Voxel modeling system based on relative point density ratio and its control method

Abstract

The present invention relates to the field of 3D point cloud-based voxel modeling and Relative Point Density Ratio (RPDR) calculation technology for autonomous driving. More specifically, the invention relates to a relative point density ratio-based voxel modeling system and a control method thereof that enables the analysis of environmental occupancy characteristics, classification of uncertainty of low-density voxels, and visualization by configuring a point cloud acquired from a 3D sensor into a voxel map, calculating the maximum number of observable points per voxel through scanning ray-voxel intersection, and then calculating the ratio with the actual number of valid points to provide a normalized spatial point density corrected for distance and material dependency. According to the concept of the present invention, a 3D sensor data acquisition and coordinate alignment unit that periodically acquires point cloud data including distance and depth information using a 3D camera and aligns the acquired point cloud to the same reference coordinate system; a voxel map generation unit that divides the aligned point cloud data into square units of a predetermined size, converts them into a voxel-unit grid, and defines the center and boundary information of each voxel to quantify the point distribution in spatial units; a scanning ray generation and voxel intersection unit that defines scanning rays emitted within the camera's field of view and determines whether there is an intersection between each ray and a voxel to calculate the maximum expected number of points per voxel; an effective point counting unit that calculates the number of actual points contained within a voxel and aggregates the number of effective points per voxel; an RPDR calculation and normalization unit that calculates the relative point density ratio using the maximum expected number of points and the number of effective points, and normalizes the point density between voxels by removing distance and reflectance dependency; and a voxel whose relative point density ratio value is below a predetermined threshold, automatically identifies the voxel as an uncertain region, and indicates a region where a spatially low-density structure exists. The present invention provides a relative point density ratio-based voxel modeling system and a control method thereof, characterized by including an uncertainty tagging and classification unit, an RPDR-based environment comparison and analysis unit that quantitatively evaluates point density characteristics and structural complexity by comparing average relative point density ratio values ​​calculated in different experimental environments, and a visualization and result output unit that intuitively displays the point density distribution in space by visualizing the relative point density ratio values ​​in color.

Images