Ford vehicle identification via shallow neural network trained by particle swarm optimization

Abstract

Automatic identification of the car manufacturer is difficult to achieve because of the similarity among the different brands. In this work, we propose a new system of Ford vehicle identification. Firstly, we captured the side view of the car image. Secondly, we employed the wavelet entropy (WE) to extract efficient features from car images. Thirdly, we employed a shallow neural network (SNN) as a classifier. Finally, we used the particle swarm optimization to train the classifier. The 10 10-fold cross validation on a data set containing 220 vehicle images showed that our Ford vehicle identification system obtained the overall sensitivity of 83.27±1.61%. The overall specificity is 83.91± 1.87%, the overall accuracy is 83.59± 0.94%. Experiment result show that the proposed system is effective for Ford vehicle identification.