Improvement of Self-Driving Algorithm with Traffic Command Recognition and Vehicle Information Interaction

Self-driving technology has been studied and developed for a long time and gradually tends to mature. However, we want to complete the fully self-driving under the smart city, whether in self-driving cars or uncrewed express vehicles and other vehicles. However, there are still many problems with traffic command and vehicle interworking during the car's driving. In this article, based on the two problems mentioned above, the authors improve the existing self-driving algorithm from these two aspects. On the one hand, the authors use the OpenPose to deal with 3-D motion and gestures and experiment on static images and static video of traffic gestures, the model can accurately segment various traffic information including traffic indication gestures in the target, and give feedback based on the set priority. On the other hand, by simulating vehicle information experiments, the algorithm can process nearby information and makes corresponding pre-processing according to the processing results. These two improvements not only make the existing self-driving algorithm more perfect but also make the surrounding road condition information predictable, which means that the self-driving technology becomes more flexible and safer.