| Abstrakt: |
In the field of high-resolution tomography, there is currently a notable increase in the volume of tomographic projections and data produced. Such a context has been demanding new computational approaches to the process of reconstruction and processing of the resulting digital images. This paper presents a new approach to meet such a demand, such as optimizing the set of tomographic projections for the reconstruction process, parallelizing algorithm reconstruction, and processing the data in a distributed manner. In this context, a customized method for the high-resolution tomographic reconstruction of agricultural samples has been validated. Hence, tomographic projections with greater amounts of information based on measurements of the spectral density of the projections can be prioritized, and the reconstructive process parallelization using the known filtered back-projection can be considered (i.e., distributed data flow and the use of the Apache Spark environment). For the operation, such an approach based on the big data environment has been organized, that is considering a cluster installed on the Amazon Web Services platform, whose configuration has been defined after the evaluation of the speedup and efficiency metrics. The developed method proved to be useful for carrying out high-resolution tomography analyses of large quantities of agricultural samples, based on the paradigms of precision agriculture for gains in sustainability and competitiveness of the production process. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink