Statistical Analysis-Based Error Models for the Microsoft KinectTM Depth Sensor

Abstract

The stochastic error characteristics of the Kinect sensing device are presented for each axis direction. Depth (z) directional error is measured using a flat surface, and horizontal (x) and vertical (y) errors are measured using a novel 3D checkerboard. Results show that the stochastic nature of the Kinect measurement error is affected mostly by the depth at which the object being sensed is located, though radial factors must be considered, as well. Measurement and statistics-based models are presented for the stochastic error in each axis direction, which are based on the location and depth value of empirical data measured for each pixel across the entire field of view. The resulting models are compared against existing Kinect error models, and through these comparisons, the proposed model is shown to be a more sophisticated and precise characterization of the Kinect error distributions.

Optimal 6D Object Pose Estimation with Commodity Depth Sensors

“…The resulting coefficients of the quadratic model for the hypersurface of standard depth errors are shown in Table 3.1, which resulted in a 0.785 R 2 value. Note, these values are a result of a recalibration to a new collection of depth data sets, different from those presented in [23]. In Figure 3.2(C), the hypersurface fits are plotted on the left for the sensor positioned in front of the wall at a depth of 1600 mm, 2400 mm, and 3200 mm.…”

“…This section presents a new set of empirical models for axial and lateral depth error to be utilized with point set registration (PSR) methods, which were presented in a journal article authored by Choo et al at the University of Virginia [23]. These models were formulated in recognition of the nonlinear pincushion distortion that contributes to an increase in depth error with an increase in radial distance from the optical center, as discussed in Section 2.3.…”

“…Through this study, it was noted that the observed error in depth images is due not only to unpredictable IR noise, but also to systematic quantization and mismatched correlation windows from measured sections of a distorted dot pattern. This work resulted in an open access Sensors journal paper, titled 'Statistical Analysis-Based Error Models for the Microsoft Kinect Depth Sensor' [23].…”

Intention-Aware Motion Planning Using Learning Based Human Motion Prediction

“…Estimated human poses from recorded videos or realtime sensor data tend to be inaccurate or imperfect due to occlusions or limited sensor ranges [4]. Furthermore, the whole-body motions and their complex dynamics with many high-DOF makes it difficult to represent them with accurate motion models [14].…”

“…However, the current state of the art in gathering motion data results in many challenges. First of all, there are errors in the data due to the sensors (e.g., point cloud sensors) or poor sampling [4]. Secondly, human motion can be sudden or abrupt and this can result in various uncertainties in terms of accurate representation of the environment.…”