Investigating the factors that influence 3D stereo depth sensor noise

Published

April 18, 2025

Author(s)

Terrence Pierce, Prem Rachakonda

Abstract

3D stereo depth sensors have a variety of applications, including sensing for autonomous vehicles, reverse engineering, and manufacturing automation. The performance of these sensors can be affected by various factors, such as sensor construction, sensor technology, sensor settings, environment, etc. Understanding parameters that affect sensor output is needed to characterize them and to develop standards. As machine learning with 3D point clouds and depth data becomes increasingly prevalent, understanding the data used with these models becomes crucial for improving the adoption rates of such depth sensors. In certain domains, sensor noise and transient effects can become dominant. Reducing or at least considering these effects before using sensor data with machine learning algorithms is necessary. To characterize depth sensors, we conducted experiments using targets with varying gloss, color, and texture/pattern. Additionally, we studied sensor data quality and noise by exploring the sensor parameters such as exposure, gain, and laser power. We have found transient effects in both 2D images and depth data captured by sensors. These experiments help inform the operating conditions that could be advised for specific applications and future standards addressing these sensors.

Proceedings Title

Proceedings of the 2025 SPIE Defense and Commercial Sensing conference

Conference Dates

April 13-17, 2025

Conference Location

Kissimmee, FL, US

Conference Title

SPIE Defense + Commercial Sensing

Pub Type

Conferences

Download Paper

Local Download

Keywords

Stereoscopic Depth, Sensor Noise, Commercial Sensing, Standards, Metrology, Machine Learning

Citation

Pierce, T. and Rachakonda, P. (2025), Investigating the factors that influence 3D stereo depth sensor noise, Proceedings of the 2025 SPIE Defense and Commercial Sensing conference, Kissimmee, FL, US, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=959666 (Accessed April 27, 2025)