Machine Vision System for Autonomous Asphalt Layer

Asphalt laying is a labor-intensive process and creates extensive air pollution. Next generation environment friendly machines need to minimize the emissions and necessary manual labor. Autonomously operating machines are expected to replace the current ones. These machines will need to measure the road profile to perform their operation. Laser based road profile measurement sensors are expensive and require a clean environment to perform their single point or line inspection with a fraction of micron accuracy. Recently, stereoscopic imaging has been selected by the automobile manufacturers for autonomous driving. In this study performance of the stereoscopic imaging technique of Intel RealSense D455 camera was evaluated for estimation of the surface profile in the narrow strip at the back of the asphalt laying machine. Laboratory and field tests were performed in the study. At the laboratory tests, four different heights were created on the floor by using boxes with different heights. RealSense D455 was used to take pictures. The height distribution of the images along a line on the floor was estimated by the camera after single shutter action. The manufacturer’s 2% accuracy at 4 m distance was verified in the study. At the field test, the profile of an actual pothole was evaluated, and similar performance was observed. The study indicated that stereoscopic imaging methods could be used to estimate the road profile if the expected accuracy is around plus minus 1 cm when the strip length is about 50 cm or less. The main advantage is getting the data from single shutter action in a very short time without any need for point to point or line scanning.