Neural volumetric reconstruction for coherent synthetic aperture sonar

A computer-implemented method comprising:

• obtaining measurements of an underwater object using high-resolution Synthetic aperture sonar (SAS);
• applying an iterative deconvolution optimization process to the measurements to generate pulse deconvolved measurements;
• performing an analysis-by-synthesis reconstruction using an implicit neural representation to predict complex-valued scatterers from the pulse deconvolved measurements;
• generating synthesized complex measurements from the complex-valued scatterers using a differentiable forward model;
• iteratively updating weights of the differentiable forward model with a computed minimized loss between the synthesized complex measurements and the complex-valued scatterers;
• generating as output from the differentiable forward model, a reconstruction of the underwater object;
• and
• returning the output to a computing device.

Non-transitory computer-readable storage media comprising instructions that, when executed, configure processing circuitry to:

• obtain measurements of an underwater object using high-resolution Synthetic aperture sonar (SAS);
• apply an iterative deconvolution optimization process to the measurements to generate pulse deconvolved measurements;
• perform an analysis-by-synthesis reconstruction using an implicit neural representation to predict complex-valued scatterers from the pulse deconvolved measurements;
• generate synthesized complex measurements from the complex-valued scatterers using a differentiable forward model;
• iteratively update weights of the differentiable forward model with a computed minimized loss between the synthesized complex measurements and the complex-valued scatterers;
• generate as output from the differentiable forward model, a reconstruction of the underwater object;
• and
• return the output to a computing device.