Inexpensive RGB-D Sensors Performance Measurement in Pavement Data Collection—Part II

Abstract

The majority of sensors used for quantifying indices related to a pavement management system are costly; however, low-cost sensors such as RGB-D sensors have been recently deployed. The performance of RGB-D sensors in pavement condition data collection has not received enough attention. By providing a thorough literature review on the specification of RGB-D sensors, this paper aims to assess and compare the performance of two RGB-D sensors in pavement condition data collection. First, by collecting data from asphalt, concrete, and mosaic surfaces in wet and dry conditions, the accuracy and precision of the sensors are evaluated. For this purpose, noise reduction techniques are applied to the central region of interest and the inherent trend of depth frames is corrected by implementing the Singular Value Decomposition algorithm. Second, dimensions of artificial objects above the pavement surface are detected via sensors by using the Lease Square Method (LSM) and edge detection techniques to evaluate RGB-D sensors' ability and performance in measurement of artificial pavement defects of known dimensions. The results of statistical analysis show the difference in error measurement via both sensors in all surfaces in wet and dry conditions. Through conducting extensive field experiments, it is concluded that Kinect one has several advantages over the Kinect Xbox 360 in pavement condition data collection.