A Novel Model Based on AdaBoost and Deep CNN for Vehicle Classification

Abstract

Real-time vehicle classification is an important issue in intelligent transport systems. In this paper, we propose a novel model to classify five distinct groups of vehicle images from actual life based on AdaBoost algorithm and deep convolutional neural networks (CNNs). The experimental results demonstrate that the proposed model attains the highest classification accuracy of 99.50% on the test data set, while it takes only 28 ms to identify a vehicle image. This performance significantly outperforms the traditional algorithms, such as SIFT-SVM, HOG-SVM, and SURF-SVM. Moreover, the proposed deep CNN-based feature extractor has less parameters, thereby occupies much smaller storage resources as compared with the state-of-the-art CNN models. The high prediction accuracy and low storage cost confirm the effectiveness of our proposed model for vehicle classification in real time.

Keywords

• Real time
• vehicle classification
• CNN
• AdaBoost
• SVM