3D Occupancy Network Modelling for Multi-view Image Fusion Techniques in Autonomous Driving

The article outlines the principle and mathematical model of multi-view image fusion technology and acquires research data based on multi-view image fusion technology. Combining research data, vehicle structural parameters, and Ackermann’s steering principle, the vehicle kinematics model is constructed to complete the task of 3D occupancy network modeling. The loss function of the selected network is discussed and its training is optimized. The simulation analysis shows that along with the increase in vehicle distance, the vehicle distance detection error increases by 6.59, and the average detection error is 3.72%, which is within the error allowance. In addition, in the path planning simulation analysis, the path planning time of this paper’s method is, which indicates that the 3D occupancy network can quickly and efficiently plan an effective and feasible path according to the change of obstacle position, and it practices the principle of safe and automatic driving well.