Toward Robust River Plastic Detection: Combining Lab and Field‐Based Hyperspectral Imagery

Plastic pollution in aquatic ecosystems has increased dramatically the last five decades, with strong impacts on human and life. Recent studies endorse need for innovative approaches to monitor presence, abundance, types of plastic these ecosystems. One approach gaining rapid traction is use multi- hyperspectral cameras. However, most experiments using this were controlled environments, making findings challenging apply natural environments. We present a method linking lab- field-based identification macroplastics data (1,150–1,675 nm). Experiments riverbank-harvested set up laboratory environment, banks Rhine River. Representative pixel selections eleven lab-based images (n = 786,264 pixels) two 40,289 used analyze differences between Next, classifier algorithms such as support vector machines (SVM), spectral angle mapper (SAM) information divergence (SID) applied, because their robustness varying light conditions high accuracies mapping similarities. Our results showed that SAM classifiers are robust separating pixels from background elements. By applying detection images, user plastics 93.6% 8,370 attained. This study provides key fundamental insights field. With paper we aim contribute development future missions detect