• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.24 no.9B
• /
• pp.1690-1696
• /
• 1999

Video sequence should be hierachically classified for the content-based retrieval. Cut detection algorithm is an essential process to classify shots. It is generally difficult for cut detection algorithms to detect cut points since a current frame is compared with a previous one, because movement of camera or object made adrupt scene change. We reduce ratio of failed cut detection so that compare the difference between frames of predicted cut point and their neighbors. In this paper, first we get predicted cut point, then we judge that the predicted cut point is true point or not. And we extracted DC images in MPEG video sequence for comparison. As a result of experiments. We confirmed that the cut detection ratio of the proposed algorithm is higher than of any other algorithms.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2006.05a
• /
• pp.197-200
• /
• 2006

A scene changing detection using histogram and header information of H.264 video stream is presented in this paper. The method using histogram is normal to be detect the changing scene. But this technique results in a lot of processing time because video data is compressed and decompressed to video stream and calculated the difference of histogram between scenes on the each frame. The method using H.264 header information is available to detect the scene change at real time without the process of calculation. Histogram and header information is more rapid for scene change detection with being the same performance in precision and recall.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.1
• /
• pp.120-124
• /
• 2010

This paper presents a method that can automatically detect variety of degradations (i.e., scratches and blotches) in old film archive. The proposed method consists of candidate detection and verification. Degradations are first identified by finding the local extreme of a frame in spatiotemporal domains, thereby using edge detector and SROD detector. Then, to remove some false alarms occurred in the first stages, the verification is performed using the texture and shape properties of scratches and blotches. The textural properties of scratches and blotches are learned using neural networks (NNs) and their shapes are represented using morphological filters. The experiments were performed on several old films, then the results demonstrated the effectiveness of the proposed method, where it has a precision of 81% and a recall of 79%.

• Journal of Korea Multimedia Society
• /
• v.24 no.5
• /
• pp.659-666
• /
• 2021

With the continuous development of deep learning, human behavior recognition algorithms have achieved good results. However, in a multi-person recognition environment, the complex behavior environment poses a great challenge to the efficiency of recognition. To this end, this paper proposes a multi-person pose estimation model. First of all, the human detectors in the top-down framework mostly use the two-stage target detection model, which runs slow down. The single-stage YOLOv3 target detection model is used to effectively improve the running speed and the generalization of the model. Depth separable convolution, which further improves the speed of target detection and improves the model's ability to extract target proposed regions; Secondly, based on the feature pyramid network combined with context semantic information in the pose estimation model, the OHEM algorithm is used to solve difficult key point detection problems, and the accuracy of multi-person pose estimation is improved; Finally, the Euclidean distance is used to calculate the spatial distance between key points, to determine the similarity of postures in the frame, and to eliminate redundant postures.

• Journal of Information Processing Systems
• /
• v.18 no.5
• /
• pp.677-687
• /
• 2022

We propose a detection algorithm based on tree-structured convolutional neural networks (TsCNNs) that finds pornography, propaganda, or other inappropriate content on a social media network. The algorithm sequentially applies the typical convolutional neural network (CNN) algorithm in a tree-like structure to minimize classification errors in similar classes, and thus improves accuracy. We implemented the detection system and conducted experiments on a data set comprised of 6 ordinary classes and 11 inappropriate classes collected from the Korean military social network. Each model of the proposed algorithm was trained, and the performance was then evaluated according to the images and videos identified. Experimental results with 20,005 new images showed that the overall accuracy in image identification achieved a high-performance level of 99.51%, and the effectiveness of the algorithm reduced identification errors by the typical CNN algorithm by 64.87 %. By reducing false alarms in video identification from the domain, the TsCNNs achieved optimal performance of 98.11% when using 10 minutes frame-sampling intervals. This indicates that classification through proper sampling contributes to the reduction of computational burden and false alarms.

• Structural Monitoring and Maintenance
• /
• v.9 no.1
• /
• pp.59-80
• /
• 2022

In the present study, the objective is to detect the structural damages using the responses obtained from the sensors at the optimal location under uncertainty conditions. Reducing the error rate in damage detection process due to responses' noise is an important goal in this study. In the proposed algorithm for optimal sensor placement, the noise of responses recorded from the sensors is initially reduced using the principal component analysis. Afterward, the optimal sensor placement is obtained by the damage detection equation based sensitivity analysis. The sensors are placed on degrees of freedom corresponding to the minimum error rate in structural damage detection through this procedure. The efficiency of the proposed method is studied on a truss bridge, a space dome, a double-layer grid as well as a three-story experimental frame structure and the results are compared. Moreover, the performance of the suggested method is compared with three other algorithms of Average Driving Point Residue (ADPR), Effective Independence (EI) method, and a mass weighting version of EI. In the examples, young's modulus, density, and cross-sectional areas of the elements are considered as uncertainty parameters. Ultimately, the results have demonstrated that the presented algorithm under uncertainty conditions represents a high accuracy to obtain the optimal sensor placement in the structures.

• ETRI Journal
• /
• v.46 no.5
• /
• pp.890-903
• /
• 2024

Anomaly detection in video surveillance is crucial but challenging due to the rarity of irregular events and ambiguity of defining anomalies. We propose a method called AONet that utilizes a spatiotemporal module to extract spatiotemporal features efficiently, as well as a residual autoencoder equipped with an attention network for effective future frame prediction in video anomaly detection. AONet utilizes a novel activation function called OptAF that combines the strengths of the ReLU, leaky ReLU, and sigmoid functions. Furthermore, the proposed method employs a combination of robust loss functions to address various aspects of prediction errors and enhance training effectiveness. The performance of the proposed method is evaluated on three widely used benchmark datasets. The results indicate that the proposed method outperforms existing state-of-the-art methods and demonstrates comparable performance, achieving area under the curve values of 97.0%, 86.9%, and 73.8% on the UCSD Ped2, CUHK Avenue, and ShanghaiTech Campus datasets, respectively. Additionally, the high speed of the proposed method enables its application to real-time tasks.

• Convergence Security Journal
• /
• v.24 no.2
• /
• pp.19-29
• /
• 2024

A crypto-jacking attack is an attack that infringes on the availability of users by stealing computing resources required for cryptocurrency mining. The target of the attack is gradually diversifying from general desktop or server environments to cloud environments. Therefore, it is essential to apply a crypto-minor detection technique suitable for various computing environments. However, since the existing detection methodologies have only been detected in a specific environment, comparative analysis has not been properly performed on the methodologies that can be applied to each environment. Therefore, in this study, classification criteria for conventional crypto-minor detection techniques are established, and a complex and integrated detection framework applicable to the cloud environment is presented through in-depth comparative analysis of existing crypto-minor detection techniques based on different experimental environments and datasets.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.6
• /
• pp.19-27
• /
• 2010

This paper proposes a method for enhancing detection performance and reducing computational cost when detecting a human face by applying B-spline active contour to the frame difference of consecutive images. Firstly, the method estimates amount of user's motion using kurtosis. If the kurtosis is smaller than a pre-defined threshold, it is considered that the amount of user's motion is insufficient and thus the contour fitting is not applied. Otherwise, the contour fitting is applied by exploiting the fact that the amount of motion is sufficient. Secondly, for the contour fitting, difference edges are detected by combining the distance transformation of the binarized frame difference and the edges of current frame. Lastly, the face is located by assigning the contour fitting process to the detected difference edges. Kurtosis-based motion amount estimation can reduce a computational cost and stabilize the results of the contour fitting. In addition, distance transformation-based difference edge detection can enhance the problems of contour lag and discontinuous difference edges. Experimental results confirm that the proposed method can reduce the face localization error caused by the contour lag and discontinuity of edges, and decrease the computational cost by omitting approximately 39% of the contour fitting.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.6
• /
• pp.413-422
• /
• 2013

In this paper, we propose a novel method for real-time lane detection that is robust for inclined roads and not require a camera parameter, the Inverse Perspective Transform of the image, and the proposed lane filter. After finding the vanishing point from the start frame of the image and storing the region surrounding the vanishing point as the Template Area(TA), our method predict the lanes by scanning toward the lower part from the vanishing point of the image and obtain the image removed the perspective effect using the Inverse Perspective Transform coefficients extracted based on the predicted lanes. To robustly determine lanes on inclined roads, the region surrounding the vanishing point is set up as the template area (TA), and, by recalculating the vanishing point by tracing the area similar to the TA (SA) in the input image through template matching, it responds to the changes on the road conditions. The proposed method for a more robust lane detection method for inclined roads is a lane detection method by applying a lane detection filter on an image removed of the perspective effect. Through this method, the processing region is reduced and the processing procedure is simplified to produce a satisfactory lane detection result of about 40 frames per second.