An algorithm for identifying organic materials using a short-wave infrared hyperspectral imager is presented. The main achievement of this algorithm is to automatically locate a clear reference within the scene, without any prior knowledge regarding the scene, and use it to extract the spectral signature of the target material. This step is required for sub-millimeter organic films since they absorb only a small fraction of the light. Therefore, the obtained spectral signature is a product of the target material and the clean background spectral signature. Detection rate as high as 90% was achieved for three model compounds in a controlled scene with false alarm rate lower than 1% in our model scene, containing a set of synthetic targets (ST). These findings lay the groundwork for further study of real (and probably more complex compared to the ST) cases and materials, required for the development of a detection system that will be capable of locating trace amounts of polluting chemicals to enable improved prevention, regulation, and mitigation acts.
All Science Journal Classification (ASJC) codes
- Electrical and Electronic Engineering