Plastic waste is one of the main factors causing environmental pollution and affecting biodiversity, and identification and detecting plastic waste is the premise of removal and treatment. Unmanned aerial vehicle (UAV) is gradually applied to identify and classify plastic waste because of its advantages of simplicity, convenience, and safe operation, but the current visual interpretation method is inefficient and cumbersome. To support the detection of plastic waste, researchers have developed various automatic and semi-automatic algorithms. Among these algorithms, deep learning technology has outstanding performance in river garbage detection, but there are also practical problems such as small floating garbage volume, sparse samples, complex garbage environment, In this paper, a classification plus target detection (C+D) model is proposed, and a lightweight floating plastic waste detection model based on deep learning is constructed. The EfficientNet classification algorithm and Yolov5 target detection algorithm are combined and improved for experimental verification, and various floating plastic wastes are automatically identified and located. In this paper, the UAV image data set obtained from the flight in Longhe River Basin, Langfang City, Hebei Province, China, is used to investigate the plastic floating garbage. The algorithm verification experiment shows that the detection accuracy of the three kinds of plastic garbage is higher than 85% (AP: plastic bag: 0.95; Plastic foam: 0.90; Plastic bottle: 0.87), which shows its excellent floating plastic recognition ability. The FPS of UAV equipment can reach 40.23 on edge, which shows that its recognition speed is fast and meets the real-time demand.
This study pioneers the integration of Sentinel-2 satellite imagery with Kolmogorov–Arnold networks (KAN) for the evaluation of chlorophyll-a (Chl-a) concentrations in inland lakes. Using Baiyangdian Lake in Hebei Province, China, as a case study, a specialized KAN architecture was designed to extract spectral features from Sentinel-2 data, and a robust algorithm was developed for Chl-a estimation. The results demonstrate that the KAN model outperformed traditional feature-engineering-based machine learning (ML) methods and standard multilayer perceptron (MLP) deep learning approaches, achieving an R2 of 0.8451, with MAE and RMSE as low as 1.1920 μg/L and 1.6705 μg/L, respectively. Furthermore, attribution analysis was conducted to quantify the importance of individual features, highlighting the pivotal role of bands B3 and B5 in Chl-a retrieval. Furthermore, spatio-temporal distributions of Chl-a concentrations in Baiyangdian Lake from 2020 to 2024 were generated leveraging the KAN model, further elucidating the underlying causes of water quality changes and examining the driving factors. Compared to previous studies, the proposed approach leverages the high spatial resolution of Sentinel-2 imagery and the accuracy and interpretability of the KAN model, offering a novel framework for monitoring water quality parameters in inland lakes. These findings may guide similar research endeavors and provide valuable decision-making support for environmental agencies.