Research on Multi-Target Tracking Model of Data Fusion between Autonomous Driving Millimeter-Wave Radar and Visual Sensors

The research mainly addresses the multi-target tracking problem encountered in the data fusion process of millimeter-wave radar and visual sensors in unmanned systems. Multi-sensor data fusion models are investigated to optimize target detection performance and improve the accuracy of tracking models. The research analyzed the accuracy of the Kalman filter and the convolutional neural network, and used joint data association technology to achieve high-precision target detection, stable real-time tracking, and excellent adaptability to dynamic environments. The tracking accuracy reached 97.87% and the false detection rate was less than 2.13% through simulation experiments in a simulated driving environment. The experimental results verified that the fusion model performs excellently in complex driving scenarios, maintaining high-precision target tracking, proving that the technical methods used are feasible.