Image dehazing technology is a crucial component in the fields of intelligent transportation and autonomous driving. However, most existing dehazing algorithms only process images in the spatial domain, failing to fully exploit the rich information in the frequency domain, which leads to residual haze in the images. To address this issue, we propose a novel Frequency-domain Information Guided Symmetric Dual-branch Dehazing Network (FIGD-Net), which utilizes the spatial branch to extract local haze features and the frequency branch to capture the global haze distribution, thereby guiding the feature learning process in the spatial branch. The FIGD-Net mainly consists of three key modules: the Frequency Detail Extraction Module (FDEM), the Dual-Domain Multi-scale Feature Extraction Module (DMFEM), and the Dual-Domain Guidance Module (DGM). First, the FDEM employs the Discrete Cosine Transform (DCT) to convert the spatial domain into the frequency domain. It then selectively extracts high-frequency and low-frequency features based on predefined proportions. The high-frequency features, which contain haze-related information, are correlated with the overall characteristics of the low-frequency features to enhance the representation of haze attributes. Next, the DMFEM utilizes stacked residual blocks and gradient feature flows to capture local detail features. Specifically, frequency-guided weights are applied to adjust the focus of feature channels, thereby improving the module’s ability to capture multi-scale features and distinguish haze features. Finally, the DGM adjusts channel weights guided by frequency information. This smooths out redundant signals and enables cross-branch information exchange, which helps to restore the original image colors. Extensive experiments demonstrate that the proposed FIGD-Net achieves superior dehazing performance on multiple synthetic and real-world datasets.

River water quality prediction is a crucial aspect of water environment management and ecological conservation, holding significant importance for ensuring the sustainable utilization of water resources. As a key indicator for assessing river self-purification capacity and aquatic ecosystem health, the accurate prediction of dissolved oxygen (DO) is particularly vital for water quality early warning. To address the challenges that single deep learning models face in collaboratively modeling long- and short-term dependencies, and that most hybrid methods fail to adequately consider the characteristic differences in various components within a time series, this paper proposes an STL-TCN-LSTM model for predicting DO concentration in river water. The proposed model first employs seasonal-trend decomposition using Loess (STL) to decompose the original time series into three components: trend, seasonality, and residual, aiming to separate features at different time scales. Then, three parallel Temporal Convolutional Networks (TCNs) are utilized to extract temporal features from each component and reconstruct the sequence. Finally, the reconstructed results are fed into a Long Short-Term Memory (LSTM) network to further model their dynamic temporal dependencies, thereby enhancing prediction accuracy. The performance of the proposed model is validated on three river water quality datasets from different river basins with varying sampling frequencies. The experimental results on the three river datasets show that the STL-TCN-LSTM model consistently outperforms all baseline models, including LSTM, TCN, BiLSTM, GRU, CNN-LSTM, and XGBoost. Specifically, the Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) are reduced by an average of 14.47%, 14.51%, and 14.27%, respectively, while the coefficient of determination (R²) improves by an average of 0.79%. The Wilcoxon signed-rank test confirms that all performance improvements are statistically significant (p < 0.05). These results demonstrate that the proposed model achieves higher prediction accuracy and exhibits stronger generalization capability in DO forecasting, thereby offering a reliable tool for water quality early warning and aquatic environmental management.

Infrared images are often degraded by complex noise due to hardware and environmental factors, posing challenges for subsequent processing and target detection. To overcome the shortcomings of existing denoising methods in balancing noise removal and detail preservation, this paper proposes a Wavelet Transform Enhanced Infrared Denoising Model (WTEIDM). Firstly, a Wavelet Transform Self-Attention (WTSA) is designed, which combines the frequency-domain decomposition ability of the discrete wavelet transform (DWT) with the dynamic weighting mechanism of self-attention to achieve effective separation of noise and detail. Secondly, a Multi-Scale Gated Linear Unit (MSGLU) is devised to improve the ability to capture detail information and dynamically control features through dual-branch multi-scale depth-wise convolution and gating strategy. Finally, a Parallel Hybrid Attention Module (PHAM) is proposed to enhance cross-dimensional feature fusion effect through the parallel cross-interaction of spatial and channel attention. Extensive experiments are conducted on five infrared datasets under different noise levels (σ = 15, 25, and 50). The results demonstrate that the proposed WTEIDM outperforms several state-of-the-art denoising algorithms on both PSNR and SSIM metrics, confirming its superior generalization capability and robustness.

In recent years, face detection has been widely applied in various intelligent monitoring systems. However, missed detections and low detection accuracy present challenges, such as small, blurred, and occluded faces in multi-face detection scenarios. To address these challenges, an adaptive multi-region fusion network is designed for dense face detection. First, in the shallow layers of the network, a multi-scale cross-stage fusion (MC4f) module is designed to replace the C3 module, which solves the issue of gradient explosion or disappearance in deep networks and promotes the effective convergence of the network on small datasets. An adaptive fusion explicit spatial vision centre (AESVC) is then designed between the backbone and neck networks to adaptively fuse local and global features to refine face information and enhance feature representation capabilities in complex tasks. Subsequently, a multi-scale parallel attention mechanism (MSPAM) is proposed to enhance the cross-scale fusion of facial features and reduce the loss of shallow features. Finally, to achieve accurate facial key point localisation and alignment, wing loss and A-loss functions are integrated, which balances the relationship between easy and difficult samples. Compared with the original model, the proposed model increases the mean average precision (mAP) by 1.75, 2.01, and 3.06% for easy, medium, and hard samples, respectively. The experimental results prove the effectiveness of the adaptive multi-region fusion network for dense face detection.