Road vehicle detection and classification based on Deep Neural Network

Abstract

The deep learning is a growing multi-layer neural network learning algorithm in the field of machine learning in recent years. Firstly, this paper analyzes the superiority of the deep learning at the aspect of feature extraction. Aimed at the lack of feature expression capacity and curse of dimensionality results from excessive feature dimensions of shallow learning, this paper proposes that using deep learning can extract high-lever features from low-lever features though its given layer structure. Secondly, the deep learning algorithm is applied in the case of road vehicle detection. Based on the traditional method, such as neural network the deep learning structure is further studied to increase the performance of feature extraction and classification recognition. Also, some tests are run in the Matlab software. The tests results show that with the increasing the amount of the data, the mean error and misclassification rate gradually decrease, so this algorithm based on the neural network has good superiority and adaptability of the deep learning. Finally, this paper proposes some suggestions for the improvement of the algorithm and prospects the development direction of the deep learning in the field of machine learning and artificial intelligence.