• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.3
• /
• pp.171-179
• /
• 2020

Recently, studies have shown that convolution neural networks are achieving the best performance in image classification, object detection, and image generation. Vision based defect inspection which is more economical than other defect inspection, is a very important for a factory automation. Although supervised anomaly detection algorithm has far exceeded the performance of traditional machine learning based method, it is inefficient for real industrial field due to its tedious annotation work, In this paper, we propose ADGAN, a unsupervised anomaly detection architecture using the variational autoencoder and the generative adversarial network which give great results in image generation task, and demonstrate whether the proposed network architecture identifies anomalous images well on MNIST benchmark dataset as well as our own welding defect dataset.

• Journal of Digital Contents Society
• /
• v.4 no.1
• /
• pp.11-21
• /
• 2003

In computer vision, the goal of stereopsis is to determine the surface structure of real world form two or more perspective views of scene. It is similar to human visual system. We can avoid obstacles, recognize objects, and manipulate machine using three-dimensional information. Until recently, only gray-level images have been used as input to computation for depth determination, but the availability of color can further enhance the performance of computational stereopsis. There are many models to provide efficient color system. The simplest model, RGB model treats color as if it were composed of separate entities. Each color channel is processed individually by the same stereopsis module as used in the gray-level model. His Model decouples intensity component from color information. So it can deal with color properties without defect intensity information. Opponent color model is based on human visual system. In this model, the red-green-blue colors are combined into three opponent channels before further processing.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.14 no.7
• /
• pp.871-877
• /
• 2004

In this paper, for manufacturing a custom-made shoes, shape of foot acquired three-dimensional measurement device which makes shoe-last data for needing a custom-made shoes is founded on artificial intelligence technique and it shows method restoring to the original shape in optimized state. the developed system for this study is based on PC which uses existing three dimensional measurement method. And it gains shoe-last and data of foot shape going through 8 CCD(Charge Coupled Device) Which equipped top and bottom, right and left sides and 4 lasers which also equipped both sides and upper and lower sides. The acquired data are processed image processing algorithm using artificial intelligence technique. And result of data management is better quality of removing noise than other system not using artificial intelligence technique and it can simplify post-processing. So, this paper is constituted hardware and software system and it used neural network for determining threshold value, when input image on pre-processing step is being stage of image binarization and present that results.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.1
• /
• pp.87-93
• /
• 2012

In this paper, we proposed a faulty pattern detection algorithm of TCP(Tape Carrier Package)/COF(Chip On Film), and implemented a real-time system for inspecting TCP/COF. Since TCP/COF has very high resolution having several micro meters, the human operator should visually inspect all the parts through microscope. In this work, we implement an inspection system to detect the faulty pattern, so the operator can visually inspect only the designated parts by the inspection system through the monitor. The proposed defects detection algorithm for TCP/COF packages is implemented by the pattern matching method based on subtracting the reference image from test image. To evaluate performance of the proposal system. we made various experiments according to type of CCD camera and light source as well as illumination projection method. From experimental results, it is confirmed that the proposed system makes it possible to detect effectively the defective TCP/COF film.

• Journal of the Korea Convergence Society
• /
• v.9 no.5
• /
• pp.1-6
• /
• 2018

In smart grid an accurate load forecasting is crucial in planning resources, which aids in improving its operation efficiency and reducing the dynamic uncertainties of energy systems. Research in this area has included the use of shallow neural networks and other machine learning techniques to solve this problem. Recent researches in the field of computer vision and speech recognition, have shown great promise for Deep Neural Networks (DNN). To improve the performance of daily electric peak load forecasting the paper presents a new deep neural network model which has the architecture of two multi-layer neural networks being serially connected. The proposed network model is progressively pre-learned layer by layer ahead of learning the whole network. For both one day and two day ahead peak load forecasting the proposed models are trained and tested using four years of hourly load data obtained from the Korea Power Exchange (KPX).

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.11
• /
• pp.79-88
• /
• 1999

In this paper we address the tracking method which tracks only target object in image sequence including moving object. We use a contour tracking algorithm based on intensity and motion boundaries. The motion of the moving object contour in the image is assumed to be well describable by an affine motion model with a translation, a change in scale and a rotation. The moving object contour is represented by B-spline, the position and motion of which is estimated along the image sequence. we use pattern recognition to identify target object. In order to use linear Kalman Filters we decompose the estimation process into two filters. One is estimating the affine motion parameters and the other the shape of moving object contour. In some experiments with dial plate we show that this method enables us to obtain the robust motion estimates and tracking trajectories even in case of including obstructive object.

• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.1
• /
• pp.35-42
• /
• 2016

The core of the machine vision that is frequently used at the pre/post-production stages is a marker alignment technology. In this paper, a method to detect the angle and position of a product at high speed by use of a unique pattern present in the marker stamped on the product, and calibrate them is proposed. In the proposed method, to determine the angle and position of a marker, the candidates of the marker are extracted by using a variation of the integral histogram, and then clustering is applied to reduce the candidates. The experimental results revealed about 5s 719ms improvement in processing time and better precision in detecting the rotation angle of a product.

• Journal of KIISE
• /
• v.45 no.1
• /
• pp.30-35
• /
• 2018

GuessWhat?! is a game in which two machine players, composed of questioner and answerer, ask and answer yes-no-N/A questions about the object hidden for the answerer in the image, and the questioner chooses the correct object. GuessWhat?! has received much attention in the field of deep learning and artificial intelligence as a testbed for cutting-edge research on the interplay of computer vision and dialogue systems. In this study, we discuss the objective function and characteristics of the GuessWhat?! game. In addition, we propose a simple solver for GuessWhat?! using a simple rule-based algorithm. Although a human needs four or five questions on average to solve this problem, the proposed method outperforms state-of-the-art deep learning methods using only two questions, and exceeds human performance using five questions.

• Journal of Information Processing Systems
• /
• v.14 no.5
• /
• pp.1176-1194
• /
• 2018

The object classification using the images' contents is a big challenge in computer vision. The superpixels' information can be used to detect and classify objects in an image based on locations. In this paper, we proposed a methodology to detect and classify the image's pixels' locations using enhanced bag of words (BOW). It calculates the initial positions of each segment of an image using superpixels and then ranks it according to the region score. Further, this information is used to extract local and global features using a hybrid approach of Scale Invariant Feature Transform (SIFT) and GIST, respectively. To enhance the classification accuracy, the feature fusion technique is applied to combine local and global features vectors through weight parameter. The support vector machine classifier is a supervised algorithm is used for classification in order to analyze the proposed methodology. The Pascal Visual Object Classes Challenge 2007 (VOC2007) dataset is used in the experiment to test the results. The proposed approach gave the results in high-quality class for independent objects' locations with a mean average best overlap (MABO) of 0.833 at 1,500 locations resulting in a better detection rate. The results are compared with previous approaches and it is proved that it gave the better classification results for the non-rigid classes.

• Journal of rehabilitation welfare engineering & assistive technology
• /
• v.11 no.4
• /
• pp.331-338
• /
• 2017

In this study, a portable x-ray generator was manufactured and a real-time image acquisition system was constructed using the image intensifier from the generated generator. We have developed a real - time position error verification system that can verify whether the artificial joint position is different from the initial image from the acquired image. The template image of the region of interest is extracted from the reference image using the pattern matching technique and compared with the image to be compared. As a result, It is shown that real - time position error verification is achieved by displaying the difference angle. This system is portable type, has a self-shielding facility, and the output of the irradiation device can be manufactured in a small size of 1kw and can be used as a portable type. In case of emergency patients in the non-destructive field for industrial use, It has proved effective for use in small areas such as feet.