Deep Learning Based Scene Parsing Algorithm For Intelligent Vehicle

Recent years, as the strongly support of much car companies as well as Internet companies, the development of the intelligent technology was extremely rapid. Wheanwhile, deep learning technology was applied in various fields widely. The technology of scene parsing is the basic technology in intelligent vehicle. This paper tried the application of deep learning in the scene parsing technology of intelligent vehicle. Based on the SIFT FLOW dataset, the Stanford Background dataset and the BJFTH dataset, the following study work was completed.1. A new scene parsing algorithm based on convolutional neural network was presented, and the advantages and disadvantages were given in this paper by the analysis of the experiment results.2. A new scene parsing algorithm based on Recurrent convolutional neural network was presented and the analysis of experiment results was given,either.3. A new scene parsing algorithm based on TD convolutional neural network was presented. We also did the analysis of the experiment results and the optimization method is given.Experiment results show that all the models get the state of the art result and the model based on recurrent neural network obtained a outstanding result. This paper indicate the feasibility to applied deep learning technology in intelligent vehicle scene parsing task.