Spatial Calibration of Diffuse LiDARs

Abstract

Diffuse direct time-of-flight LiDARs report per-pixel depth histograms formed by aggregating photon returns over a wide instantaneous field of view, violating the single-ray assumption behind standard LiDAR-RGB calibration. We present a simple spatial calibration procedure that estimates, for each diffuse LiDAR pixel, its footprint (effective support region) and relative spatial sensitivity in a co-located RGB image plane. Using a scanned retroreflective patch with background subtraction, we recover per-pixel response maps that provide an explicit LiDAR-to-RGB correspondence for cross-modal alignment and fusion. We demonstrate the method on the ams OSRAM TMF8828.

Figures (5)

• Figure 1: Diffuse LiDAR spatial calibration overview. Diffuse LiDARs, used in consumer devices and mobile robots, (a) aggregate time-of-flight returns over wide per-pixel fields-of-view, producing spatially mixed measurements that make standard LiDAR-to-RGB calibration difficult. We describe (b-d) a simple method that calibrates diffuse LiDAR pixels by estimating each pixel’s effective support region and spatial weighting over a co-located RGB image plane. The (e) resulting per-pixel response maps provide an explicit LiDAR-to-RGB correspondence for accurate cross-modal alignment and fusion.
• Figure 2: Pixel aggregation modes on the ams OSRAM TMF8828 diffuse dToF LiDAR ams_osram_tmf882x. The sensor supports multiple spatial aggregation layouts; in each mode, reported pixels integrate photon returns over a wide instantaneous field-of-view under flood illumination. Our example calibration uses 3$\times$3 Wide mode ($P{=}9$).
• Figure 3: Custom rigid mount for co-located diffuse LiDAR (TMF8828) and RGB (RealSense D435i). Left: CAD mount design (dimensions in mm), showing the fixed sensor-to-sensor geometry and mounting hole layout. Right: real capture mount used for our calibration data collection.
• Figure 4: Retroreflective patch scan grid sampled with a UR10 robot arm. We traverse an $80{\times}45$ grid ($K{=}3600$) using a snake pattern to reduce motion between points. At each grid location, we record synchronized RGB frames and per-pixel LiDAR histograms; an identical patch-removed scan is also captured for background subtraction.
• Figure 5: Per-pixel spatial response maps for the TMF8828 3$\times$3 Wide mode overlaid on the co-located RGB image. Nonzero regions show the pixel's effective support in RGB coordinates, and the response magnitudes encode relative spatial sensitivity within that support.