Potholes are one of the key defects that affect the performance of roads and highway networks. Metrological features of a pothole provide useful metrics for road distress measurement and severity analysis. This paper presents a performance analysis of Kinect as a sensor for pothole imaging and metrology. Depth images of paved surfaces are collected from concrete and asphalt roads using this sensor. Three-dimensional (3D) meshes are generated for a variety of pothole configurations in order to visualise and to calculate their different metrological features. The sensor is benchmarked using a test-rig with pothole-like depressions or artificial potholes of known dimensions to evaluate sensor performance under different real-life imaging conditions, such as through the media of clear, muddy and oily water. Error in measurement due to surface roughness is also studied. Another source of error that is due to the presence of foreign objects such as stones and pebbles in the form of negative depth, is also discussed and compensated. Results show that a mean percentage error of 2.58 and 5.47% in depth and volumetric calculations, respectively.