ABSTRACT
Herrera, F.K.; Uhrin, A.V.; Winans, W.R.; Parrish, C.E.; Murphy, P., and Battista, T., 0000. Enhanced detection and classification of beach-stranded macrodebris: A pilot study using polarimetric imagery.
Polarimetric imaging is an emerging remote-sensing technology that is proving useful in detection and characterization of anthropogenic objects within a scene. Polarimetric imagery has also been found to be effective when operating in challenging environments such as low light conditions, or when objects on land are located within a noisy/cluttered background, or there is low contrast. There are potential applications to support marine debris detection on beaches where objects can be stranded on heterogeneous substrates. This study evaluated whether mathematically derived polarimetric bands, when used in combination with red, green, and blue (RGB) spectral bands, could improve the ability to visually observe debris items on a sand shoreline and augment semi-automatic debris classification. Eight-band composite shoreline images were created and evaluated, comprised of RGB bands and five calculated polarimetric bands, using a FLIR Blackfly S RGB-polarimetric imaging camera. Polarimetric bands visually enhanced object characteristics such as edge detail, surface texture, and dimensionality, all of which improved segmentation of debris. Results also showed that the addition of the polarimetric bands improved the separability of debris classes and, when included in a K-nearest neighbors classification, substantially increased overall accuracy and kappa statistics. This study provides strong indication that polarimetric imaging is a useful asset for marine debris programs. As the commercial market for polarimetric imaging cameras continues to grow and sensors become more affordable and operationally mature, it is anticipated that polarimetric imagery will become increasingly useful for marine debris detection, particularly where substrates are unknown or highly complex.
