Point cloud noise reduction is crucial to the accuracy of light detection and ranging (LiDAR) imaging systems. In order to reduce the noise caused by receiver, multipath effect, external interference and atmospheric disturbances, a radius filtering method based on genetic algorithm was used for noise reduction, which optimized the key parameters of radius filtering (filtering radius and nearest-neighbor thresholds) by genetic algorithm. Validation in simple and complex scenes show that, the algorithm maintains denoising accuracy and point retention in simple scenes while slightly improving noise recall. The noise recall in complex scenes is improved by about 21% over traditional radius filtering and about 16% over statistical filtering, which is useful for point cloud radius filtering. The radius filtering based on genetic algorithm provides a novel and effective method for LiDAR data processing, which is valuable for improving the quality of LiDAR imaging.