• Journal of Korea Multimedia Society
• /
• v.24 no.5
• /
• pp.695-700
• /
• 2021

The Faster RCNN-based shellfish recognition algorithm is introduced for shellfish recognition studies that currently do not have any deep learning-based algorithms in a practical setting. The original feature extraction module is replaced by DenseNet, which fuses multi-level feature data and optimises the NMS algorithm, network depth and merging method; overcoming the omission of shellfish overlap, multiple shellfish and insufficient light, effectively solving the problem of low shellfish classification accuracy. In the complexifier test environment, the test accuracy was improved by nearly 4%. Higher testing accuracy was achieved compared to the original testing algorithm. This provides favourable technical support for future applications of the improved Faster RCNN approach to seafood quality classification.

• The Journal of Korean Institute of Information Technology
• /
• v.17 no.5
• /
• pp.13-20
• /
• 2019

Recently, object detection technology based on CNN has been actively studied. Object detection technology is used as an important technology in autonomous vehicles, intelligent image analysis, and so on. In this paper, we propose a rotation change robust object detection system by applying MI-FL (Moment Invariant-Feature Layer) to SSD (Single Shot Multibox Detector) which is one of CNN-based object detectors. First, the features of the input image are extracted based on the VGG network. Then, a total of six feature layers are applied to generate bounding boxes by predicting the location and type of object. We then use the NMS algorithm to get the bounding box that is the most likely object. Once an object bounding box has been determined, the invariant moment feature of the corresponding region is extracted using MI-FL, and stored and learned in advance. In the detection process, it is possible to detect the rotated image more robust than the conventional method by using the previously stored moment invariant feature information. The performance improvement of about 4 ~ 5% was confirmed by comparing SSD with existing SSD and MI-FL.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.22 no.1
• /
• pp.1-10
• /
• 2019

In this paper, we propose the track initiation algorithm based on the weighted score for TWS radar tracking. This algorithm utilizes radar velocity information to calculate the probabilistic track score and applies the Non-Maximum-Suppression(NMS) to confirm the targets to track. This approach is understood as a modification of a conventional track initiation algorithm in a probabilistic manner. Also, we additionally apply the weighted Hough transform to compensate a measurement error, and it helps to improve the track detection probability. We designed the simulator in order to demonstrate the performance of the proposed track initiation algorithm. The simulation result show that the proposed algorithm, which reduces about 40 % of a false track probability, is better than the conventional algorithm.

• Journal of Korea Multimedia Society
• /
• v.18 no.5
• /
• pp.631-640
• /
• 2015

Recently CCTV system is installed widely purpose to enhanced physical security, gathering criminal evidence and management of facilities. In spite of supporting strong management function, CCTV system has weak security function. Therefore high security management function is required. Generally it's not easy to control the devices under NAT using a NMS(Network Management System). So we design and implement alive check algorithm of CCTV devices under NAT using DVRNS address resolution and TCP session check. We evaluated and analyzed of developed system on real environment which includes about 100 DVRs under NAT. As a result of test, it showed that device alive check and DVRNS address resolution were well performed without any error.

• Nuclear Engineering and Technology
• /
• v.54 no.12
• /
• pp.4625-4635
• /
• 2022

A low-energy dielectric loaded accelerator with a non-uniform, multi-segment structure is studied and optimized. So far, no analytical solution is provided for such structures. Also, due to the existing nonlinear behavior and a large number of geometric parameters, the problem of numerical optimizations is complex. For this reason, a method is presented to design and optimize such structures using the Genetic Algorithm (GA). Moreover, the GA output results are compared with Trust Region (TR) and Nelder-Mead Simplex (NMS) methods. Comparative results show that the GA is more efficient in achieving optimization goals and also has a higher speed than the two other methods. Finally, an optimized accelerating tube is integrated into a proper coupler. Then, the accelerator is simulated for full electromagnetic investigations using the CST suite of codes. This design leads to a structure with a power of about 80 kW in the X-band, which delivers electrons to the output energy in the range of 300-459 kV. The length and outer diameter of the accelerating tube obtained are 10 cm and 1 cm, respectively.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10c
• /
• pp.781-783
• /
• 2001

동영상을 효율적으로 압축하기 위한 움직임벡터 예측에 관한 많은 연구가 진행되어 왔다. 가장 일반적인 FBMA(Full search-based Block Matching Algorithm)는 화질은 좋지만 계산량이 많기 때문에 실시간 인코딩을 요구하는 시스템에서 사용하는데 문제가 있다. 좋은 화질을 유지하면서 인코딩 속도를 해결하기 위한 많은 알고리즘들이 제안되어 왔지만 ASIC이나 소형 시스템에서 사용할 수 있는 방법이 계속 요구되고 있다. 본 논문에서는 계산량을 더욱 줄여 속도향상을 가져오면서 FBMA와 비숫한 SNR을 유지하는 방법인 NMS(New Fast Motion Estimation Search With Subsmapling Method)를 제안하였다. NMS는 서브샘플림한 값을 이용하여 SAD값을 구하고 또한 새로운 Search를 제안하여 기존 방법들이 제공하는 주관적 화질이나 PSNR을 높게 유지하면서도 속도를 10~15% 정도 개선시킬 수 있다.

• Annual Conference of KIPS
• /
• 2008.05a
• /
• pp.950-953
• /
• 2008

네트워크 망 관리를 위해 사용되는 SNMP(Simple Network Management Protocol)는 다양한 네트워크장비에 대해 표준화된 정보를 제공한다. 네트워크의 고속화, 대형화로 인해 관리를 위한 정보 수집 트래픽이 증가하고 정보 수집 시간 지연 문제가 발생한다. 본 논문에서는 기존의 SNMP를 이용한 정보수집 방법의 문제점을 파악하여 네트워크 관리를 위한 트래픽량을 줄이고, Polling 소비 시간을 최소화하는 알고리즘을 제안한다. 제안한 알고리즘은 관리 대상 시스템의 각 링크의 트래픽 변화 유무를 예측하여 불필요한 수집을 줄이는 방법이다. 관리 대상 시스템이 많고, 주기적으로 관리 정보를 수집하는 경우 제안한 알고리즘이 효율적으로 사용될 것으로 기대된다. Enterprise 네트워크 형태의 학교 Campus NMS에 적용하여 알고리즘의 타당성을 증명하였다.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.12 no.6
• /
• pp.13-23
• /
• 2008

In this study, a finite element (FE) model updating method based on the hybrid genetic algorithm (HGA) is proposed to improve the grillage FE model for plate girder bridges. HGA consists of a genetic algorithm (GA) and direct search method (DS) based on a modification of Nelder & Mead's simplex optimization method (NMS). Fitness functions based on natural frequencies, mode shapes, and static deflections making use of the measurements and analytical results are also presented to apply in the proposed method. In addition, a multi-objective function has been formulated as a linear combination of fitness functions in order to simultaneously improve both stiffness and mass. The applicability of the proposed method to girder bridge structures has been verified through a numerical example on a two-span continuous grillage FE model, as well as through an experimental test on a simply supported plate girder skew bridge. In addition, the effect of measuring error is considered as random noise, and its effect is investigated by numerical simulation. Through numerical and experimental verification, it has been proven that the proposed method is feasible and effective for FE model updating on plate girder bridges.

• Journal of Communications and Networks
• /
• v.12 no.1
• /
• pp.74-85
• /
• 2010

For the purpose of compromising hosts, attackers including infected hosts initially perform a portscan using IP addresses in order to find vulnerable hosts. Considerable research related to portscan detection has been done and many algorithms have been proposed and implemented in the network intrusion detection system (NIDS). In order to distinguish portscanners from remote hosts, most portscan detection algorithms use a fixed threshold that is manually managed by the network manager. Because the threshold is a constant, even though the network environment or the characteristics of traffic can change, many false positives and false negatives are generated by NIDS. This reduces the efficiency of NIDS and imposes a high processing burden on a network management system (NMS). In this paper, in order to address this problem, we propose an automatic portscan detection system using an fast increase slow decrease (FISD) scheme, that will automatically and adaptively set the threshold based on statistical data for traffic during prior time periods. In particular, we focus on reducing false positives rather than false negatives, while the threshold is adaptively set within a range between minimum and maximum values. We also propose a new portscan detection algorithm, rate of increase in the number of failed connection request (RINF), which is much more suitable for our system and shows better performance than other existing algorithms. In terms of the implementation, we compare our scheme with other two simple threshold estimation methods for an adaptive threshold setting scheme. Also, we compare our detection algorithm with other three existing approaches for portscan detection using a real traffic trace. In summary, we show that FISD results in less false positives than other schemes and RINF can fast and accurately detect portscanners. We also show that the proposed system, including our scheme and algorithm, provides good performance in terms of the rate of false positives.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.1795-1811
• /
• 2019

In order to achieve rapid and accurate detection of vehicle objects in complex traffic conditions, we propose a novel vehicle detection method. Firstly, more contextual and small-object vehicle information can be obtained by our Joint Feature Network (JFN). Secondly, our Evolved Region Proposal Network (EPRN) generates initial anchor boxes by adding an improved version of the region proposal network in this network, and at the same time filters out a large number of false vehicle boxes by soft-Non Maximum Suppression (NMS). Then, our Mask Network (MaskN) generates an example that includes the vehicle occlusion, the generator and discriminator can learn from each other in order to further improve the vehicle object detection capability. Finally, these candidate vehicle detection boxes are optimized to obtain the final vehicle detection boxes by the Fine-Tuning Network(FTN). Through the evaluation experiment on the DETRAC benchmark dataset, we find that in terms of mAP, our method exceeds Faster-RCNN by 11.15%, YOLO by 11.88%, and EB by 1.64%. Besides, our algorithm also has achieved top2 comaring with MS-CNN, YOLO-v3, RefineNet, RetinaNet, Faster-rcnn, DSSD and YOLO-v2 of vehicle category in KITTI dataset.