• Journal of IKEEE
• /
• v.26 no.3
• /
• pp.364-372
• /
• 2022

This paper proposes the design and implementation results for human and object classification systems utilizing frequency modulated continuous wave (FMCW) radar sensor. Such a system requires the process of radar sensor signal processing for multi-target detection and the process of deep learning for the classification of human and object. Since deep learning requires such a great amount of computation and data processing, the lightweight process is utmost essential. Therefore, binary neural network (BNN) structure was adopted, operating convolution neural network (CNN) computation in a binary condition. In addition, for the real-time operation, a hardware accelerator was implemented and verified via FPGA platform. Based on performance evaluation and verified results, it is confirmed that the accuracy for multi-target classification of 90.5%, reduced memory usage by 96.87% compared to CNN and the run time of 5ms are achieved.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.4
• /
• pp.39-46
• /
• 2021

In this paper, we propose a method for the motion recognition method required in the industrial field in mixed reality. In industrial sites, movements (grasping, lifting, and carrying) are required throughout the upper full-body, from trunk movements to arm movements. In this paper, we use a method composed of sensors and wearable devices that are not vision-based such as Kinect without using heavy motion capture equipment. We used two IMU sensors for the trunk and shoulder movement, and used Myo arm band for the arm movements. Real-time data coming from a total of 4 are fused to enable motion recognition for the entire upper body area. As an experimental method, a sensor was attached to the actual clothes, and objects were manipulated through synchronization. As a result, the method using the synchronization method has no errors in large and small operations. Finally, through the performance evaluation, the average result was 50 frames for single-handed operation on the HoloLens and 60 frames for both-handed operation.

• Journal of Broadcast Engineering
• /
• v.26 no.6
• /
• pp.738-747
• /
• 2021

Today, with the improvement of deep learning technology, computer vision areas such as image classification, object detection, object segmentation, and object tracking have shown remarkable improvements. Various applications such as intelligent surveillance, robots, Internet of Things, and autonomous vehicles in combination with deep learning technology are being applied to actual industries. Accordingly, the requirement of an efficient compression method for video data is necessary for machine consumption as well as for human consumption. In this paper, we propose an object-based compression of thermal infrared images for machine vision. The input image is divided into object and background parts based on the object detection results to achieve efficient image compression and high neural network performance. The separated images are encoded in different compression ratios. The experimental result shows that the proposed method has superior compression efficiency with a maximum BD-rate value of -19.83% to the whole image compression done with VVC.

• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.393-407
• /
• 2023

To assess structural condition in a non-destructive manner, computer vision-based structural health monitoring (SHM) has become a focus. Compared to traditional contact-type sensors, the advantages of computer vision-based measurement systems include lower installation costs and broader measurement areas. In this study, we propose a novel computer vision-based vibration measurement and coarse-to-fine damage assessment method for truss bridges. First, a deep learning model FairMOT is introduced to track the regions of interest (ROIs) that include joints to enhance the automation performance compared with traditional target tracking algorithms. To calculate the displacement of the tracked ROIs accurately, a normalized cross-correlation method is adopted to fine-tune the offset, while the Harris corner matching is utilized to correct the vibration displacement errors caused by the non-parallel between the truss plane and the image plane. Then, based on the advantages of the stochastic damage locating vector (SDLV) and Bayesian inference-based stochastic model updating (BI-SMU), they are combined to achieve the coarse-to-fine localization of the truss bridge's damaged elements. Finally, the severity quantification of the damaged components is performed by the BI-SMU. The experiment results show that the proposed method can accurately recognize the vibration displacement and evaluate the structural damage.

• Journal of Biomedical Engineering Research
• /
• v.44 no.4
• /
• pp.255-263
• /
• 2023

Clinical ultrasound (US) is a widely used imaging modality with various clinical applications. However, capturing a large field of view often requires specialized transducers which have limitations for specific clinical scenarios. Panoramic imaging offers an alternative approach by sequentially aligning image sections acquired from freehand sweeps using a standard transducer. To reconstruct a 3D volume from these 2D sections, an external device can be employed to track the transducer's motion accurately. However, the presence of optical or electrical interferences in a clinical setting often leads to incorrect measurements from such sensors. In this paper, we propose a deep learning (DL) framework that enables the prediction of scan trajectories using only US data, eliminating the need for an external tracking device. Our approach incorporates diverse data types, including correlation volume, optical flow, B-mode images, and rawer data (IQ data). We develop a DL network capable of effectively handling these data types and introduce an attention technique to emphasize crucial local areas for precise trajectory prediction. Through extensive experimentation, we demonstrate the superiority of our proposed method over other DL-based approaches in terms of long trajectory prediction performance. Our findings highlight the potential of employing DL techniques for trajectory estimation in clinical ultrasound, offering a promising alternative for panoramic imaging.

• Smart Media Journal
• /
• v.11 no.9
• /
• pp.64-74
• /
• 2022

A number of studies related to underwater robots are being conducted to utilize marine resources. However, unlike ordinary drones, underwater robots have a problem that it is not easy to locate because the medium is water, not air. The monitoring and positioning program of underwater robots, an existing study for identifying underwater locations, has difficulty in locating and monitoring in small spaces because it aims to be utilized in large spaces. Therefore, in this paper, we propose a three-dimensional positioning program for continuous monitoring and command delivery in small spaces. The proposed program consists of a multi-dimensional positioning monitoring function and a ability to control the path of travel through a three-dimensional screen so that the depth of the underwater robot can be identified. Through the performance evaluation, a robot underwater could be monitored and verified from various angles with a 3D screen, and an error within the assumed range was verified as the difference between the set path and the actual position is within 6.44 m on average.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.3
• /
• pp.117-122
• /
• 2023

Gaze estimation, which automatically recognizes where a user is currently staring, and object detection based on estimated gaze point, can be a more accurate and efficient way to understand human visual behavior. in this paper, we propose a method to detect the objects within the region of interest around the gaze point. Specifically, after estimating the 3D gaze point, a region of interest based on the estimated gaze point is created to ensure that object detection occurs only within the region of interest. In our experiments, we compared the performance of general object detection, and the proposed object detection based on region of interest, and found that the processing time per frame was 1.4ms and 1.1ms, respectively, indicating that the proposed method was faster in terms of processing speed.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.59 no.1
• /
• pp.44-54
• /
• 2023

In this study, the second-order Nomoto's nonlinear expansion model was implemented as a Tagaki-Sugeno fuzzy model based on the heading angular velocity to design the automatic steering system of a ship considering nonlinear elements. A Tagaki-Sugeno fuzzy PID controller was designed using the applied fuzzy membership functions from the Tagaki-Sugeno fuzzy model. The linear models and fuzzy membership functions of each operating point of a given nonlinear expansion model were simultaneously tuned using a genetic algorithm. It was confirmed that the implemented Tagaki-Sugeno fuzzy model could accurately describe the given nonlinear expansion model through the Zig-Zag experiment. The optimal parameters of the sub-PID controller for each operating point of the Tagaki-Sugeno fuzzy model were searched using a genetic algorithm. The evaluation function for searching the optimal parameters considered the route extension due to course deviation and the resistance component of the ship by steering. By adding a penalty function to the evaluation function, the performance of the automatic steering system of the ship could be evaluated to track the set course without overshooting when changing the course. It was confirmed that the sub-PID controller for each operating point followed the set course to minimize the evaluation function without overshoot when changing the course. The outputs of the tuned sub-PID controllers were combined in a weighted average method using the membership functions of the Tagaki-Sugeno fuzzy model. The proposed Tagaki-Sugeno fuzzy PID controller was applied to the second-order Nomoto's nonlinear expansion model. As a result of examining the transient response characteristics for the set course change, it was confirmed that the set course tracking was satisfactorily performed.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.12
• /
• pp.1794-1799
• /
• 2022

Recently, the increasing number of overloaded vehicles on the road poses a risk to traffic safety, such as falling objects, road damage, and chain collisions due to the abnormal weight distribution, and can cause great damage once an accident occurs. therefore we propose to build an object detection-based AI model to identify overloaded vehicles that cause such social problems. In addition, we present a simple yet effective method to construct an object detection model for the large-scale vehicle images. In particular, we utilize the large-scale of vehicle image sets provided by open AI-Hub, which include the overloaded vehicles. We inspected the specific features of sizes of vehicles and types of image sources, and pre-processed these images to train a deep learning-based object detection model. Also, we propose an integrated system for tracking the detected vehicles. Finally, we demonstrated that the detection performance of the overloaded vehicle was improved by about 23% compared to the one using raw data.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.10a
• /
• pp.407-409
• /
• 2021

Under the influence of the 4th industrial revolution, various technologies such as artificial intelligence and automation are being grafted into industrial sites, and accordingly, the importance of data processing is increasing. Digital images may generate noise due to various reasons, and may affect various systems such as image recognition and classification and object tracking. To compensate for these shortcomings, we propose an image restoration algorithm based on pattern information of non-noise pixels. According to the distribution of non-noise pixels inside the filtering mask, the proposed algorithm switched the filtering process by dividing the interpolation method into a pattern that can be applied, a pattern based on region division, and a randomly arranged pixel pattern. preserves and restores the image. The proposed algorithm showed superior performance compared to the existing impulse noise removal algorithm.