Marine litter weight estimation from UAV imagery: Three potential methodologies to advance macrolitter reports.

In the context of marine litter monitoring, reporting the weight of beached litter can contribute to a better understanding of pollution sources and support clean-up activities. However, the litter scaling task requires considerable effort and specific equipment. This experimental study proposes and evaluates three methods to estimate beached litter weight from aerial images, employing different levels of litter categorization. The most promising approach (accuracy of 80 %) combined the outcomes of manual image screening with a generalized litter mean weight (14 g) derived from studies in the literature. Although the other two methods returned values of the same magnitude as the ground-truth, they were found less feasible for the aim. This study represents the first attempt to assess marine litter weight using remote sensing technology. Considering the exploratory nature of this study, further research is needed to enhance the reliability and robustness of the methods.

Drones for litter monitoring on coasts and rivers: suitable flight altitude and image resolution.

Multirotor drones can be efficiently used to monitor macro-litter in coastal and riverine environments. Litter on beaches, dunes and riverbanks, along with floating litter on coastal and river waters, can be spotted and mapped from aerial drone images. Items detection and classification are prone to image resolution, which is expressed in terms of Ground Sampling Distance (GSD). The GSD is determined by drone flight altitude and camera properties. This paper investigates what is a suitable GSD value for litter survey. Drone flight altitude and camera setup should be chosen to obtain a GSD between 0.5 cm/px and 1.25 cm/px. Within this range, the lowest GSD allows litter categorization and classification, whereas the highest value should be adopted for a coarser litter census. In the vision of drawing up a global protocol for drone-based litter surveys, this work sets the ground for homogenizing data collection and litter assessments.

BePLi Dataset v1: Beach Plastic Litter Dataset version 1 for instance segmentation of beach plastic litter.

Marine plastic pollution is a pressing global issue nowadays. To address this problem, automated image analysis techniques that can identify plastic litter are necessary for scientific research and coastal management purposes. The Beach Plastic Litter Dataset version 1 (BePLi Dataset v1) comprises 3709 original images taken in various coastal environments, along with instance-based and pixel-level annotations for all plastic litter objects visible in the images. The annotations were compiled in the Microsoft Common Objects in Context (MS COCO) format, which was partially modified from the original format. The dataset enables the development of machine-learning models for instance-level and/or pixel-wise identification of beach plastic litter. All original images in the dataset were extracted from beach litter monitoring records operated by the local government of Yamagata Prefecture in Japan. Litter images were taken in different backgrounds, such as sand beaches, rocky beaches, and tetrapods. The annotations for instance segmentation of beach plastic litter were made manually, and were given for all plastics objects, including PET bottles, containers, fishing gear, and styrene foams,all of which were categorized in a single class "plastic litter". Technologies developed using this dataset have the potential to enable further scalability for the estimation of plastic litter volume. This would help researchers, including individuals, and the the government to monitor or analyze beach litter and the corresponding pollution levels.

Super-Resolution and Feature Extraction for Ocean Bathymetric Maps Using Sparse Coding.

The comprehensive production of detailed bathymetric maps is important for disaster prevention, resource exploration, safe navigation, marine salvage, and monitoring of marine organisms. However, owing to observation difficulties, the amount of data on the world's seabed topography is scarce. Therefore, it is essential to develop methods that effectively use the limited data. In this study, based on dictionary learning and sparse coding, we modified the super-resolution technique and applied it to seafloor topographical maps. Improving on the conventional method, before dictionary learning, we performed pre-processing to separate the teacher image into a low-frequency component that has a general structure and a high-frequency component that captures the detailed topographical features. We learn the topographical features by training the dictionary. As a result, the root-mean-square error (RMSE) was reduced by 30% compared with bicubic interpolation and accuracy was improved, especially in the rugged part of the terrain. The proposed method, which learns a dictionary to capture topographical features and reconstructs them using a dictionary, produces super-resolution with high interpretability.

The BeachLitter dataset for image segmentation of beach litter.

This dataset consists of 3500 images of beach litter and 3500 corresponding pixel-wise labelled images. Although performing such pixel-by-pixel semantic masking is expensive, it allows us to build machine-learning models that can perform more sophisticated automated visual processing. We believe this dataset may be of significance to the scientific communities concerned with marine pollution and computer vision, as this dataset can be used for benchmarking in the tasks involving the evaluation of marine pollution with various machine learning models. The beach litter images were obtained from coastal environment surveys conducted between 2011 and 2019 by the Yamagata Prefectural Government, Japan. These images were originally obtained owing to the reporting guidelines concerning regular coastal-environmental-cleanup and beach-litter-monitoring surveys. Based on these images, the Japan Agency for Marine-Earth Science and Technology created 3500 images comprising eight classes of semantic masks for beach litter detection [1].

Pixel-level image classification for detecting beach litter using a deep learning approach.

Mitigating and preventing beach litter from entering the ocean is urgently required. Monitoring beach litter solely through human effort is cumbersome, with respect to both time and cost. To address this problem, an artificial intelligence technique that can automatically identify different-sized beach litter is proposed. The technique was established by training a deep learning model that enables pixel-wise classification (semantic segmentation) using beach images taken by an observer on the beach. Eight segmentation classes that include two beach litter classes were defined, and the results were qualitatively and quantitatively verified. Segmentation performance was adequately high based on three metrics: Intersection over Union (IoU), precision, and recall, although there is room for further improvement. The potency of the method was demonstrated when it was applied to images taken in different places from training data images, and the coverage of artificial litter calculated and discussed using drone images provided ground truth.