Evaluating the Accuracy of the Azure Kinect and Kinect v2.

Area of Science:

• Computer Vision
• Robotics
• Sensor Technology

Background:

• The Microsoft Kinect revolutionized depth sensing for consumer applications.
• Advancements in depth camera technology are crucial for 3D scanning and spatial computing.
• The Azure Kinect is the successor to the Kinect v2, necessitating a performance comparison.

Purpose of the Study:

• To evaluate and compare the depth and spatial accuracy of the Azure Kinect against the Kinect v2.
• To quantify the temporal and spatial error characteristics of both depth sensing devices.
• To determine the suitability of the Azure Kinect as a replacement for the Kinect v2 in 3D scanning.

Main Methods:

• Comparative analysis of Azure Kinect and Kinect v2 depth and spatial accuracy.
• Utilized laser scanner data as ground truth for sensor measurements.
• Captured planar and flat surfaces at various distances and positions to assess error.
• Analyzed temporal random depth error using histograms and spatial error using a cone of accuracy.

Main Results:

• Azure Kinect demonstrates superior spatial and temporal accuracy compared to Kinect v2 at distances greater than 2.5 meters.
• For distances under 2.5 meters, both sensors exhibit comparable spatial and temporal accuracies.
• Azure Kinect shows enhanced temporal accuracy over Kinect v2 between 2.5m and 3.5m near the optical axis.

Conclusions:

• The Azure Kinect offers improved performance in specific operational ranges compared to its predecessor.
• The Azure Kinect is confirmed as a suitable substitute for the Kinect v2 in 3D scanning tasks.
• Understanding sensor error characteristics is vital for effective 3D reconstruction and application development.