• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.8
• /
• pp.1823-1830
• /
• 2003

This paper presents a multiple face detector based on a robust pupil detection technique. The pupil detector uses active illumination that exploits the retro­reflectivity property of eyes to facilitate detection. The detection range of this method is appropriate for interactive desktop and kiosk applications. Once the location of the pupil candidates are computed, the candidates are filtered and grouped into pairs that correspond to faces using heuristic rules. To demonstrate the robustness of the face detection technique, a dual mode face tracker was developed, which is initialized with the most salient detected face. Recursive estimators are used to guarantee the stability of the process and combine the measurements from the multi­face detector and a feature correlation tracker. The estimated position of the face is used to control a pan­tilt servo mechanism in real­time, that moves the camera to keep the tracked face always centered in the image.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.11
• /
• pp.1519-1527
• /
• 2020

AGeneral radiation measuring devices have been developed in the form of spatial dose rate detection devices that measure dose rates to radioactive contaminant and 2D or 3D imaging devices for radioactive contamination information. Each of these radiation detection techniques has advantages. The advantages of both detection devices are necessary to minimize personal injury and rapid decontamination in the area of a radioactive accident. In this paper, we proposed a technique that can measure the dose rate and direction information about the radioactive pollutant source in real time using a detection sensor, a rotating body, and a directional shield for radioactive pollutant detection. The rotational-based detection device is configured to check the dose rate and direction using the location information of the rotator and measurement value. We proposed a measurement technique for vertical and horizontal directions through multiple holes. It was confirmed that the measurement error for direction information was less than 1% when detected in the horizontal direction.

• Restorative Dentistry and Endodontics
• /
• v.46 no.3
• /
• pp.43.1-43.9
• /
• 2021

Objectives: This study aimed to evaluate the detection rate of apical radiolucencies in 2-dimensional images using cone-beam computed tomography (CBCT) as the reference standard, and to determine which factors related to the apical radiolucencies and the teeth could influence its detection. Materials and Methods: The sample consisted of exams of patients who had panoramic (PAN) and/or periapical (PERI) radiography and CBCT. The exams were assessed by 2 oral radiologists and divided into PAN+CBCT (227 teeth-285 roots) and PERI+CBCT (94 teeth-115 roots). Radiographic images were evaluated for the presence of apical radiolucency, while CBCT images were assessed for presence, size, location, and involvement of the cortical bone (thinning, expansion, and destruction). Diagnostic values were obtained for PERI and PAN. Results: PERI and PAN presented high accuracy (0.83 and 0.77, respectively) and specificity (0.89 and 0.91, respectively), but low sensitivity, especially for PAN (0.40 vs. 0.65 of PERI). The size of the apical radiolucency was positively correlated with its detection in PERI and PAN (p < 0.001). For PAN, apical radiolucencies were 3.93 times more frequently detected when related to single-rooted teeth (p = 0.038). The other factors did not influence apical radiolucency detection (p > 0.05). Conclusions: PERI presents slightly better accuracy than PAN for the detection of apical radiolucency. The size is the only factor related to radiolucency that influences its detection, for both radiographic exams. For PAN, apical radiolucency is most often detected in single-rooted teeth.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.89-106
• /
• 2022

Recently, the amount of video data collected from smartphones, CCTVs, black boxes, and high-definition cameras has increased rapidly. According to the increasing video data, the requirements for analysis and utilization are increasing. Due to the lack of skilled manpower to analyze videos in many industries, machine learning and artificial intelligence are actively used to assist manpower. In this situation, the demand for various computer vision technologies such as object detection and tracking, action detection, emotion detection, and Re-ID also increased rapidly. However, the object detection and tracking technology has many difficulties that degrade performance, such as re-appearance after the object's departure from the video recording location, and occlusion. Accordingly, action and emotion detection models based on object detection and tracking models also have difficulties in extracting data for each object. In addition, deep learning architectures consist of various models suffer from performance degradation due to bottlenects and lack of optimization. In this study, we propose an video analysis system consists of YOLOv5 based DeepSORT object tracking model, SlowFast based action recognition model, Torchreid based Re-ID model, and AWS Rekognition which is emotion recognition service. Proposed model uses single-linkage hierarchical clustering based Re-ID and some processing method which maximize hardware throughput. It has higher accuracy than the performance of the re-identification model using simple metrics, near real-time processing performance, and prevents tracking failure due to object departure and re-emergence, occlusion, etc. By continuously linking the action and facial emotion detection results of each object to the same object, it is possible to efficiently analyze videos. The re-identification model extracts a feature vector from the bounding box of object image detected by the object tracking model for each frame, and applies the single-linkage hierarchical clustering from the past frame using the extracted feature vectors to identify the same object that failed to track. Through the above process, it is possible to re-track the same object that has failed to tracking in the case of re-appearance or occlusion after leaving the video location. As a result, action and facial emotion detection results of the newly recognized object due to the tracking fails can be linked to those of the object that appeared in the past. On the other hand, as a way to improve processing performance, we introduce Bounding Box Queue by Object and Feature Queue method that can reduce RAM memory requirements while maximizing GPU memory throughput. Also we introduce the IoF(Intersection over Face) algorithm that allows facial emotion recognized through AWS Rekognition to be linked with object tracking information. The academic significance of this study is that the two-stage re-identification model can have real-time performance even in a high-cost environment that performs action and facial emotion detection according to processing techniques without reducing the accuracy by using simple metrics to achieve real-time performance. The practical implication of this study is that in various industrial fields that require action and facial emotion detection but have many difficulties due to the fails in object tracking can analyze videos effectively through proposed model. Proposed model which has high accuracy of retrace and processing performance can be used in various fields such as intelligent monitoring, observation services and behavioral or psychological analysis services where the integration of tracking information and extracted metadata creates greate industrial and business value. In the future, in order to measure the object tracking performance more precisely, there is a need to conduct an experiment using the MOT Challenge dataset, which is data used by many international conferences. We will investigate the problem that the IoF algorithm cannot solve to develop an additional complementary algorithm. In addition, we plan to conduct additional research to apply this model to various fields' dataset related to intelligent video analysis.

• Journal of Korea Multimedia Society
• /
• v.16 no.10
• /
• pp.1147-1155
• /
• 2013

In this study, we designed and implemented a step detection analyzer that can compare and analyze the step detection rates and results among the step detection algorithms. The step detection analyzer converts 3-axes accelerometer data into continuous energy stream through SVM operation, shows the horizontal comparison among the step detection results for each step detection algorithms, and can make elemental detection analyses. For these processes, the step detection analyzer presents the continuous energy stream as energy waveform, checks the peak values and time location of the detected steps with step detection algorithms, and gives visual interface to get some possible causes in cases of step detection miss. It can also give the threshold graph for each algorithm to check the threshold value on missed cases directly and can help to get more appropriate threshold values or other adjustable parameters in step detection algorithm. This step detection analyzer can be applied efficiently on performance enhancement of step detection algorithm, on deciding an appropriate algorithm for a specific step counter system in the various step counter filed operations.

• Journal of Digital Convergence
• /
• v.14 no.5
• /
• pp.301-307
• /
• 2016

The location measurement technique of moving worker in dangerous areas, is necessary for safety in the mines, basements, warehouses, etc. There are various measurement techniques about moving node of position in a restricted environment. Trigonometric Method, one of measurement techniques, is commonly used because of its easiness. However, errors occur frequently when measuring distance and position due to radio interference and physical disability with measuring instruments. This paper proposed a method which is more accurate and shows reduced margin of error than existing trigonometric method by recalculating distance between Anchor and moving node with various measuring instruments. By adding Anchor when calculating distance and position of moving node's estimated point, suggested technique obtains at least 41% efficiency compared to existing method.

• Journal of KIISE
• /
• v.43 no.3
• /
• pp.389-397
• /
• 2016

In this paper, we propose a highway safety system that comprises a small number of central servers and smart mobile devices. To implement this system, we constructed a central server that collects GPS location information on cars, whose update messages are decreased via the car location estimation algorithm. The in-car mobile devices use the accelerometer sensors to detect hazardous situations; this information is updated to the central server that relays the information to the corresponding endangered cars via location-based unicast using LTE communication. To evaluate the proposed algorithm, we equipped a mobile device app on a real car and conducted real experiments in various environments such as city streets, rural areas, and highway roads. Furthermore, we conducted simulations to evaluate the propagation of danger information. Finally, we conducted simulated experiments to detect car collisions as well as exceptions, such as falling of the mobile device from the cradle.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.5
• /
• pp.84-92
• /
• 2020

This paper describes the estimation method of cable fault location in rocket motors using M-sequence (Maximal Length Sequence). In order to estimate the location of a cable fault, three methods have been usually used: TDR (Time Domain Reflectometry), FDR (Frequency Domain Reflectometry), and TFDR (Time-Frequency Domain Reflectometry). However, these methods suffer the disadvantage of requiring users to be close to a test field, which is dangerous. The estimation method of cable fault location using M-sequence is proposed to solve this problem. The proposed method can make use of DAS (Data Acquisition System). The experiments were three cases: damaged, open, and short. The RG-58 coaxial cable was used in the experiments. As a result, the proposed method has better performance than that of conventional methods such as TDR and TFDR.

• Journal of Korea Multimedia Society
• /
• v.12 no.1
• /
• pp.156-167
• /
• 2009

Computing environment is evolving rapidly to ubiquitous environment. It thus makes people access and utilize any resource on the move and makes location of objects rise to the rank of the most important information. However provision of location information is too restrictive in small environments such as a home or building. In this paper we propose a home entertainment system which supports user mobility based on sensor networks and distributed devices. The proposed system estimates a user location by using signal strength received from Bluetooth ad-hoc networks. This estimated location information can tell which device a user can take advantage of while moving. In order to prove possibility of user mobility support, the system is implemented on our UPnP A/V framework home entertainment system to provide content streaming and the result shows that a provision of service for mobile users could be possible in small ubiquitous computing environments.

• Composites Research
• /
• v.13 no.5
• /
• pp.1-9
• /
• 2000

An intensity-based optical fiber vibration sensor is applied to monitor the structural vibration and detect impact locations on a plate. Optical fiber vibration sensor is constructed by placing two cleaved fiber end, one of which is cantilevered in a hollow glass tube. The movement of the cantilevered section lags behind the rest of the sensor in response to an applied vibration and the amount of light coupled between the two fibers is thereby modulated. For vibration sensing, optical fiber vibration sensor is mounted on the carbon fiber composite beam and its response is investigated to free and forced vibration. In impact location detection, four optical fiber vibration sensors whose location is predetermined are placed at chosen positions and the different arrival times of impact-generated vibration signal are recorded by an FFT analyzer. Impact location can be calculated from these time delays. Experimental results show that optical fiber vibration sensor signals coincide with gap sensor in vibration sensing. The precise location of impact can be detected on an acrylate plate.