Abstract Automatic identification of the brand of the car is difficult to accomplish because the different brands have a lot of similarities. In this work, we developed a new system of Volkswagen vehicle identification. Firstly, the original car images were preprocessed by the watershed algorithm and manual revision. Secondly, we employed the wavelet entropy (WE) to extract efficient features from the car images. Thirdly, we used multilayer perceptron (MLP) as a classifier. At last, we chose the artificial bee colony (ABC) algorithm to train the MLP. The original ABC is good at exploration but poor at exploitation because of its equation system. So, we proposed a new model of the ABC called improved artificial bee colony (IABC) to balance exploration and exploitation. We used the 5 × 5-fold cross-validation for fair comparison. The experiment result showed that the overall specificity is 88.62%, the overall sensitivity is 89.16%, and the overall accuracy is 89.17%. Therefore, the proposed method is effective for Volkswagen vehicle identification. The result of the IABC provides better performance than ordinary ABC.
Volkswagen Vehicle Identification via Multilayer Perceptron Trained by Improved Artificial Bee Colony Algorithm
2019-10-02
11 pages
Article/Chapter (Book)
Electronic Resource
English
Volkswagen vehicle , Identification , Wavelet entropy , Multilayer perceptron , Improved artificial bee colony algorithm , k-fold cross-validation Engineering , Computational Intelligence , Data Mining and Knowledge Discovery , Artificial Intelligence , Complexity , Information Systems Applications (incl. Internet)
Modular, Multilayer Perceptron
NTRS | 1991
|An improved artificial bee colony algorithm for solving parameter identification problems
British Library Online Contents | 2017
|Improved multi-strategy artificial bee colony algorithm
British Library Online Contents | 2016
|Research on Sound Source Localization Algorithm Based on Multilayer Perceptron
Springer Verlag | 2021
|