• Journal of the Korea Society of Computer and Information
• /
• v.25 no.3
• /
• pp.19-26
• /
• 2020

Steel production requires high-quality surfaces with minimal defects. Therefore, the detection algorithms for the surface defects of steel strip should have good generalization performance. To meet the growing demand for high-quality products, the use of intelligent visual inspection systems is becoming essential in production lines. In this paper, we proposed a ShuffleDefectNet defect detection system based on deep learning. The proposed defect detection system exceeds state-of-the-art performance for defect detection on the Northeastern University (NEU) dataset obtaining a mean average accuracy of 99.75%. We train the best performing detection with different amounts of training data and observe the performance of detection. We notice that accuracy and speed improve significantly when use the overall architecture of ShuffleDefectNet.

• Journal of the Korean Society for Precision Engineering
• /
• v.22 no.3 s.168
• /
• pp.170-178
• /
• 2005

The purpose of this study is to develop a gait-event detection system, which is necessary for the cycle-to-cycle FES control of locomotion. Proposed gait event detection system consists of a signal measurement part and gait event detection part. The signal measurement was composed of the sensors and the LabVIEW program for the data acquisition and synchronization of the sensor signals. We also used a video camera and a motion capture system to get the reference gait events. Machine learning technique with ANN (artificial neural network) was adopted for automatic detection of gait events. 2 cycles of reference gait events were used as the teacher signals for ANN training and the remnants ($2\sim5$ cycles) were used fur the evaluation of the performance in gait-event detection. 14 combinations of sensor signals were used in the training and evaluation of ANN to examine the relationship between the number of sensors and the gait-event detection performance. The best combinations with minimum errors of event-detection time were 1) goniometer, foot-switch and 2) goniometer, foot-switch, accelerometer x(anterior-posterior) component. It is expected that the result of this study will be useful in the design of cycle-to-cycle FES controller.

• Journal of Korea Multimedia Society
• /
• v.18 no.12
• /
• pp.1519-1527
• /
• 2015

TFT-LCD automatic defect inspection system for detecting defects in place of the visual tester does pre-processing, candidate defect pixel detection, and recognition and classification through a blob analysis. An over-detection result of defects acts as an undue burden of blob analysis for recognition and classification. In this paper, we propose defect detection method based on the histogram distribution modeling of TFT-LCD image to minimize over-detection of candidate defective pixels. Primary defect candidate pixels are detected estimating the skewness of the luminance distribution histogram of the background pixels. Based on the detected defect pixels, the defective pixels other than noise pixels are detected using the distribution histogram model of the local area. Experimental results confirm that the proposed method shows an excellent defect detection result on the image containing the various types of defects and the reduction of the degree of over-detection as well.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.7
• /
• pp.1117-1122
• /
• 2017

Pedestrian detection is a very important and valuable part of artificial intelligence and computer vision. It can be used in various areas for example automatic drive, video analysis and others. Many works have been done for the pedestrian detection. The accuracy of pedestrian detection on multiple pedestrian image has reached high level. It is not easily get more progress now. This paper proposes a new structure based on the idea of HOG and convolutional filters to do the pedestrian detection in single pedestrian image. It can be a method to increase the accuracy depend on the high accuracy in single pedestrian detection. In this paper, we use Multiblock HOG and magnitude of the pixel as the feature and use convolutional filter to do the to extract the feature. And then use NEWFM to be the classifier for training and testing. We use single pedestrian image of the INRIA data set as the data set. The result shows that the Convolutional Multiblock HOG we proposed get better performance which is 0.015 miss rate at 10-4 false positive than the other detection methods for example HOGLBP which is 0.03 miss rate and ChnFtrs which is 0.075 miss rate.

• Journal of Sensor Science and Technology
• /
• v.29 no.3
• /
• pp.194-200
• /
• 2020

Human detection is an important aspect in many video-based sensing and monitoring systems. Studies have been actively conducted for the automatic detection of humans in camera images, and various methods have been proposed. However, there are still problems in terms of performance and computational cost. In this paper, we describe a method for efficient human detection in the field of view of a camera, which may be static or moving, through multiple processing steps. A detection line is designated at the position where a human appears first in a sensing area, and only the one-dimensional gray pixel values of the line are monitored. If any noticeable change occurs in the detection line, corner detection and optical flow computation are performed in the vicinity of the detection line to confirm the change. When significant changes are observed in the corner numbers and optical flow vectors, the final determination of human presence in the monitoring area is performed using the Histograms of Oriented Gradients method and a Support Vector Machine. The proposed method requires processing only specific small areas of two consecutive gray images. Furthermore, this method enables operation not only in a static condition with a fixed camera, but also in a dynamic condition such as an operation using a camera attached to a moving vehicle.

• Journal of IKEEE
• /
• v.26 no.4
• /
• pp.622-627
• /
• 2022

This paper proposes an intrusion detection system for in-vehicle network to improve detection performance considering attack counts and attack types. In intrusion detection system, both FNR (False Negative Rate), where intrusion frame is misjudged as normal frame, and FPR (False Positive Rate), where normal frame is misjudged as intrusion frame, seriously affect vechicle safety. This paper proposes a novel intrusion detection algorithm to improve both FNR and FPR, where data frame previously detected as intrusion above certain attack counts is automatically detected as intrusion and the automatic intrusion detection method is adaptively applied according to attack types. From the simulation results, the propsoed method effectively improve both FNR and FPR in DoS(Denial of Service) attack and spoofing attack.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.34 no.6
• /
• pp.619-627
• /
• 2016

Cells of a PV (photovoltaic) module can suffer defects due to various causes resulting in a loss of power output. As a malfunctioning cell has a higher temperature than adjacent normal cells, it can be easily detected with a thermal infrared sensor. A conventional method of PV cell inspection is to use a hand-held infrared sensor for visual inspection. The main disadvantages of this method, when applied to a large-scale PV power plant, are that it is time-consuming and costly. This paper presents an algorithm for automatically detecting defective PV panels using images captured with a thermal imaging camera from an UAV (unmanned aerial vehicle). The proposed algorithm uses statistical analysis of thermal intensity (surface temperature) characteristics of each PV module to verify the mean intensity and standard deviation of each panel as parameters for fault diagnosis. One of the characteristics of thermal infrared imaging is that the larger the distance between sensor and target, the lower the measured temperature of the object. Consequently, a global detection rule using the mean intensity of all panels in the fault detection algorithm is not applicable. Therefore, a local detection rule was applied to automatically detect defective panels using the mean intensity and standard deviation range of each panel by array. The performance of the proposed algorithm was tested on three sample images; this verified a detection accuracy of defective panels of 97% or higher. In addition, as the proposed algorithm can adjust the range of threshold values for judging malfunction at the array level, the local detection rule is considered better suited for highly sensitive fault detection compared to a global detection rule. In this study, we used a panel area extraction method that we previously developed; fault detection accuracy would be improved if panel area extraction from images was more precise. Furthermore, the proposed algorithm contributes to the development of a maintenance and repair system for large-scale PV power plants, in combination with a geo-referencing algorithm for accurate determination of panel locations using sensor-based orientation parameters and photogrammetry from ground control points.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.11 no.2
• /
• pp.29-38
• /
• 2012

The data quality and measurements along consecutive detector stations can vary much even in the same traffic conditions due to variety in detector types, calibration and maintenance effort, field operation periods, minor geometric changes of roads and so on. These faulty situations often create 10% or more of inherent difference in important traffic measurements between two stations even under stable low flow condition. Low detection rates(DR) and high false alarm rates(FAR) therefore sets in among many popular Automatic Incident Detection Algorithms(AIDA). This research is two-folded and aims mainly to develop a new AIDA for uninterrupted flow. For this purpose, a technique which utilizes a Simple Arithmetic Operation(SAO) of traffic variables is introduced. This SAO technique is designed to address the inherent discrepancy of detector data observed successive stations, and to overcome the degradation of AIDA performance. It was found that this new algorithm improves DR as much as 95 percent and above. And mean time to detection(MTTD) is found to be 1 minutes or less. When it comes to FAR, this new approach compared to existing AIDAs reduces FAR up to 31.0 percent. And capability in persistency check of on-going incidents was found excellent as well.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.695-699
• /
• 2003

It is general to use the control network in control systems in order to reduce the complexity of the related wiring harnesses and to improve the system flexibility. CAN becomes one of the most popular network protocols because of its low price, multiple sources, high performance and reliability. This paper describes a CAN based real-time control of the fire detection system for the intelligent building system. The proposed fire detection and alarm system is stronger than the previous one against noises and communication media faults and can solve many problems such as complex cabling and increment of I/O ports by using many sensors. Furthermore, MMI can be achieved easily with the personal computer that is used for replacing the traditional monitoring system. The proposed system is implemented and the experimental results are given.

• International Journal of Contents
• /
• v.14 no.1
• /
• pp.1-6
• /
• 2018

Text, as one of the most influential inventions of humanity, has played an important role in human life since ancient times. The rich and precise information embodied in text is very useful in a wide range of vision-based applications such as the text data extracted from images that can provide information for automatic annotation, indexing, language translation, and the assistance systems for impaired persons. Therefore, natural-scene text detection with active research topics regarding computer vision and document analysis is very important. Previous methods have poor performances due to numerous false-positive and true-negative regions. In this paper, a fully-convolutional-network (FCN)-based method that uses supervised architecture is used to localize textual regions. The model was trained directly using images wherein pixel values were used as inputs and binary ground truth was used as label. The method was evaluated using ICDAR-2013 dataset and proved to be comparable to other feature-based methods. It could expedite research on text detection using deep-learning based approach in the future.