• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.2
• /
• pp.33-44
• /
• 2010

This paper propose the structure of real-time face detection hardware architecture for robot vision processing applications. The proposed architecture is robust against illumination changes and operates at no less than 60 frames per second. It uses Modified Census Transform to obtain face characteristics robust against illumination changes. And the AdaBoost algorithm is adopted to learn and generate the characteristics of the face data, and finally detected the face using this data. This paper describes the face detection hardware structure composed of Memory Interface, Image Scaler, MCT Generator, Candidate Detector, Confidence Comparator, Position Resizer, Data Grouper, and Detected Result Display, and verification Result of Hardware Implementation with using Virtex5 LX330 FPGA of Xilinx. Verification result with using the images from a camera showed that maximum 32 faces per one frame can be detected at the speed of maximum 149 frame per second.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.2
• /
• pp.343-350
• /
• 2022

It is worth verifying the effectiveness of data integration between data with different features. This study investigated whether the data integration affects the accuracy of deep neural network (DNN), and which integration method shows the best improvement. This study used two different public datasets. One public dataset was taken in an actual farm in India. And another was taken in a laboratory environment in Korea. Leaf images were selected from two different public datasets to have five classes which includes normal and four different types of plant diseases. DNN used pre-trained VGG16 as a feature extractor and multi-layer perceptron as a classifier. Data were integrated into three different ways to be used for the training process. DNN was trained in a supervised manner via the integrated data. The trained DNN was evaluated by using a test dataset taken in an actual farm. DNN shows the best accuracy for the test dataset when DNN was first trained by images taken in the laboratory environment and then trained by images taken in the actual farm. The results show that data integration between plant images taken in a different environment helps improve the performance of deep neural networks. And the results also confirmed that independent use of plant images taken in different environments during the training process is more effective in improving the performance of DNN.

• Smart Media Journal
• /
• v.5 no.1
• /
• pp.15-23
• /
• 2016

This paper proposes an effective algorithm of Korean license plate recognition. By applying Haar-like feature and Canny edge detection on a captured vehicle image, it is possible to find a connected rectangular, which is a strong candidate for license plate. The color information of license plate separates plates into white and green. Then, OTSU binary image processing and foreground neighbor pixel propagation algorithm CLNF will be applied to each license plates to reduce noise except numbers and letters. Finally, through labeling, numbers and letters will be extracted from the license plate. Letter and number regions, separated from the plate, pass through mesh method and thinning process for extracting feature vectors by X-Y projection method. The extracted feature vectors are classified using neural networks trained by backpropagation algorithm to execute final recognition process. The experiment results show that the proposed license plate recognition algorithm works effectively.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.3
• /
• pp.845-850
• /
• 2023

In this paper, we propose a method for diagnosing ball bearing vibration using transfer learning. STFT, which can analyze vibration signals in time-frequency, was used as input to CNN to diagnose failures. In order to rapidly learn CNN-based deep artificial neural networks and improve diagnostic performance, we proposed a transfer learning-based deep learning learning technique. For transfer learning, the feature extractor and classifier were selectively learned using a VGG-based image classification model, the data set for learning was publicly available ball bearing vibration data provided by Case Western Reserve University, and performance was evaluated by comparing the proposed method with the existing CNN model. Experimental results not only prove that transfer learning is useful for condition diagnosis in ball bearing vibration data, but also allow other industries to use transfer learning to improve condition diagnosis.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.2
• /
• pp.94-101
• /
• 2012

This paper proposes an improved image steganalysis scheme to raise the detection rate of stego images out of cover images. To improve the detection rate of stego image in the steganalysis, tiny variation caused by data hiding should be amplified. For this, we extract feature vectors of cover image and stego image by two steps. First, we separate image into upper 4 bit subimage and lower 4 bit subimage. As a result, stego noise is expanded more than two times. We decompose separated subimages into twelve subbands by applying 3-level Haar wavelet transform and calculate co-occurrence probabilities of two different subbands in the same scale. Since co-occurrence probability of the two wavelet subbands is affected by data hiding, it can be used as a feature to differentiate cover images and stego images. The extracted feature vectors are used as the input to the multilayer perceptron(MLP) classifier to distinguish between cover and stego images. We test the performance of the proposed scheme over various embedding rates by the LSB, S-tool, COX's SS, and F5 embedding method. The proposed scheme outperforms the previous schemes in detection rate to existence of hidden message as well as exactness of discrimination.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.2
• /
• pp.397-404
• /
• 2013

This paper propose the hardware architecture of face detection hardware system using the AdaBoost algorithm. The proposed structure of face detection hardware system is possible to work in 30frame per second and in real time. And the AdaBoost algorithm is adopted to learn and generate the characteristics of the face data by Matlab, and finally detected the face using this data. This paper describes the face detection hardware structure composed of image scaler, integral image extraction, face comparing, memory interface, data grouper and detected result display. The proposed circuit is so designed to process one point in one cycle that the prosed design can process full HD($1920{\times}1080$) image at 70MHz, which is approximate $2316087{\times}30$ cycle. Furthermore, This paper use the reducing the word length by Overflow to reduce memory size. and the proposed structure for face detection has been designed using Verilog HDL and modified in Mentor Graphics Modelsim. The proposed structure has been work on 45MHz operating frequency and use 74,757 LUT in FPGA Xilinx Virtex-5 XC5LX330.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.21 no.1
• /
• pp.54-61
• /
• 2001

A software package to classify acoustic emission (AE) signals using the wavelet transform and the neural network was developed Both of the continuous and the discrete wavelet transforms are considered, and the error back-propagation neural network is adopted as m artificial neural network algorithm. The signals acquired during the 3-point bending test of specimens which have artificial defects on weld zone are used for the classification of the defects. Features are extracted from the time-frequency plane which is the result of the wavelet transform of signals, and the neural network classifier is tamed using the extracted features to classify the signals. It has been shown that the developed software package is useful to classify AE signals. The difference between the classification results by the continuous and the discrete wavelet transforms is also discussed.

• Journal of the Institute of Electronics Engineers of Korea TE
• /
• v.42 no.1
• /
• pp.13-20
• /
• 2005

In this paper, I recognized Korean isolated words using CVEM which is suggested by M. Bilginer et al. CVEM is an algorithm which is easy to extract the common properties from training voice signals and also doesn't need complex calculation. In addition CVEM shows high accuracy in recognition results. But, CVEM has couple of problems which are impossible to use for many training voices and no discriminant information among extracted common vectors. To get the optimal common vectors from certain voice classes, various voices should be used for training. But CVEM is impossible to get continuous high accuracy in recognition because CVEM has a limitation to use many training voices and the absence of discriminant information among common vectors can be the source of critical errors. To solve above problems and improve recognition rate, k-clustering subspace method and DCVEM suggested. And did various experiments using voice signal database made by ETRI to prove the validity of suggested methods. The result of experiments shows improvements in performance. And with proposed methods, all the CVEM problems can be solved with out calculation problem.

• Journal of the Korea Convergence Society
• /
• v.10 no.5
• /
• pp.35-42
• /
• 2019

Recently, as the Internet and various wearable devices have appeared, Internet technology has contributed to obtaining more convenient information and doing business. However, as the internet is used in various parts, the attack surface points that are exposed to attacks are increasing, Attempts to invade networks aimed at taking unfair advantage, such as cyber terrorism, are also increasing. In this paper, we propose a feature selection method to improve the classification performance of the class to classify the abnormal behavior in the network traffic. The UNSW-NB15 dataset has a rare class imbalance problem with relatively few instances compared to other classes, and an undersampling method is used to eliminate it. We use the SVM, k-NN, and decision tree algorithms and extract a subset of combinations with superior detection accuracy and RMSE through training and verification. The subset has recall values of more than 98% through the wrapper based experiments and the DT_PSO showed the best performance.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.21 no.6
• /
• pp.749-754
• /
• 2011

In this study, we propose the design of optimized pRBFNNs-based face recognition system using two-dimensional Image and ASM algorithm. usually the existing 2 dimensional face recognition methods have the effects of the scale change of the image, position variation or the backgrounds of an image. In this paper, the face region information obtained from the detected face region is used for the compensation of these defects. In this paper, we use a CCD camera to obtain a picture frame directly. By using histogram equalization method, we can partially enhance the distorted image influenced by natural as well as artificial illumination. AdaBoost algorithm is used for the detection of face image between face and non-face image area. We can butt up personal profile by extracting the both face contour and shape using ASM(Active Shape Model) and then reduce dimension of image data using PCA. The proposed pRBFNNs consists of three functional modules such as the condition part, the conclusion part, and the inference part. In the condition part of fuzzy rules, input space is partitioned with Fuzzy C-Means clustering. In the conclusion part of rules, the connection weight of RBFNNs is represented as three kinds of polynomials such as constant, linear, and quadratic. The essential design parameters (including learning rate, momentum coefficient and fuzzification coefficient) of the networks are optimized by means of Differential Evolution. The proposed pRBFNNs are applied to real-time face image database and then demonstrated from viewpoint of the output performance and recognition rate.