A Quantitative Comparison of Calibration Methods for RGB-D Sensors Using Different Technologies

Area of Science:

• Computer Vision
• Sensor Technology
• Metrology

Background:

• RGB-D sensors provide simultaneous color and depth data, finding broad applications in robotics, CAD, and entertainment.
• Despite commercial growth and low cost, RGB-D sensors often operate at the limit of their sensitivity, necessitating accurate calibration.
• Existing research lacks comparative studies of calibration algorithms across multiple RGB-D sensor types.

Purpose of the Study:

• To conduct a comparative analysis of the three most prevalent calibration methods for RGB-D sensors.
• To evaluate the impact of different calibration techniques on depth measurement accuracy.
• To assess sensor performance in object reconstruction tasks at the edge of sensitivity limits.

Main Methods:

• Applied three common calibration algorithms to three distinct RGB-D sensors (structured light and time-of-flight technologies).
• Designed experiments to quantitatively measure depth accuracy across different calibration methods.
• Utilized an object reconstruction application to test sensor performance under demanding conditions.

Main Results:

• The study provides a comparative evaluation of calibration method performance on multiple RGB-D sensors.
• Quantitative measurements and visual inspection were used to assess the accuracy of depth data and reconstructed objects.
• Results highlight the critical role of calibration in achieving required accuracy for sensitive applications.

Conclusions:

• Calibration is essential for enhancing the accuracy of RGB-D sensors, particularly in applications near their sensitivity limits.
• The comparative study offers valuable insights for selecting appropriate calibration methods based on sensor type and application requirements.
• This research addresses a gap in the literature by providing a multi-sensor, multi-method calibration performance analysis.