• Journal of Cadastre & Land InformatiX
• /
• v.51 no.1
• /
• pp.67-77
• /
• 2021

Spatial data is becoming very important as a medium that connects various data produced in smart cities, digital twins, autonomous driving, smart construction, and other applications. In addition, the rapid construction and update of spatial information is becoming a hot topic to satisfy the diverse needs of consumers in this field. This study developed a software prototype that can match the position of an image-based 3D building model produced without Ground Control Points using existing spatial data. As a result of applying this software to the test area, the 3D building model produced based on the image and the existing spatial data show a high positional matching rate, so that it can be widely used in applications requiring the latest 3D spatial data.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.20 no.4
• /
• pp.95-105
• /
• 2021

Road surface damage detection is essential for a comfortable driving environment and the prevention of safety accidents. Road management institutes are using automated technology-based inspection equipment and systems. As one of these automation technologies, a sensor to detect road surface damage plays an important role. For this purpose, several studies on sensors using deep learning have been conducted in recent years. Road images and label images are needed to develop such deep learning algorithms. On the other hand, considerable time and labor will be needed to secure label images. In this paper, the adversarial learning method, one of the semi-supervised learning techniques, was proposed to solve this problem. For its implementation, a lightweight deep neural network model was trained using 5,327 road images and 1,327 label images. After experimenting with 400 road images, a model with a mean intersection over a union of 80.54% and an F1 score of 77.85% was developed. Through this, a technology that can improve recognition performance by adding only road images was developed to learning without label images and is expected to be used as a technology for road surface management in the future.

• Journal of the Institute of Convergence Signal Processing
• /
• v.22 no.4
• /
• pp.171-178
• /
• 2021

Caravans are easily affected by external physical factors and often cause dangerous situations for passengers. Therefore, in order to secure the stability of the passenger, there is a need to develop a sway reduction device capable of preventing the sway phenomenon in advance. This paper aims to minimize the hitch angle between the tow vehicle and the caravan. Specifically, the initial instability of the caravan is detected through an IMU sensor mounted on each of the tow vehicle and the caravan, and a control value is calculated to reduce errors from the Hitch angle and Hitch yaw rate using a PID controller. Different braking torques are generated, distributed, and controlled on the left and right brakes of the caravan according to the calculated control value. It could be verified through the driving experiment that the hitch angle was decreased compared to the case where the performance of the sway reduction device was not controlled, and the transverse stability improvement rate was improved by 94.49% compared to before control.

• The Journal of Korea Robotics Society
• /
• v.17 no.2
• /
• pp.142-151
• /
• 2022

With the advent of the 4th industrial revolution, autonomous driving technology is being commercialized in various industries. However, research on autonomous driving so far has focused on platforms with wheel-type platform. Research on a tracked platform is at a relatively inadequate step. Since the tracked platform has a different driving and steering method from the wheel-type platform, the existing research cannot be applied as it is. Therefore, a path-tracking algorithm suitable for a tracked platform is required. In this paper, we studied a path-tracking algorithm for a tracked platform based on a GPS sensor. The existing Pure Pursuit algorithm was applied in consideration of the characteristics of the tracked platform. And to compensate for "Cutting Corner", which is a disadvantage of the existing Pure Pursuit algorithm, an algorithm that changes the LAD according to the curvature of the path was developed. In the existing pure pursuit algorithm that used a tracked platform to drive a path including a right-angle turn, the RMS path error in the straight section was 0.1034 m and the RMS error in the turning section was measured to be 0.2787 m. On the other hand, in the variable LAD algorithm, the RMS path error in the straight section was 0.0987 m, and the RMS path error in the turning section was measured to be 0.1396 m. In the turning section, the RMS path error was reduced by 48.8971%. The validity of the algorithm was verified by measuring the path error by tracking the path using a tracked robot platform.

• The Journal of Korea Robotics Society
• /
• v.18 no.3
• /
• pp.251-259
• /
• 2023

In the path traveling of differential-drive robots, the steering controller plays an important role in determining the path-following performance. When a robot with a pure-pursuit algorithm is used to continuously drive a right-angled driving path in an unstructured environment without turning in place, the robot cannot accurately follow the right-angled path and stops driving due to the ground and motor load caused by turning. In the case of pure-pursuit, only the current robot position and the steering angle to the current target path point are generated, and the steering component does not reflect the speed plan, which requires improvement for precise path following. In this study, we propose a driving algorithm for differentially driven robots that enables precise path following by planning the driving speed using the radius of curvature and fusing the planned speed with the steering angle of the existing pure-pursuit controller, similar to the Model Predict Control control that reflects speed planning. When speed planning is applied, the robot slows down before entering a right-angle path and returns to the input speed when leaving the right-angle path. The pure-pursuit controller then fuses the steering angle calculated at each path point with the accelerated and decelerated velocity to achieve more precise following of the orthogonal path.

• Journal of Broadcast Engineering
• /
• v.27 no.1
• /
• pp.20-30
• /
• 2022

Recently, as VR (Virtual Reality) technology has been in the spotlight, 360° panoramic images that can view lively VR contents are attracting a lot of attention. Image stitching technology is a major technology for producing 360° panorama images, and many studies are being actively conducted. Typical stitching algorithms are based on feature point-based image stitching. However, conventional feature point-based image stitching methods have a problem that stitching results are intensely affected by feature points. To solve this problem, deep learning-based image stitching technologies have recently been studied, but there are still many problems when there are few overlapping areas between images or large parallax. In addition, there is a limit to complete supervised learning because labeled ground-truth panorama images cannot be obtained in a real environment. Therefore, we produced three fisheye images with different camera centers and corresponding ground truth image through carla simulator that is widely used in the autonomous driving field. We propose image stitching model that creates a 360° panorama image with the produced fisheye image. The final experimental results are virtual datasets configured similar to the actual environment, verifying stitching results that are strong against various environments and large parallax.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.01a
• /
• pp.35-38
• /
• 2023

포트홀은 주행하는 자동차와 접촉이 이뤄지면 차체나 운전자에게 충격을 주고 제어를 잃게 하여 도로 위 안전을 위협할 수 있다. 포트홀의 검출을 위한 국내 동향으로는 진동을 이용한 방식과 신고시스템 이용한 방식과 영상 인식을 기반한 방식이 있다. 이 중 영상 인식 기반 방식은 보급이 쉽고 비용이 저렴하나, 컴퓨터 비전 알고리즘은 영상의 품질에 따라 정확도가 달라지는 문제가 있었다. 이를 보완하기 위해 영상 인식 기반의 딥러닝 모델을 사용한다. 따라서, 본 논문에서는 사전 학습된 딥러닝 모델의 정확도 향상을 위한 Feature Level Ensemble 기법을 제안한다. 제안된 기법은 사전 학습된 CNN 모델 중 Test 데이터의 정확도 기준 Top-3 모델을 선정하여 각 딥러닝 모델의 Feature Map을 Concatenate하고 이를 Fully-Connected(FC) Layer로 입력하여 구현한다. Feature Level Ensemble 기법이 적용된 딥러닝 모델은 평균 대비 3.76%의 정확도 향상을 보였으며, Top-1 모델인 ShuffleNet보다 0.94%의 정확도 향상을 보였다. 결론적으로 본 논문에서 제안된 기법은 사전 학습된 모델들을 이용하여 각 모델의 다양한 특징을 통해 기존 모델 대비 정확도의 향상을 이룰 수 있었다.

• Information Systems Review
• /
• v.18 no.4
• /
• pp.1-15
• /
• 2016

Electric cars are expected to increase quality of life by reducing air pollution and to contribute to economic growth by creating new businesses. However, electric car adoption has lagged and has not satisfied public expectation. One of the primary reasons for this outcome is the slow charging speed or inconvenience of charging a battery. Under the insufficient diffusion of electric cars, pushing business entities to construct charging facilities is undesirable for a policy maker to increase the adoption rate because of cost and management issues. This study adopts the design science methodology to interpret the problem of deploying electric car charging stations in the view of information systems. A trip planning algorithm is suggested on the basis of the theory of range anxiety. We investigate issues related to the current charging locations using data from drivers' car navigation devices. We also review its applicability to trip planning to obtain insights.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.5
• /
• pp.903-910
• /
• 2024

Complex semantic segmentation tasks are primarily performed in server environments equipped with high-performance graphics hardware such as GPUs and TPUs. This cloud-based AI inference method operates by transmitting processed results to the client. However, this approach is dependent on network communication and raises concerns about privacy infringement during the process of transmitting user data to servers. Therefore, this paper proposes a Generalized On-Device Framework for Semantic Segmentation that can operate in mobile environments with high accessibility to people. This framework supports various semantic segmentation models and enables direct inference in mobile environments through model conversion and efficient memory management techniques. It is expected that this research approach will enable effective execution of semantic segmentation algorithms even in resource-constrained situations such as IoT devices, autonomous vehicles, and industrial robots, which are not cloud computing environments. This is expected to contribute to the advancement of real-time image processing, privacy protection, and network-independent AI application fields.

• Journal of IKEEE
• /
• v.28 no.3
• /
• pp.296-302
• /
• 2024

In this paper, we propose a driver condition monitoring system using FMCW in-cabin radar, which is free from wearing inconvenience and privacy issues. Using 77GHz high-precision radar, the system detects changes in eye blinking patterns according to changes in the driving environment and the driver's condition using an adaptive multiple filtering algorithm, and accurately determines drowsy driving by measuring the number of eye blinks and the time it takes to open and close the eyes through the detected data. With the emergence of high-performance radars that are becoming more and more miniaturized, it is possible to embed them in the instrument panel or rearview mirror of the vehicle, and if the driver is judged to be drowsy, it can wake up the driver through an alarm or interlock with the vehicle's driving system to slow down and make an emergency stop to prevent accidents and promote driver safety.