An effective vehicle road scene recognition based on improved deep learning

Abstract

In order to improve the recognition for autonomous vehicles with respect to road scenes, a new activation function called ReLU sigmoid is proposed based on ReLU and the sigmoid model, which resolves the issue of neuron necrosis in the ReLU model. Through an analysis of the action principle of the activation function, the key aspects of the activation function design are determined. The sigmoid and ReLU are combined in the positive and negative half axes of the semantic axis to optimize the road scene recognition model. The experimental results reveal that compared with the ReLU and LReLU models, the ReLU sigmoid model effectively improves the recognition accuracy of the convolutional neural network for the road scene, from 75.12% to 67.15%, respectively, to 83.70%. It is proved that the algorithm can improve the recognition performance of the deep learning model for the road scene and alleviate the phenomenon of neuron necrosis, along with improving the vehicle's perception of the road environment.