• Journal of Korea Multimedia Society
• /
• v.20 no.5
• /
• pp.769-781
• /
• 2017

Deep learning networks like Convolutional Neural Networks (CNNs) show successful performances in many computer vision applications such as image classification, object detection, and so on. For implementation of deep learning networks in embedded system with limited processing power and memory, deep learning network may need to be simplified. However, simplified deep learning network cannot learn every possible scene. One realistic strategy for embedded deep learning network is to construct a simplified deep learning network model optimized for the scene images of the installation place. Then, automatic training will be necessitated for commercialization. In this paper, as an intermediate step toward automatic training under fisheye camera environments, we study more precise human localization in fisheye images, and propose an accurate human localization method, Automatic Ground-Truth Labelling Method (AGTLM). AGTLM first localizes candidate human object bounding boxes by utilizing GoogLeNet-LSTM approach, and after reassurance process by GoogLeNet-based CNN network, finally refines them more correctly and precisely(tightly) by applying saliency object detection technique. The performance improvement of the proposed human localization method, AGTLM with respect to accuracy and tightness is shown through several experiments.

• Progress in Medical Physics
• /
• v.26 no.1
• /
• pp.52-58
• /
• 2015

The automated dental cavity detection program for a new concept intra-oral dental x-ray imaging device, an auxiliary diagnosis system, which is able to assist a dentist to identify dental caries in an early stage and to make an accurate diagnosis, was to be developed. The primary theory of the automatic dental cavity detection program is divided into two algorithms; one is an image segmentation skill to discriminate between a dental cavity and a normal tooth and the other is a computational method to analyze feature of an tooth image and take an advantage of it for detection of dental cavities. In the present study, it is, first, evaluated how accurately the DRLSE (Direct Regularized Level Set Evolution) method extracts demarcation surrounding the dental cavity. In order to evaluate the ability of the developed algorithm to automatically detect dental cavities, 7 tooth phantoms from incisor to molar were fabricated which contained a various form of cavities. Then, dental cavities in the tooth phantom images were analyzed with the developed algorithm. Except for two cavities whose contours were identified partially, the contours of 12 cavities were correctly discriminated by the automated dental caries detection program, which, consequently, proved the practical feasibility of the automatic dental lesion detection algorithm. However, an efficient and enhanced algorithm is required for its application to the actual dental diagnosis since shapes or conditions of the dental caries are different between individuals and complicated. In the future, the automatic dental cavity detection system will be improved adding pattern recognition or machine learning based algorithm which can deal with information of tooth status.

• Journal of Advanced Navigation Technology
• /
• v.21 no.6
• /
• pp.638-642
• /
• 2017

FOD is a generic term for substances with potential threats that can pose a fatal risk to aircraft. Therefore, FOD should be noted in all areas of the airport. Especially, the method of detecting and collecting FOD in runway and aircraft movements is very low efficiency and economical efficiency of airport operation, so it is essential to develop FOD automatic detection system suitable for domestic environment. As part of the aviation safety technology development project, the development of an automatic detection system for foreign matter in the moving area of the aircraft inside the airport is underway. In this paper, it is confirmed that EO / IR camera is used for detection of foreign objects at Taean Airfield of Hanseo University. EO camera is used during the day and IR camera is used at night.

• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.2
• /
• pp.83-90
• /
• 2015

Almost of current Automatic Incident Detection(AID) algorithms involve the vulnerability that detects the traffic accident in open road or in tunnel as the traffic jam not as the traffic accident. This paper proposes the improved accident detection algorithm to enhance the detection probability based on accident detection algorithms applied in open roads. The improved accident detection algorithm provides the preliminary judgment of potential accident by detecting the stopped object by Gaussian Mixture Model. Afterwards, it measures the detection area is divided into blocks so that the occupancy rate can be determined for each block. All experimental results of applying the new algorithm on a real incident was detected image without error.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.28 no.3
• /
• pp.271-280
• /
• 2004

Assessment of image registration for Pressure Sensitive Paint (PSP) was performed. A 16 bit camera and LED lamp were used with Uni-FIB paint (ISSI). Because of model displacement and deformation at 'wind-on' condition, a large error of the intensity ratio was induced between 'wind-on' and' wind-off images. To correct the error, many kinds of image registrations were tested. At first, control points were marked on the model surface to find the coefficients of polynomial transform functions between the 'wind-off' 'wind-on' images. The 2nd-order polynomial function was sufficient for representing the model displacement and deformation. An automatic detection scheme was introduced to find the exact coordinates of the control points. The present automatic detection algorithm showed more accurate and user-friendly than the manual detection algorithm. Since the coordinates of transformed pixel were not integer, five interpolation methods were applied to get the exact pixel intensity after transforming the 'wind-on' image. Among these methods, the cubic convolution interpolation scheme gave the best result.

• Journal of Broadcast Engineering
• /
• v.7 no.2
• /
• pp.107-113
• /
• 2002

In this paper, we propose a method and a software system for automatic summarization of baseball game videos. The proposed system pursues fast execution and high accuracy of summarization. To satisfy the requirement, the detection of important events in baseball video is performed through DC-based shot boundary detection algorithm and simple caption recognition method. Furthermore, the proposed system supports a hierarchical description so that users can browse and navigate videos in several levels of summarization. In this paper, we propose a method and a software system for automatic summarization of baseball game videos. The proposed system pursues fast execution and high accuracy of summarization. To satisfy the requirement, the detection of important events in baseball video is performed through DC-based shot boundary detection algorithm and simple caption recognition method. Furthermore, the proposed system supports a hierarchical description so that users can browse and navigate videos in several levels of summarization.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.482-486
• /
• 2009

Scale-invariant feature is an effective method for retrieving and classifying images. In this study, we analyze a scale-invariant planar curve features for developing 2D shapes. Scale-space filtering is used to determine contour structures on different scales. However, it is difficult to track significant points on different scales. In mathematics, curvature is considered to be fundamental feature of a planar curve. However, the curvature of a digitized planar curve depends on a scale. Therefore, automatic scale detection for curvature analysis is required for practical use. We propose a technique for achieving automatic scale detection based on difference of curvature. Once the curvature values are normalized with regard to the scale, we can calculate difference in the curvature values for different scales. Further, an appropriate scale and its position are detected simultaneously, thereby avoiding tracking problem. Appropriate scales and their positions can be detected with high accuracy. An advantage of the proposed method is that the detected significant points do not need to be located in the same contour. The validity of the proposed method is confirmed by experimental results.

• The KIPS Transactions:PartB
• /
• v.10B no.2
• /
• pp.127-134
• /
• 2003

This paper is intended to develop an algorithm for automatic detection of traffic accidents using image sequences. This algorithm is designed for detecting stopped vehicles traffic accidents, break down, illegal stop in the road shoulder - on the range of camera view. Virtual traps are set on accident-prone spots. We analyze the changes in gray levels of pixels on the virtual traps which represent the motion of vehicles on the corresponding spots. We verify the proposed algorithm by simulating some situations and checking if it detect them correctly.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.25 no.5
• /
• pp.39-44
• /
• 2011

Automatic fire alarm system is set up to automatically detect fire on buildings. Because of economic reasons, P-type receiver and a conventional type fire detector is normally used for automatic fire alarm system in Korea. Because early detection of fire is regarded as important, the need of finding technique of fire location increases. This paper is studied a method to improve a reliability and add a function of fire location detection on a conventional type fire detector and P-type receiver. Fire location is detected by a method that controller attached on the receiver and the detector is read with a time lag. A reliability of fire detection alarm system is improved with a method that false fire alarm is able to decrease using two different principle detector together. This paper is studied for basic data of improvement of low-cost addressable automatic fire alarm system.

• International Journal of Advanced Culture Technology
• /
• v.10 no.1
• /
• pp.230-235
• /
• 2022

This study is to select an optimal object detection algorithm for designing a self-checkout counter to improve the inconvenience of payment systems for products without existing barcodes. To this end, a performance comparison analysis of YOLO v2, Tiny YOLO v2, and the latest YOLO v5 among deep learning-based object detection algorithms was performed to derive results. In this paper, performance comparison was conducted by forming learning data as an example of 'donut' in a bakery store, and the performance result of YOLO v5 was the highest at 96.9% of mAP. Therefore, YOLO v5 was selected as the artificial intelligence object detection algorithm to be applied in this paper. As a result of performance analysis, when the optimal threshold was set for each donut, the precision and reproduction rate of all donuts exceeded 0.85, and the majority of donuts showed excellent recognition performance of 0.90 or more. We expect that the results of this paper will be helpful as the fundamental data for the development of an automatic payment system using AI self-service technology that is highly usable in the non-face-to-face era.