high performance computing on graphics processing units: hgpu.org

This paper describes an implementation of a sonar image simulator optimized for running on a computer’s Graphics Processing Unit (GPU). GPUs are hardware-optimized to obtain maximum performance on computer graphics applications. Because these applications generally simulate focal plane images (i.e. optical systems, video, etc), some specific adaptations are required to render range images such as those generated by a sonar sensor. Considerations for the simulation of side-scan and synthetic aperture images are discussed, including a thorough explanation of the imaging geometry. The use of the GPU to implement the sonar simulation process is described in detail, including the different render stages required to construct the final image and the addition of more refined effects such as noise, multi-path and the sensor’s point spread function (PSF). Other advanced uses of the GPU simulator core are discussed, such as Fast Fourier Transform (FFT) computation and fast image correlation. Finally, results and performance figures are presented.