• Korean Journal of Radiology
• /
• v.22 no.4
• /
• pp.612-623
• /
• 2021

Objective: To evaluate the diagnostic performance of a deep learning algorithm for the automated detection of developmental dysplasia of the hip (DDH) on anteroposterior (AP) radiographs. Materials and Methods: Of 2601 hip AP radiographs, 5076 cropped unilateral hip joint images were used to construct a dataset that was further divided into training (80%), validation (10%), or test sets (10%). Three radiologists were asked to label the hip images as normal or DDH. To investigate the diagnostic performance of the deep learning algorithm, we calculated the receiver operating characteristics (ROC), precision-recall curve (PRC) plots, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) and compared them with the performance of radiologists with different levels of experience. Results: The area under the ROC plot generated by the deep learning algorithm and radiologists was 0.988 and 0.988-0.919, respectively. The area under the PRC plot generated by the deep learning algorithm and radiologists was 0.973 and 0.618-0.958, respectively. The sensitivity, specificity, PPV, and NPV of the proposed deep learning algorithm were 98.0, 98.1, 84.5, and 99.8%, respectively. There was no significant difference in the diagnosis of DDH by the algorithm and the radiologist with experience in pediatric radiology (p = 0.180). However, the proposed model showed higher sensitivity, specificity, and PPV, compared to the radiologist without experience in pediatric radiology (p < 0.001). Conclusion: The proposed deep learning algorithm provided an accurate diagnosis of DDH on hip radiographs, which was comparable to the diagnosis by an experienced radiologist.

• Smart Structures and Systems
• /
• v.12 no.2
• /
• pp.181-207
• /
• 2013

In this paper, a statistical reference-free real-time damage detection methodology is proposed for detecting joint and member damage of truss bridge structures. For the statistical damage sensitive index (DSI), wavelet packet decomposition (WPD) in conjunction with the log likelihood ratio was suggested. A sensitivity test for selecting a wavelet packet that is most sensitive to damage level was conducted and determination of the level of decomposition was also described. Advantages of the proposed method for applications to real-time health monitoring systems were demonstrated by using the log likelihood ratios instead of likelihood ratios. A laboratory truss bridge structure instrumented with accelerometers and a shaker was used for experimental verification tests of the proposed methodology. The statistical reference-free real-time damage detection algorithm was successfully implemented and verified by detecting three damage types frequently observed in truss bridge structures - such as loss of bolts, loosening of bolts at multiple locations, sectional loss of members - without reference signals from pristine structure. The DSI based on WPD and the log likelihood ratio showed consistent and reliable results under different damage scenarios.

• Journal of the Korean Society of Industry Convergence
• /
• v.22 no.3
• /
• pp.315-324
• /
• 2019

This paper present a novel 6-axis robotic base platform with force/moment sensing. The robotic base platform is made up of six loadcells connecting the moving plate to the fixed plate by spherical joints at the both ends of loadcells. The statics relation is derived, the robotic base platform prototype and the loadcell measurement system are developed. The force/moment calibrations in joint and Cartesian spaces are performed. The algorithm to detect external force applied at a working robot is derived, and using a 6-DOF robot mounted on the robotic base platform, force/moment measurement experiments have been performed.

• Journal of Korea Multimedia Society
• /
• v.22 no.10
• /
• pp.1142-1148
• /
• 2019

Reconstruction of the three-dimensional (3D) environment is a key aspect of augmented reality and augmented virtuality, which utilize and incorporate a user's surroundings. Such reconstruction can be easily realized by employing a Kinect device. However, multiple Kinect devices are required for enhancing the reconstruction density and for spatial expansion. While employing multiple Kinect devices, they must be calibrated with respect to each other in advance, and a marker is often used for this purpose. However, a marker needs to be placed at each calibration, and the result of marker detection significantly affects the calibration accuracy. Therefore, a user-friendly, efficient, accurate, and marker-less method for calibrating multiple Kinect devices is proposed in this study. The proposed method includes a joint tracking algorithm for approximate calibration, and the obtained result is further refined by applying the iterative closest point algorithm. Experimental results indicate that the proposed method is a convenient alternative to conventional marker-based methods for calibrating multiple Kinect devices. Hence, the proposed method can be incorporated in various applications of augmented reality and augmented virtuality that require 3D environment reconstruction by employing multiple Kinect devices.

• ETRI Journal
• /
• v.41 no.4
• /
• pp.494-505
• /
• 2019

In this paper, we propose a new framework for detecting the unauthorized dumping of garbage in real-world surveillance camera. Although several action/behavior recognition methods have been investigated, these studies are hardly applicable to real-world scenarios because they are mainly focused on well-refined datasets. Because the dumping actions in the real-world take a variety of forms, building a new method to disclose the actions instead of exploiting previous approaches is a better strategy. We detected the dumping action by the change in relation between a person and the object being held by them. To find the person-held object of indefinite form, we used a background subtraction algorithm and human joint estimation. The person-held object was then tracked and the relation model between the joints and objects was built. Finally, the dumping action was detected through the voting-based decision module. In the experiments, we show the effectiveness of the proposed method by testing on real-world videos containing various dumping actions. In addition, the proposed framework is implemented in a real-time monitoring system through a fast online algorithm.

• Structural Engineering and Mechanics
• /
• v.71 no.1
• /
• pp.57-63
• /
• 2019

During the last two decades, much joint research regarding vibration based methods has been done, leading to developing various algorithms and techniques. These algorithms and techniques can be divided into modal methods and signal methods. Although modal methods have been widely used for health monitoring and damage detection, signal methods due to higher efficiency have received considerable attention in various fields, including aerospace, mechanical and civil engineering. Signal-based methods are derived directly from the recorded responses through signal processing algorithms to detect damage. According to different signal processing techniques, signal-based methods can be divided into three categories including time domain methods, frequency domain methods, and time-frequency domain methods. The frequency domain methods are well-known and interest in using them has increased in recent years. To determine dynamic behaviours, to identify systems and to detect damages of bridges, different methods and algorithms have been proposed by researchers. In this study, a new algorithm to detect seismic damage in the bridge's piers is suggested. To evaluate the algorithm, an analytical model of a bridge with simple spans is used. Based on the algorithm, before and after damage, the bridge is excited by a sine force, and the piers' responses are measured. The dynamic specifications of the bridge are extracted by Power Spectral Density function. In addition, the Least Square Method is used to detect damage in the bridge's piers. The results indicate that the proposed algorithm can identify the seismic damage effectively. The algorithm is output-only method and measuring the excitation force is not needed. Moreover, the proposed approach does not need numerical models.

• Korean Journal of Applied Biomechanics
• /
• v.22 no.1
• /
• pp.123-129
• /
• 2012

Simple camera-based system for evaluation of lower limb alignment as a part of an automated cycling fitting system was developed and verified in this study. Developed imaging system can evaluate lower limb alignment quantitatively during pedaling using a general camcorder and single marker attached on the knee. Threshold-based marker detection algorithm was proposed in this study. Experiment was carried out to compare the trajectory data from marker detection algorithm of the developed imaging system with the trajectory data from 3-D motion capture system. Results showed that average error between trajectories was 2.33 mm (0.92 %) in the vertical direction and 0.62 mm (1.86 %) in the medio-lateral direction. There existed significant correlation between two measured values (r=0.9996 in the vertical direction and r=0.9975 in the medio-lateral direction). It can be concluded that developed imaging system be applied to evaluate lower limb alignment which is an important factor for dynamic bicycle fitting.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.15 no.4
• /
• pp.851-859
• /
• 2011

In this paper, we propose the method that automatically extracts the silhouette and the joints of consecutive input image, and track joints to trace object for interaction between human and computer. Also the proposed method presents the action of human being to map human body using joints. To implement the algorithm, we model human body using 14 joints to refer to body size. The proposed method converts RGB color image acquired through a single camera to hue, saturation, value images and extracts body's silhouette using the difference between the background and input. Then we automatically extracts joints using the corner points of the extracted silhouette and the data of body's model. The motion of object is tracted by applying block-matching method to areas around joints among all image and the human's motion is mapped using positions of joints. The proposed method is applied to the test videos and the result shows that the proposed method automatically extracts joints and effectively maps human body by the detected joints. Also the human's action is aptly expressed to reflect locations of the joints

• The Journal of Korea Robotics Society
• /
• v.7 no.4
• /
• pp.259-266
• /
• 2012

In these days, researches about underwater robots have been actively in progress for the purposes of ocean detection and resource exploration. Unlike general underwater robots such as ROV(Remotely Operated Vehicle) and AUV(Autonomous Underwater Vehicle) which have propellers, an articulated underwater robot which is called Crabster has been being developed in KORDI(Korea Ocean Research & Development Institute) with many cooperation organizations since 2010. The robot is expected to be able to walk and swim under the sea with its legs. Among many researching fields of this project, we are focusing on a swimming section. In order to find effective swimming locomotion for the robot, we approached this subject in terms of Biomimetics. As a model of optimized swimming organism in nature, diving beetles were chosen. In the paper, swimming motions of diving beetles were analyzed in viewpoint of robotics for applying them into the swimming motion of the robot. After modeling the kinematics of diving beetle through robotics engineering technique, we obtained swimming patterns of the one of living diving beetles, and then compared them with calculated optimal swimming patterns of a robot leg. As the first trial to compare the locomotion data of legs of the diving beetle with a robot leg, we have sorted two representative swimming patterns such as forwarding and turning. Experimental environment has been set up to get the motion data of diving beetles. The experimental equipment consists of a transparent aquarium and a high speed camera. Various swimming motions of diving beetles were recorded with the camera. After classifying swimming patterns of the diving beetle, we can get angular data of each joint on hind legs by image processing software, Image J. The data were applied to an optimized algorithm for swimming of a robot leg which was designed by robotics engineering technique. Through this procedure, simulated results which show trajectories of a robot leg were compared with trajectories of a leg of a diving beetle in desired directions. As a result, we confirmed considerable similarity in the result of trajectory and joint angles comparison.

• Journal of the Optical Society of Korea
• /
• v.20 no.1
• /
• pp.22-28
• /
• 2016

The monitoring and measurement of the motion of a human joint is very important in screening for degenerative brain diseases and tracking the rehabilitation process. Since there are various medical fields to benefit from angular motion measurement, the necessity for monitoring of human joint movement is increasing. In this study, the optical sensor is composed of a light emission unit with a red LED and an optical fiber, and a reception unit with an arrangement of three photodiodes. The angular detection range was widened with the use of multiple photodiodes and the developed algorithm. The result will be useful for designing an effective angular sensor with low cost and small size.