methods;    SVM classifiers;    Training efficiency;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/397/1/012119/pdf
DOI  :  10.1088/1757-899X/397/1/012119

来源: IOP

【 摘 要 】

In order to improve the accuracy of railway fastener image recognition and training efficiency in the process of machine learning. Two traditional feature extraction methods, MB-LBP features and PHOG features are fused to make up a new image features for training in this paper. At the same time, to solve the problem of low training speed caused by increased feature dimension, introducing Adaboost-SVM classifier, the classification efficiency of Adaboost classifier will be reduced after a certain set of conditions, then the remaining samples are sent to the SVM classifier for classification. The results show that the accuracy of the fusion image feature is 3.9% and 5.8% higher than that of the individual MB-LBP and PHOG, respectively, and the training speed is also significantly improved by using the Adaboost-SVM classifier.