Digital holography (DH) has been demonstrated as a very powerful tool for micro-plastics (MPs) imaging and recognition, thanks to its unique capabilities such as label-free 3D imaging, flexible focusing and high-throughput. Moreover, the use of machine learning approaches has permitted to surpass main processing limitations in classifying MPs. In particular, the quantitative phase signature provided by DH permits to identify the unique fingerprint for MPs that is crucial to improve the accuracy in features based classification task. In this paper, we investigate new optical, morphological and texture features that can be calculated from phase images of MPs only.
|