Yong Woon Park

3D Reconstruction Framework for Multiple Remote Robots on Cloud System

This paper proposes a cloud-based framework that optimizes the three-dimensional (3D) reconstruction of multiple types of sensor data captured from multiple remote robots. A working environment using multiple remote robots requires massive amounts of data processing in real-time, which cannot be achieved using a single computer. In the proposed framework, reconstruction is carried out in cloud-based servers via distributed data processing. Consequently, users do not need to consider computing resources even when utilizing multiple remote robots. The sensors’ bulk data are transferred to a master server that divides the data and allocates the processing to a set of slave servers. Thus, the segmentation and reconstruction tasks are implemented in the slave servers. The reconstructed 3D space is created by fusing all the results in a visualization server, and the results are saved in a database that users can access and visualize in real-time. The results of the experiments conducted verify that the proposed system is capable of providing real-time 3D scenes of the surroundings of remote robots.

Motion-based skin region of interest detection with a real-time connected component labeling algorithm

This paper presents a motion-based skin Region of Interest (ROI) detection method using a real-time connected component labeling algorithm to provide real-time and adaptive skin ROI detection in video images. Skin pixel segmentation in video images is a pre-processing step for face and hand gesture recognition, and motion is a cue for detecting foreground objects. We define skin ROIs as pixels of skin-like color where motion takes place. In the skin color estimation phase, RGB color histograms are utilized to define the skin color distribution and specify the threshold to segment skin-like regions. A parallel computed connected component labeling algorithm is also proposed to group the segmentation results into several clusters. If a cluster covers any motion pixel, this cluster is identified as a skin ROI. The method’s results for real images are shown, and its speed is evaluated for various parameters. This technology is compatible with monitoring systems, scene understanding, and natural user interfaces.

Federated-filter-based unmanned ground vehicle localization using 3D range registration with digital elevation model in outdoor environments

An integrated GPS/INS does not guarantee localization robustness in outdoor environments, because GPS is vulnerable to external disturbances. However, a digital elevation model (DEM) contains 3D data on the terrain over a specified area and hence can provide in-depth localization information during GPS blockage. This paper proposes federated-filter-based localization using three-dimensional (3D) range registration with a DEM. A no-reset-feedback method is used and a 3D LIDAR sensor, magnetic compass, and odometer are used to correct INS errors in GPS blockage. For 3D range registration with DEM, this paper presents a framework based on the weighted registration scheme of two transformations, pairwise registration and registration with DEM, with the INS position and attitude information. The transformation is first determined by comparing the results of two registration methods with the INS position and is then modified to replace the orientation result of 3D registration with the INS attitude. A multilayered DEM approach using the height of the integrated system is also used to constrain the search range of DEM into three layers near the current unmanned ground vehicle (UGV) position when the corresponding point is searched for in the DEM. Experimental results show that the proposed localization algorithm can greatly enhance the robustness and accuracy of UGV localization in outdoor environments.

Simulation framework of ubiquitous network environments for designing diverse network robots

Abstract Smart homes provide residents with services that offer convenience using sensor networks and a variety of ubiquitous instruments. Network robots based on such networks can perform direct services for these residents. Information from various ubiquitous instruments and sensors located in smart homes is shared with network robots. These robots effectively help residents in their daily routine by accessing this information. However, the development of network robots in an actual environment requires significant time, space, labor, and money. A network robot that has not been fully developed may cause physical damage in unexpected situations. In this paper, we propose a framework that allows the design and simulation of network robot avatars and a variety of smart homes in a virtual environment to address the above problems. This framework activates a network robot avatar based on information obtained from various sensors mounted in the smart home; these sensors identify the daily routine of the human avatar residing in the smart home. Algorithms that include reinforcement learning and action planning are integrated to enable the network robot avatar to serve the human avatar. Further, this paper develops a network robot simulator to verify whether the network robot functions effectively using the framework.

Robust Corner Detection Based on Image Structure

In this paper, we propose a robust corner detection method to improve both detection rate and localization accuracy by modifying the structure tensor-based corner detection method in two ways. First, we introduce a connected component analysis (CCA) method for constructing a CCA structure tensor in order to make the structure tensor adaptive to the structure of the image. Second, the normalized cross-correlation (NCC) method is applied for false corner rejection with the observation that the patch of a true corner has a distinctive characteristic compared with connected neighboring patches. The proposed method is compared with several corner detection methods over a number of images. Experimental results show that the proposed method shows better performance in terms of both detection rate and localization accuracy.

LIDAR Simulation Method for Low-Cost Repetitive Validation

Developments in light detection and ranging (LIDAR) have enabled its application in unmanned automotive technology, and various methods using LIDAR are now being proposed. However, it is more difficult to obtain a ground truth dataset to evaluate the performance of algorithms that use a quantity of three-dimensional (3D) points as compared to those that require only 2D images. This paper describes an approach to creating a ground truth dataset for verifying a variety of algorithms by recording the data on detected objects through simulation in virtual space. This approach is able to verify the performance of algorithms in a variety of environments with less cost than the use of actual LIDAR.

Performance Analysis of Software Architectures with Real-time Kernel Patches for the Rescue Robot

This paper introduces software architectures of the Rescue Robot that lifts and moves wounded soldiers in battle fields or hospitals. Since the Rescue Robot interacts with humans, the robot needs a real-time operation for human safety. Moreover, it is necessary that a large number of calculations and various sensing data to move the soldiers without any damages. One of challenging problems of designing the robot software architecture is that a large number of calculations and data would disturb to decrease a response time taken until robot reacting from users commands or emergency sensing data. The Rescue Robot uses Linux with real-time kernel patches as a real-time operating system and EtherCAT as a real-time network. We test three kinds of software architectures, Linux, Xenomai, and RTAI, for the Rescue Robots about the time taken until output commands are sent to joint motors just after the sensing data or input commands are delivered. Furthermore, we suggest a method to handle urgent commands or data quickly with a slightly modified kernel and a kernel module that reduces the time dramatically. This paper also analysis the difference of the real-time kernel patches.

Fast point cloud segmentation based on flood-fill algorithm

With the aim of providing a fast and effective segmentation method based on the flood-fill algorithm, in this study, we propose a new approach to segment a 3D point cloud gained by a 3D multi-channel laser range sensor into different objects. First, we divide the point cloud into two groups: ground and nonground points. Next, we segment clusters in each scanline dataset from the group of nonground points. Each scanline cluster is joined with other scanline clusters using the flood-fill algorithm. In this manner, each group of scanline clusters represents an object in the 3D environment. Finally, we obtain each object separately. Experiments show that our method has the ability to segment objects accurately and in real time.

Design of hand gesture interaction framework on clouds for multiple users

To develop cooperative content based on the hand gestures of multiple users, typical frameworks must be separately used for the communication among computers and hand gesture recognition. In this paper, we propose a framework that enables users who are far apart to interact using hand gestures in the same virtual environment. This approach makes possible remote cooperative education content. We present the design of a server and client, as well as techniques for respectively managing multiple users, controlling crashes among multiple users, and managing gesture data for activating the proposed framework on clouds. To verify that the framework enables multiple users to interact, we developed a virtual chemical experiment based on the proposed design and performed the simulation with it. The framework can be used for a variety of educational content using interactions among multiple users.

Development of simulator for invoked reality environmental design

Invoked Reality (IR) research uses a variety of sensors and projection equipment for Natural User Interface (NUI) interaction. Such equipment requires an appropriate action region, which needs to be considered when implementing an IR environment. Accordingly, the implementation of an IR environment is problematic because of the arrangement of this equipment. The research presented in this paper overcomes this difficulty by proposing the “IR Simulator.” This simulator facilitates the design of and experimentation within the IR environment in a virtual environment. IR Simulator comprises an “experimental space editor” and “simulation manager.” IR Simulator enables design in diverse kinds of IR environments using its GUI-based experimental space editor. It also provides interaction with external IR libraries through the external IR library-interworking interface. Moreover, it enables simulation on the library connected on the basis of the virtual sensor data generated by the implementation of the proposed simulator. In accordance with experiments using the proposed simulator, a variety of IR environments could be generated and an arm motion recognition library could be simulated. Furthermore, the same environment generated by the simulator was easily implemented in the actual environment on the basis of the design results from the simulator. The experiment on the arm motion recognition library used in the simulator in the real environment was similar to the results obtained by the simulation.