Analysis of Target Detection Technology in the Field of Autonomous Driving by Integrating Lidar and Camera

Environmental perception is the core support for the implementation of autonomous driving technology. The heterogeneous fusion of LiDAR and cameras, through the complementary advantages of geometric and semantic information, effectively breaks through the performance bottleneck of a single sensor. This paper systematically reviews the technological evolution in this field, constructs a dual-dimensional classification system of the fusion stage and data form, and deeply analyzes the innovative mechanisms and dataset performance of representative technologies such as BEVFusion and BEVFormer. Research shows that the core technology achieves an NDS level of 69.2% to 74.1% in 3D object detection tasks, while revealing core bottlenecks such as insufficient cross-modal spatiotemporal homogeneity robustness, difficulty in balancing real-time performance and accuracy, and limited generalization ability. Finally, we look forward to the development directions such as dynamic elastic integration, software and hardware collaborative optimization, generalization and expansion, and standardized implementation, providing important references for the research and development and engineering application of high-level autonomous driving perception systems.