Hyper-Spectral Imaging for Overlapping Plastic Flakes Segmentation
This work addresses a practical issue in plastic sorting for recycling applications, but it is incremental as it builds on existing hyper-spectral imaging techniques with a novel encoding approach.
The paper tackles the problem of segmenting overlapping plastic flakes in hyper-spectral imaging by proposing a bitfield encoding method, which improves over a baseline single-label approach and demonstrates potential for predicting multiple labels even when trained only on non-overlapping classes.
Given the hyper-spectral imaging unique potentials in grasping the polymer characteristics of different materials, it is commonly used in sorting procedures. In a practical plastic sorting scenario, multiple plastic flakes may overlap which depending on their characteristics, the overlap can be reflected in their spectral signature. In this work, we use hyper-spectral imaging for the segmentation of three types of plastic flakes and their possible overlapping combinations. We propose an intuitive and simple multi-label encoding approach, bitfield encoding, to account for the overlapping regions. With our experiments, we show that the bitfield encoding improves over the baseline single-label approach and we further demonstrate its potential in predicting multiple labels for overlapping classes even when the model is only trained with non-overlapping classes.