The deep learning is a growing multi-layer neural network learning algorithm in the field of machine learning in
recent years. Firstly, 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
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