|Abstract:||Time of Flight (ToF) cameras are measurement devices which have been present in the market for over a decade. They enable one to acquire three-dimensional point cloud; providing in addition to usual spatial coordinates, other parameters related to the quality of the data set (amplitude image & confidence value, etc.).
The current paper describes the preliminary studies that were carried out in order to understand if there were any recurring behaviours at each take. In the current literature, much has been said about calibration, errors and denoising, etc., but there is nothing in the literature dealing with the following basic tests:.. the former, concerned an analysis of dispersion of the depth data over one set of 100 frames. All pixels were examined, specifically how the interquartile ranges changed according to the position of the pixel on the sensor and changes in distance. What resulted was a map highlighting the accuracy and precision of each pixel. In the central area we measured an accuracy less than or equal to 1 cm and a very high precision rate, instead at the corners the accuracy was greater than 4 cm but with a low repeatability.
The latter test was focused on checking if the distribution was normal since it has always been assumed to be normal, but without any in depth inquiry into the matter. As demonstrated in the following paragraphs, Quantile Quantile Plot were created to verify if there was any correspondence between standard quantiles and the quantiles of data collected. Moreover, in order to make our thesis more consistent, two hypothesis’ tests were implemented, the Shapiro-Wilk and the Cramer-Von Mises.. Indeed, knowing the distribution, it’s possible to make probabilistic assumptions and to notice if there is a global or local behaviour happening. The result showed that the distribution was normal for all of the pixels of the sensor.
The first contribution of this paper is to fully investigate the behaviour of ToF Cameras in laboratory conditions so as to evaluate a possible application in the geomatics field. The second aim is to deal with the issue from a different point of view, namely the statistical one, by introducing a new and systematic approach that can be applied to every device to be tested.|