LiDAR Noise Elimination via Statistical Point Density Analysis

The patent extracts and removes noise points from the point cloud data using statistical outlier detection methods. By calculating the density of adjacent points for each point and comparing it against dynamic thresholds, the system identifies and eliminates noise points while preserving valid object points, thereby reducing noise without requiring complex system architecture

The patent replaces complex learning-based noise reduction systems with a statistical calculation approach. Instead of using trained neural networks or complex segmentation algorithms, the system uses straightforward statistical metrics (point density, distance calculations) to identify and remove noise, significantly reducing computational complexity while maintaining effective noise reduction

The patent applies partial action by focusing computational efforts only on points that are likely to be noise based on their local density characteristics. Rather than processing the entire point cloud uniformly or using computationally intensive learning models, the system calculates statistical metrics only where necessary and applies dynamic thresholding to efficiently identify noise points for removal

The patent segments the point cloud processing into distinct stages: first identifying candidate noise points through density calculation, then applying dynamic threshold filtering, and finally removing identified noise. This segmented approach allows the system to process only relevant portions of the data at each stage, reducing overall processing time compared to applying complex learning-based methods to the entire point cloud