Occlusion Modeling for Coherent Echo Data Simulation: A Comparison Between Ray-Tracing and Convex-Hull Occlusion Methods

The ability to simulate realistic coherent data setsfor synthetic aperture imaging systems is crucial for the design,development, and evaluation of sensors and their signal processingpipelines, machine learning algorithms, and autonomy systems. Inthe case of synthetic aperture sonar (SAS), collecting experimentaldata is expensive, and it is rarely possible to obtain ground truthof the sensor's path, the speed of sound in the medium, and thegeometry of the imaged scene. Simulating sonar echo data allowssignal processing algorithms to be tested with known ground truth,enabling rapid and inexpensive development and evaluation ofsignal processing algorithms. The de facto standard for simulatingconventional high-frequency (i.e.,>100 kHz) SAS echo data froman arbitrary sensor, path, and scene is to use a point- or facet-baseddiffractionmodel.Acrucialpartofthis process is acousticocclusionmodeling.ThisarticledescribesaSASsimulationpipelineandcom-pares implementations of two occlusion methods: 1) a ray-tracingmethod and 2) a newer approximate method based on finding theconvex hull of a transformed point cloud. The full capability ofthe simulation pipeline is demonstrated using an example scenebased on a high-resolution 3-D model of the SS Thistlegorm ship-wreck, which was obtained using photogrammetry. The 3-D modelspans a volume of 220x130x25 m and is comprised of over30 million facets that are decomposed into a cloud of almost 1billion points. The convex-hull occlusion model was found to resultin simulated SAS imagery that is qualitatively indistinguishablefrom the ray-tracing approach and quantitatively very similar,demonstrating that the use of this alternative method has potentialto improve speed while retaining high fidelity of simulation. Theconvex-hull approach was found to be up to four times faster ina fair speed comparison with serial and parallel central process-ing unit (CPU) implementations for both the methods, with thelargest performance increase for wide-beam systems. The fastestocclusion modeling algorithm was found to be graphics processingunit (GPU)-accelerated ray tracing over the majority of scene scalestested, which was found to be up to two times faster than the parallelCPU convex-hull implementation. Although GPU implementationsof convex-hull algorithms are not currently readily available, thefuture development of GPU-accelerated convex-hull finding couldmake the new approach much more viable. However, in the mean-time, ray tracing is still preferable, since it has higher accuracy andcan leverage the existing implementations for high-performancecomputing architectures for better performance