Research on Environmental Feature Recognition Algorithm of Emergency Braking System for Autonomous Vehicles

Abstract

In order to quickly and accurately identify the environmental characteristics of emergency braking systems for autonomous vehicles, it is convenient to build and evaluate emergency braking algorithms for autonomous vehicles. Aiming at the current deficiency of yellow and white lane line feature recognition, a method for rapid lane line calibration and detection recognition is proposed. The yellow and white lane line figures of the actual road are identified in LAB and LUV color spaces, and a quadratic curve model is established to restore the lane geometry model and provide a basis for rapid geometric positioning of the vehicle when the automatic emergency braking system algorithm is built. In order to handle with problems such as heavy calculation and low accuracy of vehicle target recognition, the deep neural network learning theory is analyzed, and a vehicle feature detection and recognition algorithm based on a convolutional neural network algorithm is built to detect and classify environmental vehicle targets. The network training recognition accuracy rates are up to 87.1%. The images of typical road conditions in South China were collected, and the network was tested and verified. The results show that the built environmental feature recognition algorithm can meet the requirements of accuracy and real-time performance.