Deep learning methods, and especially convolutional neural networks (CNNs), have made a considerable breakthrough in various fields of machine vision, basically by employing large-scale labeled databases. However, deep learning methods applied in finger-vein area are basically implemented on small-scale datasets, which are probably faced with challenges such as overfitting, susceptible to finger position, unstable performance on various datasets and son on. In this study, we present a lightweight and fully convolutional Generative Adversarial Network (GAN) architecture, which is named FCGAN, using preliminary batch normalization, and tightly-constrained loss function for implementing finger-vein image augmentation. In addition, we present a novel scheme FCGAN-CNN for finger-vein classification, which reveals that synthetic finger-vein images using FCGAN are capable of improving the property of CNN for finger-vein image classification. The experiment of sample augmentation shows that the training accuracy using FCGAN-augmented samples could go beyond 99%, which is higher than 96.34% obtained using only classic sample augmentation. Furthermore, the well-trained CNN is further evaluated on a totally different dataset, which indicates that the proposed scheme FCGAN-CNN is capable of improving the ability of CNN to extract deep features. We consider that the proposed method for sample augmentation could be extended to other biometric systems.

Suffering from uneven illumination and variation of finger position, it is still a tough challenge to effectively distinguish the vein networks and non-venous regions in a fingervein image. Methods based on active contour have achieved an excellent result in medical image segmentation, despite facing with several challenges such as vulnerable to the initial contour and prone to local minimum. In this paper, we propose a novel method which is effective for finger-vein image segmentation based on active contour. Since venous and non-venous areas in captured finger-vein images are hard to distinguish, we design a dehazing algorithm and an edge fitting term to improve the segmentation procedure. Moreover, we employ the kernel fuzzy C-means (KFCM) algorithm to conduct the initialization, which is able to solve the problem that the active contour based methods are susceptible to initial contours. The experimental results show that compared with latest methods, the proposed method achieves a better performance in segmenting finger-vein images and is able to improve the recognition accuracy of finger-vein identification system.

In the past, GAN-based face reenactment methods concentrated mostly on transferring the facial expressions and positions of the source. However, the generated results were susceptible to blurring in some minute details of the face, such as teeth and hair, and their backgrounds were also not guaranteed to be consistent with the manipulated images in terms of light and shadow. Because of these issues, the generated results could be distinguishable as fakes. In this paper, we proposed a landmark based method named HR-Net, which can render source facial expressions and postures on any identity and simultaneously generate realistic face details. Firstly, a lightweight landmark identity conversion module (LIC) was designed to address the identity leakage problem, and it represented facial expressions and poses with only 68 2D landmarks. On this basis, a boundary-guided face reenactment module (BFR) was presented to only learn the background of the reference images; thus, the results generated by BFR can be consistent with the reference images’ light and shadow. Moreover, a novel local perceptual loss function was implemented to support the BFR module in generating more realistic details. Extensive experiments demonstrated that our method achieved the state of the art.

Cloud detection for ground-based sky images has attracted much attention in cloud-related fields. In this paper, we proposed a cloud detection algorithm that reduced the sunlight interference in the image. The solar location method was introduced to track the sun in the image used for feature calculation, which was suitable for the case where the camera could not be calibrated. Following this, the adjustable red green difference (ARGD) feature using red and green channels was proposed. The red weight in the feature was determined by the layering region division, which classified the degree of sunlight interference in the image, and the sky state, which discriminated whether there was sunlight interference in the image. Finally, a fixed zero threshold was applied to feature images in order to obtain the cloud detection results. Experimental results showed that the proposed algorithm performs better than the other algorithms and can effectively reduce the sunlight interference.

Ground-based cloud image categorization is crucial for ground-cloud observation and significantly impacts activities like photovoltaic power forecasting. Current techniques fail to effectively integrate novel cloud image texture features with image classification algorithms. Enhancing cloud categorization precision in publicly available datasets and research tailored for photovoltaic power forecasting remains in its early stages. This study introduces a novel Cloud-ConvNeXt classification approach utilizing a comprehensive texture feature map. We generate a detailed texture feature map for significant areas of ground-based cloud images using a gray-level co-occurrence matrix (GLCM) and local binary patterns (LBP). This image serves as input for an enhanced ConvNeXt classification network, producing precise cloud categories based on meteorological standards. Additionally, cloud occlusion feature and cloud factor feature are extracted from ground-based cloud images to improve cloud categories properties. Results show that the Cloud-ConvNeXt technique surpasses other models, achieving an accuracy of 85.40%. We have constructed a Support Vector Machine (SVM)-Informer model to predict the power output of highly efficient photovoltaic (PV) systems in the next 5 minutes. The results demonstrate that our predictions outperform other methods that depend on cloud images acquired from the ground, with a coefficient of determination (R²) of 0.9668. Furthermore, ablation experiments on attributes linked to different cloud categories confirm their efficacy, thus building a strong foundation for precise solar production forecasting.

A substation planning method that accounts for the widespread introduction of distributed generators (DGs) in a low-carbon economy is proposed. With the proliferation of DGs, the capacity that DGs contribute to the distribution network has become increasingly important. The capacity of a DG is expressed as a capacity credit (CC) that can be evaluated according to the principle that the reliability index is unchanged before and after the introduction of the DG. A method that employs a weighted Voronoi diagram is proposed for substation planning considering CC. A low-carbon evaluation objective function is added to the substation planning model to evaluate the contribution of DGs to a low-carbon economy. A case study is analyzed to demonstrate the practicality of the proposed method.