Inversion of Water Quality Parameters from UAV Hyperspectral Data Based on Intelligent Algorithm Optimized Backpropagation Neural Networks of a Small Rural River

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(1), P. 119 - 119

Published: Jan. 2, 2025

The continuous and effective monitoring of the water quality small rural rivers is crucial for sustainable development. In this work, machine learning models were established to predict a typical river based on quantity measured data UAV hyperspectral images. Firstly, spectral preprocessed using fractional order derivation (FOD), standard normal variate (SNV), normalization (Norm) enhance response characteristics parameters. Second, method combining Pearson’s correlation coefficient variance inflation factor (PCC–VIF) was utilized decrease dimensionality features improve input data. Again, screened features, back-propagation neural network (BPNN) model optimized mixture genetic algorithm (GA) particle swarm optimization (PSO) as means estimating parameter concentrations. To intuitively evaluate performance hybrid algorithm, its prediction accuracy compared with that conventional algorithms (Random Forest, CatBoost, XGBoost, BPNN, GA–BPNN PSO–BPNN). results show GA–PSO–BPNN turbidity (TUB), ammonia nitrogen (NH3-N), total (TN), phosphorus (TP) exhibited optimal coefficients determination (R2) 0.770, 0.804, 0.754, 0.808, respectively. Meanwhile, also demonstrated good robustness generalization ability from different periods. addition, we used visualize parameters in study area. This work provides new approach refined rivers.

Language: Английский

Smart Monitoring Method for Land-Based Sources of Marine Outfalls Based on an Improved YOLOv8 Model

Water, Journal Year: 2024, Volume and Issue: 16(22), P. 3285 - 3285

Published: Nov. 15, 2024

Land-based sources of marine outfalls are a major source pollution. The monitoring land-based is an important means for environmental protection and governance. Traditional on-site manual methods inefficient, expensive, constrained by geographic conditions. Satellite remote sensing spectral analysis can only identify pollutant plumes affected discharge timing cloud/fog interference. Therefore, we propose smart method based on improved YOLOv8 model, using unmanned aerial vehicles (UAVs). This accurately classify outfalls, offering high practical application value. Inspired the sparse sampling in compressed sensing, incorporated multi-scale dilated attention mechanism into model integrated dynamic snake convolutions C2f module. approach enhanced model’s detection capability occluded complex-feature targets while constraining increase computational load. Additionally, proposed new loss calculation combining Inner-IoU (Intersection over Union) MPDIoU (IoU with Minimum Points Distance), which further regression speed its ability to predict targets. final experimental results show that achieved mAP50 (mean Average Precision at 50) 87.0%, representing 3.4% from original effectively enabling outlets.

Language: Английский