Real-time structure-light-based 3D terrain sensing for mobile robot using CUDA

3D surface inspection has been widely used in the industry to detect all kinds of surface defects and to measure the overall quality of a produced piece. It gives us the idea of using this technique on mobile robots for terrain sensing. 3D surface inspection provides reliable and accurate measurements, but offline. So the challenge of using surface inspection on mobile robots is how to implement it for real-time measurement, for mobile robots need to get information from the environment online and make responses immediately. For this purpose, a real-time terrain sensing system for mobile robots based on structure light is proposed. To satisfy the requirement of online measurement, a one-shot pattern using monochromatic light is used in this system, and the parallel implementation of the required algorithms is made using CUDA to accelerate the processing. The one-shot pattern is based on the pattern primitive which is similar to the checkerboard corner, and it provides a high accuracy performance. The use Graphics Processing Units (GPUs) for data processing tasks facilitates the construction of cost-effective and real-time systems. Such a system has been developed in our laboratory, and we start to use it on a 6 wheel robot. Last, the experimental results demonstrate the performance of this system.