Current 3D object detection models are predominantly based on data-driven deep learning techniques， so that dataset quality plays a pivotal role in model performance. Aiming at the scarcity of coal dust environment data and the time-consuming and labor-intensive construction process of real-world datasets， a point cloud data augmentation method was proposed on the basis of scattering and absorption effects of coal dust on Light Detection and Ranging （LiDAR） electromagnetic waves. In the method， by considering the optical characteristics of coal dust particles， a propagation simulation model of LiDAR electromagnetic waves was built to characterize LiDAR signal attenuation and scattering phenomena in coal dust environment. Secondly， based on real point cloud data acquired under clear weather conditions， 3D coordinates and reflection intensities were adjusted via the simulation model， so as to generate simulated point cloud data that conforms to the perception characteristics of coal dust environment. Finally， five mainstream 3D models （PV-RCNN++， PV-RCNN， PointRCNN， PointPillars， Voxel-RCNN_Car） were trained and tested on the augmented dataset. The results demonstrate that the proposed method improves the detection precision of these five detection models in coal dust environment. For the most complex model， PV-RCNN， improved performance is 1.88， 1.74， and 0.84 percentage points in the car， pedestrian， and cyclist categories， respectively. It can be seen that in comparison with models trained in clear weather conditions， using the augmented point cloud data to train object detection models improves detection precision in coal dust environment significantly， so as to enable more reliable perception in complex environments in open-pit mines， thereby providing data support for stable operation of autonomous mine carts.