Image optimization for chemical species tomography with an irregular and sparse beam array

High-speed tomographic imaging of hostile engineering processes using absorption-based measurements presents a number of difficulties. In some cases, these challenges include severe limitations on the number of available measurement paths through the subject and the process of designing the geometrical arrangement of these paths for best imaging performance. This paper considers the case of a chemical species tomography system based on near-IR spectroscopic absorption measurements, intended for application to one cylinder of a multi-cylinder production engine. Some of the results, however, are also applicable to other hard-field tomographic modalities in applications where similar constraints may be encountered. A hitherto unreported design criterion is presented for optimal beam geometry for imaging performance, resulting in an irregular array with only 27 measurement paths through the subject for the engine application. Image reconstruction for this severely limited geometry is considered at length, using both simulated and experimental phantom data. Novel methods are presented for the practical generation of gaseous phantoms for calibration and testing of the system. The propane absorption coefficient at 1700 nm is measured. Quantitative imaging of propane plumes in air is demonstrated, showing good localization of circular plumes with diameter as small as 1/5 of the subject diameter and excellent imaging of multiple plumes.