Adaptive Modeling and Compression of Bathymetric Data With Variable Density

Abstract

This article presents an original approach to bathymetric data modeling consisting of three main elements: first, it uses a very high-resolution grid; second, it employs cells decomposition into smaller blocks depending on local roughness; and third, it features lossy data compression with given accuracy. While developing this approach, the applicable International Hydrographic Organization (IHO) standards (describing guidelines to the accuracy of the models) were taken into account. The main scientific contribution is a strategy of adaptive nonuniform decomposition for digital terrain models joint with controlled lossy compression. We present three methods of grid decomposition, based on the depth gradient (MaxMin method), surface slope variation (curvature method), and spectral coefficients analysis (spectrum method). Decomposed blocks are then compressed in such a way that guarantees a required accuracy of the output model. The near-lossless compression employs methods based on the discrete cosine transform, discrete wavelet transform, or Karhunen–Loeve transform. To ensure high accuracy of the model, we also propose a new way of determining the compression error depending on a local roughness of sea bottom. The presented approach was tested on four representative surfaces created from real sea bottom surveys, giving very prospective results. The developed method can be used in sea survey computer systems used for processing and storing bathymetric data, especially in cases of data from many measurements and data covering extensive areas.