This paper by Kourosh Khoshelham and Sander Oude Elberink from the University of Twente, Netherlands, focuses on the accuracy and resolution of depth data from the Kinect sensor for indoor mapping applications. The authors provide a detailed analysis of the calibration process and the mathematical model for depth measurement, derived from disparity images. They derive a theoretical random error model and verify it through experiments, showing that the random error in depth measurements increases quadratically with the distance from the sensor, reaching about 4 cm at the maximum range of 5 meters. The quality of the data is also influenced by the low resolution of the depth measurements, which results in larger point spacing at greater distances. The paper concludes with a comparison of the Kinect point cloud with a high-end laser scanner, demonstrating that the Kinect data is generally accurate within 1-3 meters but degrades at larger distances due to noise and low resolution. The authors emphasize the importance of accurate stereo calibration to eliminate distortions and misalignments between color and depth data, and suggest that for mapping applications, data should be acquired within 1-3 meters to ensure the best quality.This paper by Kourosh Khoshelham and Sander Oude Elberink from the University of Twente, Netherlands, focuses on the accuracy and resolution of depth data from the Kinect sensor for indoor mapping applications. The authors provide a detailed analysis of the calibration process and the mathematical model for depth measurement, derived from disparity images. They derive a theoretical random error model and verify it through experiments, showing that the random error in depth measurements increases quadratically with the distance from the sensor, reaching about 4 cm at the maximum range of 5 meters. The quality of the data is also influenced by the low resolution of the depth measurements, which results in larger point spacing at greater distances. The paper concludes with a comparison of the Kinect point cloud with a high-end laser scanner, demonstrating that the Kinect data is generally accurate within 1-3 meters but degrades at larger distances due to noise and low resolution. The authors emphasize the importance of accurate stereo calibration to eliminate distortions and misalignments between color and depth data, and suggest that for mapping applications, data should be acquired within 1-3 meters to ensure the best quality.