Jang Yeop Kim

Park JY, et al. J Vet Sci. 2025 Dec;27:e5. https://doi.org/10.4142/jvs.25144

License plate recognition is a computer vision technology that plays a crucial role in intelligent transportation systems and vehicle management. However, in real-world road environments, recognition accuracy significantly decreases due to distortions caused by various viewing angles. In particular, existing systems exhibit severe performance degradation when processing license plate images captured at steep angles. This paper proposes a new approach to solve the license plate recognition problem in such unconstrained environments. To accurately recognize text on distorted license plates, it is crucial to precisely locate the four corners of the plate and correct the distortion. For this purpose, the proposed system incorporates vehicle and license plate detection based on YOLOv8 and integrates a Corner Enhancement Module (CEM) utilizing a Deformable Convolutional Network (DCN) into the model’s neck to ensure robust feature extraction against geometric transformations. Additionally, the system significantly improves corner detection accuracy through parallel ensemble processing of three license plate images: the original and two aspect ratio-adjusted versions (2:1 and 1.5:1). Furthermore, we verified the system’s versatility in real road environments by implementing a real-time license plate recognition system using Raspberry Pi 4 and a camera module.

Image quality plays a critical role in medical image analysis, significantly impacting diagnostic outcomes. Sharp and detailed images are essential for accurate diagnoses, but acquiring high-resolution medical images often demands sophisticated and costly equipment. To address this challenge, this study proposes a convolutional neural network (CNN)-based super-resolution architecture, utilizing a melanoma dataset to enhance image resolution through deep learning techniques. The proposed model incorporates a convolutional self-attention block that combines channel and spatial attention to emphasize important image features. Channel attention uses global average pooling and fully connected layers to enhance high-frequency features within channels. Meanwhile, spatial attention applies a single-channel convolution to emphasize high-frequency features in the spatial domain. By integrating various attention blocks, feature extraction is optimized and further expanded through subpixel convolution to produce high-quality super-resolution images. The model uses L1 loss to generate realistic and smooth outputs, outperforming existing deep learning methods in capturing contours and textures. Evaluations with the ISIC 2020 dataset—containing 33126 training and 10982 test images for skin lesion analysis—showed a 1–2% improvement in peak signal-to-noise ratio (PSNR) compared to very deep super-resolution (VDSR) and enhanced deep super-resolution (EDSR) architectures.

Real wars involve a considerable number of uncertainties when determining firing scheduling. This study proposes a robust optimization model that considers uncertainties in wars. In this model, parameters that are affected by enemy’s behavior and will, i.e., threats from enemy targets and threat time from enemy targets, are assumed as uncertain parameters. The robust optimization model considering these parameters is an intractable model with semi-infinite constraints. Thus, this study proposes an approach to obtain a solution by reformulating this model into a tractable problem; the approach involves developing a robust optimization model using the scenario concept and finding a solution in that model. Here, the combinations that express uncertain parameters are assumed by scenarios. This approach divides problems into master and subproblems to find a robust solution. A genetic algorithm is utilized in the master problem to overcome the complexity of global searches, thereby obtaining a solution within a reasonable time. In the subproblem, the worst scenarios for any solution are searched to find the robust solution even in cases where all scenarios have been expressed. Numerical experiments are conducted to compare robust and nominal solutions for various uncertainty levels to verify the superiority of the robust solution.

The current study aims to establish an optimal Generation System Expansion Plan that can satisfy the increasing electricity demand while maintaining operational elements and the stability of the energy supply. The architecture is composed of plan-level and operation-level models, which are basically based on optimization. In the first step, we estimated future power demand data through time series analysis. In addition, power plant data were defined and verified data were collected. In the next step, the previous Generation System Expansion Plan methodology was used to deduce a feasible solution and construction costs that satisfy the reserve rate. In the third step, mixed integer programming (MIP)-based power generation system operation plan methodology was used to deduce numbers on the operation of power generation system. In addition, power plants with similar characteristics were grouped to reduce the calculation complexity of unit commitment. In the last step, a feasible solution for the duration of the plan (deduced in Stage II) and operations and maintenance cost information were combined to produce the optimal solution that minimizes the total cost. Experiments were conducted to demonstrate the proposed integrated generation system expansion planning architecture for establishing the optimal generation system expansion planning. This study has academic implications for the establishment of optimal power plant expansion plans to meet future increasing power demand while maintaining operational considerations and supply stability. The effectiveness of the proposed methodology is also illustrated through comparison and verification with the National Plan for Electricity Supply and Demand.