Earticle

Stereo vision is actively researched as a solution for distance measurement in autonomous driving. This technique involves triangulation-based distance calculation using left and right images acquired from two image sensors. A kernel window is employed to determine the disparity between these left and right images. To achieve optimal disparity in various environments, it is essential to support different kernel sizes. Moreover, there is a need for research to integrate high-throughput memory, such as high bandwidth memory (HBM), for processing real-time highresolution stereo vision images from sensors. In this paper, we propose a stereo vision accelerator structure utilizing HBM, which supports various kernel sizes.

Image classification is an important task in computer vision. The methods based on spatial information generally employ some low-level features for image classification, such as gray scale, color, texture and location. It is difficult for vision system to understand and the single feature is too limited to obtain correct classification results. In this paper, an algorithm based on multi-kernel feature learning is proposed and used for image classification. First, the kernel function is used to produce a kernel descriptor, which aggregates the pixel attributes into patch-level features; Then, through the multi-kernel learning, these descriptors are further aggregated to obtain hierarchical multi-feature descriptors; Finally, the label of each image is given by the fusion strategy of on multi-classifiers, which effectively utilizes the advantages of multi-kernel learning and takes the complementary among the classifiers into account. The experimental results show that the proposed method is efficient in promoting the classification results.

Because of the feature points can describe the local characteristics of the image in a reasonable manner, effective use of feature point of content based image retrieval become the current hot issues in the field of computer vision. Aiming at this problem, we put forward a kind of combination clustering based on feature points, a new method of image retrieval. The method includes the combination of feature point clustering algorithm and based on the algorithm of local color histogram construction strategy. With the existing and local color histogram retrieval method based on feature points, compared to the method can effectively solve the current method of feature point location information and feature point center relying too much on the problem. Subjectivity and as a result of the manual annotation image accuracy, the traditional image retrieval methods cannot meet the needs of the user. Multidimensional indexing technology is only from the perspective of how to improve the indexing algorithm to adapt to the large-scale database to consider a problem, in content-based image retrieval. Our research combines the advantages of the semantic analysis and kernel clustering which will enhance the performance of the traditional image retrieval methods and strengthen the feasibility of the algorithm.

The goal of multi-output regression is to predict multiple realvalued output variables. Research in this area is lacking compared to multi-output classification . In this approach multi-output kernel regression based on kernel regression in which different outputs are assumed to be correlated. We also introduce a modelselection method employs generalized crossvalidation function for choosing optimal values of hyperparameters. Numerical results from synthetic and real datasets are then obtained to illustrate that the proposed outperforms the other methods on multi-output regression problems.

Conventional time series modeling may not satisfy the model validity for short-period time series data. In this study, we apply the Kernel Variant Multi-Way Principal Component Analysis (KMPCA) to cluster multivariate time series data which havemultiple dimensions with auto- and cross-correlations. We then check whether this method works well in clustering those data by employing simulation for generalization. Two simulation studies with two different mean structures with nine combinations of auto- and cross-correlations were conducted. The results showed that KMPCA cluster two different mean structure groups over 90% success rates with an appropriate kernel function. We also found that when the mean structures are the same, auto-correlation, the number of temporal points, and the kernel function parameter have the statistically significant effects on clustering performance. The second and third order interaction effects with each of those factors also have effects on clustering success rates. Among the effects of the main factors, the kernel function parameter is the most critical factor to consider for obtaining better performance. A similar error structure may obstruct the clustering performance: strong cross-correlation, weak auto-correlation, and a larger number of temporal points. The paper also discussed some limitations of the KMPCA model and suggested directions for future research that could improve the model.

In this paper, we construct a complex reproducing kernel space for singular multi-point BVPs, and skillfully obtain reproducing kernel expressions. Then, we transform the problem into an equivalent operator equation, and give a numerical algorithm to provide the approximate solution. The uniform convergence of this algorithm is proved, and complexity analysis is done. Lastly, we show the validity and feasibility of the numerical algorithm by two numerical examples.

태양광 발전 시스템의 효율적인 운영을 위해서는 발전 이상을 조기에 탐지하는 것이 필수적이다. 그러나 기존 이상 탐지 모델들은 특정 데이터셋에 최적화되어 있어 다양한 발전 환경에 범용적으로 적용하기 어렵다는 한계가 있다. 본 연구에서는 다양한 태양광 발전 데이터에 범용적으로 적용 가능한 이상 탐지 프레임워크인 MKEAD(Multi-Kernel Ensemble Adaptive Detector)를 제안한다. 제안하는 프레임워크는 다중 커널 One-Class SVM, Isolation Forest, Local Outlier Factor, Autoencoder를 결합한 앙상블 구조와 데이터 특성에 따라 최적의 탐지 컴포넌트를 자동으로 선택하는 적응형 모델 선택 메커니즘을 특징으로 한다. 4개의 데이터셋에서 모델 성능을 검증한 결과, 제안 방법은 모든 데이터셋에서 1위를 달성하였으며 기존 최고 성능 대비 평균 11.0%, 최대 27.9%의 F1 Score 향상을 보였고, 임계값 독립적 지표인 AUC-ROC에서도 0.888~0.999로 모든 데이터셋에서 최고 성능을 달성하였다. 또한, 이상 탐지 모델을 시간 시프트 기법을 통해 이상 예측 모델로 확장하여 1일~7일 후의 이상 발생을 예측하는 실험을 수행하였으며, OPSD-Europe 데이터셋에서 7일 후 예측에서도 F1 0.787을 유지하여 전조 패턴 학습 능력을 입증하였다. 탐지된 이상 사례를 분석한 결과, 패널 효율 저하 및 인버터 이상 등 주요 결함을 성공적으로 탐지하였으며, 유지보수 활동을 적시에 수행하여 성능 회복에 유용할 것임을 확인하였다.

Early detection of generation anomalies is essential for efficient operation of solar photovoltaic (PV) systems. However, existing anomaly detection models are optimized for specific datasets, limiting their universal applicability across diverse generation environments. In this study, we propose MKEAD (Multi-Kernel Ensemble Adaptive Detector), a universally applicable anomaly detection framework for various solar PV generation data. The proposed framework features an ensemble structure combining multi-kernel One-Class SVM, Isolation Forest, Local Outlier Factor, and Autoencoder, along with an adaptive model selection mechanism that automatically selects the optimal detection component based on data characteristics. Validation on four datasets demonstrated that the proposed method achieved first place in all datasets with an average improvement of 11.0% and a maximum improvement of 27.9% in F1 Score compared to the best existing methods, and also achieved the highest AUC-ROC (0.888--0.999) across all datasets, confirming threshold-independent superiority. Furthermore, we extended the anomaly detection model to an anomaly prediction model using a time-shift technique and conducted experiments to predict anomaly occurrences 1 to 7 days ahead, achieving an F1 score of 0.787 for 7-day prediction on the OPSD-Europe dataset, demonstrating the model's ability to learn precursor patterns. Analysis of detected anomaly cases successfully identified major defects such as panel efficiency degradation and inverter anomalies, confirming that timely maintenance activities would be useful for performance recovery.

최근, 감시시스템, 게임, 영화등 다양한 분야에서 영상을 이용한 실시간 객체 추적의 필요성이 높아짐에 따라, 커널기반 mean-shift 추적 기법에 대한 관심이 높아지고 있다. 커널 기반 mean-shift 객체 추적에 있어서 주요한 몇 가지 문제점들 중, 첫번째로 추적 목표 객체에 대한 부분 가림 흑은 전체 가림 상황에서의 객체 추적의 문제를 들 수 있다. 본 논문에서는 멀티모드 지역적 커널 가중치를 적용함드로써 부분 가림 상황에서도 안정적드로 객체를 추적할 수 있는 실시간 mean-shift 추적 기법을 제안한다. 제안기법에서는 단일 커널을 사용하는 대신 여러 개의 서브 커널들로 구성된 커널을 사용하고, 각 서브 커널의 위치에 따른 지역적 커널 가중치를 적용한다. 기존의 멀티모드 커널 기반의 방법과 비교한 실힘을 통하여 본 제안 방법이 보다 안정적드로 객체를 추적할 수 있음을 보였다.

As the needs of real-time visual object tracking are increasing in various kinds of application fields such as surveillance, entertainment, etc., kernel-based mean-shift tracking has received more interests. One of major issues in kernel-based mean-shift tracking is to be robust under partial or full occlusion status. This paper presents a real-time mean-shift tracking which is robust in partial occlusion by applying multi-mode local kernel weight. In the proposed method, a kernel is divided into multiple sub-kernels and each sub-kernel has a kernel weight to be determined according to the location of the sub-kernel. The experimental results show that the proposed method is more stable than the previous methods with multi-mode kernels in partial occlusion circumstance.

The multi-core technology has become pervasive in embedded systems. An implementation of the Doppler Beam Sharpening algorithm that improves the azimuth resolution by using doppler frequency shift is possible only in multi-core environment because of the amount of calculation. In this paper, we design of multi-core architecture for a real time implementation of DBS algorithm. And based on designed structure, we produce a DBS image on P4080 board.

Emotion Recognition is one of the important part to develop in human-human and human computer interaction. In this paper, we have focused on the performance of multi-class SVM (Support Vector Machine) with Gaussian RFB (Radial Basis function) kernel, which has been used to solve the problem of emotion recognition from physiological signals and to improve the accuracy of emotion recognition. The experimental paradigm for data acquisition, visual-stimuli of IAPS (International Affective Picture System) are used to induce emotional states, such as fear, disgust, joy, and neutral for each subject. The raw signals of acquisited data are splitted in the trial from each session to pre-process the data. The mean value and standard deviation are employed to extract the data for feature extraction and preparing in the next step of classification. The experimental results are proving that the proposed approach of multi-class SVM with Gaussian RBF kernel with OVO (One-Versus-One) method provided the successful performance, accuracies of classification, which has been performed over these four emotions.

In this paper, we address a novel system-level power management technique which is specifically targeted for an ARM 11 MPCore system. Our proposed solution is a DPM technique which includes process monitoring, real time power analysis, and policy application to reduce the power consumption while meeting the performance requirement. One of the main contributions of this paper is that we systematically infer QoS requirements of processes without getting any additional information from the application. When multiple processes are running under various user level policies, priorities of the policy application are determined in such a way that the overall system performance is maintained while power consumption is effectively managed. Experimental results show that our DPM technique is very effective in reducing power consumption without violating system's QoS requirements.

This paper presents a semi-automatic methodology to synthesize executable digital controller saftware in a multi-loop control system. A digital controller is described by a task graph and end-to-end timing requirements. A task graph denotes the software structure of the controller, and the end-to-end requirements establish timing relationships between external inputs and outputs. Our approach translates the end-to-end requirements into a set of task attributes such as task periods and deadlines using nonlinear optimization techniques. Such attributes are essential for control engineers to implement control programs and schedule them in a control system with limited resources. In current engineering practice, human programmers manually derive those attributes in an ad hoc manner: they often resort to radical over-sampling to safely guarantee the given timing requirements, and thus render the resultant system poorly utilized. After task-specific attributes are derived, the tasks are scheduled on a single CPU and the compiled kernel is synthesized. We illustrate this process with a non-trivial servo motor control system.

병렬 클러스터 컴퓨팅 시스템에서는 노드간의 효율적인 통신이 시스템의 전체 성능을 좌우하는 중요한 요소로 인식되어 왔다. 따라서 지금까지의 많은 연구들은 노드간 통신(inter-node communication)의 성능 향상에 초점을 맞췄다. 하지만 최근 등장한 멀티 코어 프로세서(multi-core processor)는 노드간 통신 외에도 노드내 통신(intra-node communication)의 중요성을 크게 부각시키고 있다. 이와 같이 그 중요성이 점점 더 증가하고 있는 노드내 통신의 성능을 향상시키기 위해서 여러 가지 노드내 통신향상 기법들이 제안되어 왔다. 본 논문에서는 운영체제 커널의 도움으로 노드내 통신 시 발생하는 데이터 복사를 최소화하는 기법을 제안한다. 제안된 기법은 프로세스의 통신 버퍼를 상대 프로세스의 메모리 영역에 매핑하여 데이타 복사가 한번만 발생하도록 한다. 특히 제안된 기법은 리눅스 커널 버전 2.6을 위해서 설계된다. 성능 측정은 멀티 코어 프로세서를 장착한 시스템에서 이루어 졌으며, 기존 구현과 비교하여 본 논문에서 구현된 커널 모듈이 중간 및 작은 데이타 크기에 대해서 지연시간과 처리율을 각각 최대 62%와 144% 향상시킴을 보인다. 또한 프로세스가 수행되는 코어의 위치에 따라서 다른 성능을 보일 수 있음을 보인다.

In parallel cluster computing systems, the efficiency of communication between computing nodes is one of important factors that decide overall system performance. Accordingly, many researchers have studied on high-performance inter-node communication. The recently launched multi-core processor, however. increases the importance of intra-node communication as well because the more the number of cores in a node, the more the number of parallel processes running in the same node. Though there have been studies on intra-node communications, these have limited considerations on the state-of-the-art systems. In this paper, we propose a Linux kernel module that minimizes the number of data copy by exploiting the memory mapping mechanism for high-performance intra-node communication. The proposed kernel module supports the Linux kernel version 2.6. The performance measurements over a multi-core system present that the proposed kernel module can achieve lower latency up to 62% and higher throughput up to 144% than an existing kernel module approach. In addition, the measurements reveal that the performance of intra-node communication can vary significantly based on whether the cores that run the communication processes are belong to the same processor package (i.e., sharing the L2 cache).