Study on Road Condition Recognition Method Based on Convolutional Neural Network

Feng Jiang, Ning Wei, Yan Zhou, Hengjun Niu
2018 Proceedings of the 2018 International Conference on Network, Communication, Computer Engineering (NCCE 2018)   unpublished
In recent years, frequent traffic accidents, traffic accidents captured countless people's lives, so the road safety problem attracts more and more attention of common people. This paper studies the problem of automatic recognition of road condition, the purpose is to improve the traffic information detection accuracy and robustness. In general, the automatic identification of vehicle targets is vulnerable to all kinds of weather, such as rainy weather, snow weather and sandstorm weather which
more » ... ill weaken the radar target visibility to the public, so it has potential danger. For a long time, the situation caused by this natural deterioration of road safety has not been effectively resolved, or that there is no substantive solution, this is obviously an embarrassing question. From a certain extent, can accurately and effectively carry out the automatic identification of traffic information, thereby reducing the incidence of traffic accidents, but also a symbol of a country's comprehensive strength. In order to avoid the traditional target recognition characteristics easily affected by bad weather caused instability and robustness is not good, in order to overcome this difficulty, we propose a method of road object recognition based on convolution neural network. A large number of experimental results show that this method is very effective, has a high recognition rate, but also has good robustness; the simulation result is a good proof. This method has good adaptability to harsh natural weather, can overcome these unfavorable factors, believe that this method for the implementation of the project to a great extent guide convenience, this is a very meaningful recognition method for the road of the goal.
doi:10.2991/ncce-18.2018.21 fatcat:milczgq4avh5pewzpoep3lc4py