• Journal of KIISE
• /
• v.44 no.12
• /
• pp.1282-1289
• /
• 2017

Since the study of deep convolutional neural network became prevalent, one of the important discoveries is that a feature map from a convolutional network can be extracted before going into the fully connected layer and can be used as a saliency map for object detection. Furthermore, the model can use features from each different layer for accurate object detection: the features from different layers can have different properties. As the model goes deeper, it has many latent skip connections and feature maps to elaborate object detection. Although there are many intermediate layers that we can use for semantic segmentation through skip connection, still the characteristics of each skip connection and the best skip connection for this task are uncertain. Therefore, in this study, we exhaustively research skip connections of state-of-the-art deep convolutional networks and investigate the characteristics of the features from each intermediate layer. In addition, this study would suggest how to use a recent deep neural network model for semantic segmentation and it would therefore become a cornerstone for later studies with the state-of-the-art network models.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.46 no.7
• /
• pp.13-23
• /
• 2009

In this paper, the network performance of a turbo coded optical code division multiple access (CDMA) system with cross-layer, which is between physical and network layers, concept is analyzed and simulated. We consider physical and MAC layers in a cross-layer concept. An intensity-modulated/direct-detection (IM/DD) optical system employing pulse position modulation (PPM) is considered. In order to increase the system performance, turbo codes composed of parallel concatenated convolutional codes (PCCCs) is utilized. The network performance is evaluated in terms of bit error probability (BEP). From the simulation results, it is demonstrated that turbo coding offers considerable coding gain with reasonable encoding and decoding complexity. Also, it is confirmed that the performance of such an optical CDMA network can be substantially improved by increasing e interleaver length and e number of iterations in e decoding process. The results of this paper can be applied to implement the indoor optical wireless LANs.

• Journal of Biosystems Engineering
• /
• v.37 no.4
• /
• pp.271-278
• /
• 2012

Worldwide trends in animal welfare have resulted in an increased interest in individual management of sows housed in groups within hog barns. Estrus detection has been shown to be one of the greatest determinants of sow productivity. Purpose: We conducted this study to develop a method that can automatically detect the estrus state of a sow by selecting optimal texture parameters from images of a sow's pudendum and by optimizing the number of neurons in the hidden layer of an artificial neural network. Methods: Texture parameters were analyzed according to changes in a sow's pudendum in estrus such as mucus secretion and expansion. Of the texture parameters, eight gray level co-occurrence matrix (GLCM) parameters were used for image analysis. The image states were classified into ten grades for each GLCM parameter, and an artificial neural network was formed using the values for each grade as inputs to discriminate the estrus state of sows. The number of hidden layer neurons in the artificial neural network is an important parameter in neural network design. Therefore, we determined the optimal number of hidden layer units using a trial and error method while increasing the number of neurons. Results: Fifteen hidden layers were determined to be optimal for use in the artificial neural network designed in this study. Thirty images of 10 sows were used for learning, and then 30 different images of 10 sows were used for verification. Conclusions: For learning, the back propagation neural network (BPN) algorithm was used to successful estimate six texture parameters (homogeneity, angular second moment, energy, maximum probability, entropy, and GLCM correlation). Based on the verification results, homogeneity was determined to be the most important texture parameter, and resulted in an estrus detection rate of 70%.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.47 no.5
• /
• pp.62-68
• /
• 2010

This study is to identify target locations with low false alarms on thermal infrared images obtained from natural environment. The proposed method is different from the previous researches because it uses morphology filters for Gabor response images instead of an intensity image in initial detection stage. This method does not need precise extracting a target silhouette to distinguish true targets or clutters. It comprises three distinct stages. First, morphological operations and adaptive thresholding are applied to the summation image of four Gabor responses of an input image to find out salient regions. The locations of extracted regions can be classified into targets or clutters. Second, local texture features are computed from salient regions of an input image. Finally, the local texture features are compared with the training data to distinguish between true targets and clutters. The multi-layer perceptron having three layers is used as a classifier. The performance of the proposed method is proved by using natural infrared images. Therefore it can be applied to real automatic target detection systems.

• Membrane Journal
• /
• v.15 no.3
• /
• pp.198-205
• /
• 2005

Double-layered polymer films to assay the radioactive contamination were formulated using polysulfone (PSF) and cerium activated yttrium silicate (CAYS), consisting of a dense support layer and a CAYS-holding top layer prepared via the diffusion-induced phase inversion. As the vapor exposure process was omitted, the CAYS-holding layer showed a typical asymmetric structure, with CAYS being transfixed into the polymer network spread with large macropores. With the increase in vapor exposure time before immersion, morphology of the films transformed from asymmetric to sponge-like structures, with CAYS being localized in cellular structure. The border structure between the two layers reflects the phase inversion behavior of a cast solution during the coagulation. In the radioactive detection, the polymer phase in a film holding a sponge-like structure is so dense that the radionuclides, deposited on the film, could not filter through the phase, consequently resulting in the loss in the detection efficiency of the film.

• Journal of Broadcast Engineering
• /
• v.26 no.2
• /
• pp.184-196
• /
• 2021

This paper proposes statistics adaptive linear regression-based object size prediction method for object detection. YOLOv2 and YOLOv3, which are typical deep learning-based object detection algorithms, designed the last layer of a network using statistics adaptive exponential regression model to predict the size of objects. However, an exponential regression model can propagate a high derivative of a loss function into all parameters in a network because of the property of an exponential function. We propose statistics adaptive linear regression layer to ease the gradient exploding problem of the exponential regression model. The proposed statistics adaptive linear regression model is used in the last layer of the network to predict the size of objects with statistics estimated from training dataset. We newly designed the network based on the YOLOv3tiny and it shows the higher performance compared to YOLOv3 tiny on the UFPR-ALPR dataset.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.6
• /
• pp.1330-1334
• /
• 2012

The submarines reach their weakest point when they sail on the surface to operate snorkel and periscope. At this period, however, there lies a high possibility that the submarines are detected by non-acoustic sensors such as radars, IR signatures, and human observations. In this paper, the non-acoustic stealth was adopted on the mast and periscope of submarines so as to overcome their vulnerability of being easily detected in this given situation. First of all, the non-acoustic detection sensors were investigated and the stealth methods were analyzed. And multi-layered structures consisting of RAM layer, IR layer, and Camouflage layer were proposed on the surface of the submarine. As a results, multi-layered structure was suggested with 3~5 mm of a magnetic material such as ferrite for RAM layer, 1~2 mm of ceramic or nickel for IR layer, and sea-blue paint for Camouflage layer.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.3
• /
• pp.19-27
• /
• 2019

This paper proposes a new method using multi-layer overlapped windows to improve the performance of YOLO network which is vulnerable to detect small dense objects. In particular, the proposed method uses the YOLO Network based on the multi-layer overlapped windows to track small dense vehicles that approach from long distances. The method improves the detection performance for location and size of small vehicles. It allows crossing area of two multi-layer overlapped windows to track moving vehicles from a long distance to a short distance. And the YOLO network is optimized so that GPU computation time due to multi-layer overlapped windows should be reduced. The superiority of the proposed algorithm has been proved through various experiments using captured images from road surveillance cameras.

• ETRI Journal
• /
• v.42 no.3
• /
• pp.411-419
• /
• 2020

In real-world intelligent transportation systems, accuracy in vehicle license plate detection and recognition is considered quite critical. Many algorithms have been proposed for still images, but their accuracy on actual videos is not satisfactory. This stems from several problematic conditions in videos, such as vehicle motion blur, variety in viewpoints, outliers, and the lack of publicly available video datasets. In this study, we focus on these challenges and propose a license plate detection and recognition scheme for videos based on a temporal matching prior network. Specifically, to improve the robustness of detection and recognition accuracy in the presence of motion blur and outliers, forward and bidirectional matching priors between consecutive frames are properly combined with layer structures specifically designed for plate detection. We also built our own video dataset for the deep training of the proposed network. During network training, we perform data augmentation based on image rotation to increase robustness regarding the various viewpoints in videos.

• Journal of the Korean Society of Safety
• /
• v.19 no.1
• /
• pp.49-55
• /
• 2004

This study presents an analysis of comparison of P-type fire detection system with fuzzy logic-applied fire detection system. The fuzzy logic-applied fire detection system has input variables obtained by fire experiment of small scale with K-type temperature sensor and optical smoke sensor. And the antecedent part of fuzzy rules consists of temperature and smoke density, and the consequent part consists of fire probability. Also triangular fuzzy membership function is used for input variables and fuzzy rules. To calculate the final fire probability a centroid method is introduced. A fire experiment is conducted with controlling wood crib layer, cigarette to simulate actual fire and false alarm situation. The results show that peak fire probability is 25[%] for non-fire and is more than 80[%] for fire situation, respectively. The fuzzy logic-applied fire detection system suggested here is able to distinguish fire situation and non-fire situation very precisely.