• ETRI Journal
• /
• v.45 no.5
• /
• pp.795-810
• /
• 2023

Unmanned aerial vehicles (UAV) and ground vehicles (UGV) require advanced video analytics for various tasks, such as moving object detection and segmentation; this has led to increasing demands for these methods. We propose a zero-shot video object segmentation method specifically designed for UAV and UGV applications that focuses on the discovery of moving objects in challenging scenarios. This method employs a background memory model that enables training from sparse annotations along the time axis, utilizing temporal modeling of the background to detect moving objects effectively. The proposed method addresses the limitations of the existing state-of-the-art methods for detecting salient objects within images, regardless of their movements. In particular, our method achieved mean J and F values of 82.7 and 81.2 on the DAVIS'16, respectively. We also conducted extensive ablation studies that highlighted the contributions of various input compositions and combinations of datasets used for training. In future developments, we will integrate the proposed method with additional systems, such as tracking and obstacle avoidance functionalities.

• Journal of Korean Society on Water Environment
• /
• v.26 no.6
• /
• pp.954-962
• /
• 2010

The objective of this study was to develop automatic event detection algorithm for groundwater level rise. The groundwater level data and rainfall data in July and August at 37 locations nationwide were analyzed to develop the algorithm for groundwater level rise due to rainfall. In addition, the algorithm for groundwater level rise by ice melting and ground freezing was developed through the analysis of groundwater level data in January. The algorithm for groundwater level rise by rainfall was composed of three parts, including correlation between previous rainfall and groundwater level, simple linear regression analysis between previous rainfall and groundwater level, and diagnosis of groundwater level rise due to new rainfall. About 49% of the analyzed data was successfully simulated for groundwater level rise by rainfall. The algorithm for groundwater level rise due to ice melting and ground freezing included graphic analysis for groundwater level versus time (day), simple linear regression analysis for groundwater level versus time, and diagnosis of groundwater level rise by new ice melting and ground freezing. Around 37% of the analyzed data was successfully simulated for groundwater level rise due to ice melting and ground freezing. The algorithms from this study would help develop strategies for sustainable development and conservation of groundwater resources.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.23 no.3
• /
• pp.261-272
• /
• 2005

3D models in urban areas are essential for a variety of applications, such as virtual visualization, GIS, and mobile communications. LIDAR (Light Detection and Ranging) is a relatively new technology for obtaining Digital Terrain Models (DTM) of the earth's surface since manual 3D data reconstruction is very costly and time consuming. In this paper an approach to extract ground and non-ground points data from LIDAR data by using filtering is presented and the accuracy for generating DTM from ground points data is evaluated. Numerous filter algorithms have been developed to date. To determine the performance of filtering, we selected three filters which are based on the concepts for height difference, slope, and morphology, and also were applied two different data acquired from high raised apartments areas and low house areas. From the results it has been found that the accuracy for generating DTM from LIDAR data are 0.16 m and 0.59 m in high raised apartments areas and low house areas respectively. We expect that LIDAR data is used to generate the accurate DTM in urban areas.

• Journal of Radiation Protection and Research
• /
• v.43 no.3
• /
• pp.85-96
• /
• 2018

Background: A variety of inorganic scintillators have been developed and improved for use in radiation detection and measurement, and in situ gamma-ray spectrometry in the environment remains an important area in nuclear safety. In order to verify the feasibility of promising scintillators in an actual environment, a performance test is necessary to identify gamma-ray peaks and calculate the radioactivity from their net count rates in peaks. Materials and Methods: Among commercially available scintillators, $LaBr_3(Ce)$ scintillators have so far shown the highest energy resolution when detecting and identifying gamma-rays. However, the intrinsic background of this scintillator type affects efficient application to the environment with a relatively low count rate. An algorithm to subtract the intrinsic background was consequently developed, and the in situ calibration factor at 1 m above ground level was calculated from Monte Carlo simulation in order to determine the radioactivity from the measured net count rate. Results and Discussion: The radioactivity of six natural radionuclides in the environment was evaluated from in situ gamma-ray spectrometry using an $LaBr_3(Ce)$ detector. The results were then compared with those of a portable high purity Ge (HPGe) detector with in situ object counting system (ISOCS) software at the same sites. In addition, the radioactive cesium in the ground of Jeju Island, South Korea, was determined with the same assumption of the source distribution between measurements using two detectors. Conclusion: Good agreement between both detectors was achieved in the in situ gamma-ray spectrometry of natural as well as artificial radionuclides in the ground. This means that an $LaBr_3(Ce)$ detector can produce reliable and stable results of radioactivity in the ground from the measured energy spectrum of incident gamma-rays at 1 m above the ground.

• Journal of the Korean GEO-environmental Society
• /
• v.20 no.2
• /
• pp.49-58
• /
• 2019

A study on the analysis of underground ground condition using 3D-GPR exploration was carried out in this paper. The test bed was constructed similar to the field, and the detection analysis was carried out for each depth of cavity and underground burial. Through this, we were able to know the permittivity of the ground by inversion, and we could confirm the depth of detection for the joint by accurate calculation. We confirmed the signal waveforms in the cavity under the road through 3D-GPR exploration, analyzed more quantitatively in subjective and empirical analysis. The subsidence and depth of the subsurface settlement can be observed through 3D-GPR survey, and ground condition change after the ground reinforcement can be confirmed through the exploration section.

• Proceedings of the Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography Conference
• /
• 2003.10a
• /
• pp.403-409
• /
• 2003

LiDAR(Light Detection and Ranging) is considered to be a very accurate and useful tool for detection and reconstruction of ground objects. LiDAR data has information about both intensity and x,y,z position of the ground objects. LiDAR data can be collected from both first and last-return, which are called multi-return, with up to 5 different returns simultaneously. In this paper, an approach to reconstruct buildings in urban area using LiDAR multi-return data is presented. The reconstructed buildings are combined with DEM(Digital Elevation Model) produced from DSM(Digital Surface Model) in given area to implement 3D modeling. As a result, it is shown that buildings in urban area can be reconstructed and classified by the integration of the multi-return and intensity data of LiDAR.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.164-164
• /
• 2000

This paper proposes a moving window technique that extracts lanes and vehicles using the images captured by a CCD camera equipped inside an automobile in real time. For the purpose, first of all the optimal size of moving window is determined based upon speed of the vehicle, road curvature, and camera parameters. Within the moving windows that are dynamically changing, lanes and vehicles are extracted, and the vehicles within the driving lanes are classified as obstacles. Assuming highway driving, there are two sorts of image-objects within the driving lanes: one is ground mark to show the limit speed or some information for driving, and the other is the vehicle as an obstacle. Using characteristics of three-dimension objects, a neural network can be trained to distinguish the vehicle from ground mark. When it is recognized as an obstacle, the distance from the camera to the front vehicle can be calculated with the aids of database that keeps the models of automobiles on the highway. The correctness of this measurement is verified through the experiments comparing with the radar and laser sensor data.

• Journal of Institute of Control, Robotics and Systems
• /
• v.21 no.8
• /
• pp.773-780
• /
• 2015

Navigation in outdoor environments is a fundamental and challenging problem for unmanned ground vehicles. Detecting lane markings or boundaries on the road may be one of the solutions to make navigation easy. However, because of various environments and road conditions, a robust lane detection is difficult. In this paper, we propose a new approach for estimating virtual lanes on a traversable region. Estimating the virtual lanes consist of two steps: (i) we detect virtual road region through road model selection based on traversability at current frame and similarity between the interframe and (ii) we estimate virtual lane using the number of lane on the road and results of previous frame. To improve the detection performance and reduce the searching region of interests, we use a probability map representing the traversability of the outdoor terrain. In addition, by considering both current and previous frame simultaneously, the proposed method estimate more stable virtual lanes. We evaluate the performance of the proposed approach using real data in outdoor environments.

• Proceedings of the IEEK Conference
• /
• 2000.07b
• /
• pp.877-880
• /
• 2000

In this paper, we develop an algorithm of soccer image sequences mosaicing using reverse affine transform. The continuous mosaic images of soccer ground field allows the user/viewer to view a “wide picture” of the player’s actions The first step of our algorithm is to automatic detection and tracking player, ball and some lines such as center circle, sideline, penalty line and so on. For this purpose, we use the ground field extraction algorithm using color information and player and line detection algorithm using four P-rules and two L-rules. The second step is Affine transform to map the points from image to model coordinate using predefined and pre-detected four points. General Affine transformation has many holes in target image. In order to delete these holes, we use reverse Affine transform. We tested our method in real image sequence and the experimental results are given.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.786-789
• /
• 2003

The detection of sandstorms and industrial pollutants has been the emphasis of this study. Data obtained from meteorological satellites, NOAA and GMS, have been used for detailed analysis. MODIS and Landsat images are also used for the application of future KOMPSAT- 2. Verification of satellite observations has been made with air pollution data obtained by ground-level monitors. It was found that satellite measurements agree well with concentrations and variations of air pollutants measured on the ground, and that satellite technique is a very useful device for monitoring large-scale air pollution in East Asia. The quantitative analysis of satellite image data on air pollution is the goal in the future studies.