Feature Selection for Identifying Optimal Microwave Frequencies to Detect Floating Macroplastic Litter in C and X Bands

Recently, the utilisation of microwave (MW) frequencies in remote sensing has emerged as a promising and complementary technology to optical methods for effectively detecting and monitoring floating plastic litter. Still, there is a scarce number of existing studies evaluating the optimal MW band for radar detection, particularly making use of machine learning (ML). To contribute to this topic, we propose a feature selection (FS) workflow based on the weighted principal component analysis (WPCA) algorithm to study the tabular backscattering response of floating macroplastic clusters (made of plastic bottles, straws, lids, and cylinder foams) in C- and X-bands. Specific backscattering radio measurements (units) of sequential frequency points within the MW subbands (features) were carried out in a controlled indoor scenario that mimics deep sea conditions. The experimental results show that, under the tested conditions, the X-band frequencies are more relevant in the presence of floating macroplastic.