With the rapid development of artificial intelligence and autonomous driving technology, the demand for high-precision, high-reliability and continuous positioning services has become increasingly obvious. However, in complex urban environments, GNSS signals are prone to the non-line-of-sight (NLOS) propagation effect, which leads to systematically large observation errors and then significantly reduces the navigation accuracy. To address this, we propose a tightly coupled LiDAR/IMU/GNSS navigation framework based on raw GNSS observations. Additionally, we incorporate LiDAR point cloud data to develop a NLOS satellite detection and correction module. This module constructs a 3D LiDAR point cloud map of the sensor’s surroundings and identifies NLOS signals by analysing the geometric relationships between the sensor, satellites, and the environmental map. Furthermore, reflection points from the surrounding environment are extracted and utilized for NLOS correction. The results of two groups of independent experiments show that the system positioning error after NLOS correction is reduced by 16.15%. Compared with the conventional integration system that adopts pseudorange difference information, the proposed framework achieves a 32.57% improvement in navigation accuracy under complex urban scenarios, demonstrating its effectiveness.