• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.335-337
• /
• 2024

This paper presents a new open-world object detection method emphasizing uncertainty representation in machine learning models. The focus is on adapting to real-world uncertainties, incrementally updating the model's knowledge repository for dynamic scenarios. Applications like autonomous vehicles benefit from improved multi-class classification accuracy. The paper reviews challenges in existing methodologies, stressing the need for universal detectors capable of handling unknown classes. Future directions propose collaboration, integration of language models, to improve the adaptability and applicability of open-world object detection.

• Journal of Institute of Control, Robotics and Systems
• /
• v.8 no.6
• /
• pp.477-483
• /
• 2002

This paper deals with a workspace analysis of multi-legged walking robots in velocity domain(velocity workspace analysis). Noting that when robots are holding the same object in multiple cooperating robotic arm system the kinematic structure of the system is basically the same with that of a multi-legged walking robot standing on the ground, we invented a way ot applying the technique for multiple arm system to multi-legged walking robot. An important definition of reaction velocity is made and the bounds of velocities achievable by the moving body with multi-legs is derived from the given bounds on the capabilities of actuators of each legs through Jacobian matrix for given robot configuration. After some assumption of hard-foot-condition is adopted as a contact model between feet of robot and the ground, visualization process for the velocity workspace is proposed. Also, a series of application examples will be presented including continuous walking gaits as well as several different stationary posture of legged walking robots, which validate the usefulness of the proposed technique.

• Transactions of Materials Processing
• /
• v.19 no.5
• /
• pp.320-324
• /
• 2010

Multi-Layer Ceramic Circuit(MLCC) in the face of thousands of fine pitch multi hole is processed. However, the fine pitch multi hole has a size of only a few micrometers. Therefore, in order to curtail the measurement time and reduce error, the image processing measurement method is required. So, we proposed an image processing measurement algorithm which is required to accurately measure the fine pitch multi hole. The proposed algorithm gets image of the fine pitch multi hole, extracts object from the image by morphological process, and extracts the parameters of its position and feature by edge detecting process. In addition, we have used the sub-pixel algorithm to improve accuracy. As a result, the proposed algorithm shows 97% test-retest measurement reliability within 2 ${\mu}m$. We found that the algorithm was wellsuited for measuring the fine pitch multi hole.

• Information Systems Review
• /
• v.1 no.1
• /
• pp.19-33
• /
• 1999

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2013.05a
• /
• pp.365-366
• /
• 2013

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• v.23 no.7
• /
• pp.513-519
• /
• 2011

The object of this study is to investigate the performance characteristics of refrigerators using a single-circuit multi cycle and a bypass two-circuit multi cycle. Each refrigeration cycle was tested by varying secondary fluid mass flow rate and temperature. Based on the experimental data, the optimum refrigerant charge was 48 g and the COP at the optimum secondary fluid mass flow rate was 1.53 for the single-circuit multi cycle. For freezer(F)-only mode, both the single-circuit multi cycle and the bypass two-circuit multi cycle were operated at overcharge conditions, resulting in an increase of the secondary fluid mass flow rate. The maximum COPs of the single-circuit multi cycle and the bypass two-circuit multi cycle were 1.22 and 1.35, respectively. The COP increased by 10.7% with the application of the bypass two-circuit multi cycle.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.29 no.2 s.233
• /
• pp.188-200
• /
• 2005

Control of a multi-fingered robotic hand is usually based on the theoretical analysis for kinematics and dynamics of fingers and of object. However, the implementation of such analyses to robotic hands is difficult because of errors and uncertainties in the real situations. This article presents the control method for estimating the kinematic constraint of an unknown object by active sensing. The experimental system has a two-fingered robotic hand suspended vertically for manipulation in the vertical plane. The fingers with three degrees-of-freedom are driven by wires directly connected to voice-coil motors without reduction gears. The fingers are equipped with three-axis force sensors and with dynamic tactile sensors that detect slippage between the fingertip surfaces and the object. In order to make an accurate estimation for the kinematic constraint of the unknown object, i.e. the constraint direction and the constraint center, four kinds of the active sensing and feedback control algorithm were developed: two position-based algorithms and two force-based algorithms. Furthermore, the compound and effective algorithm was also developed by combining two algorithms. Force sensors are mainly used to adapt errors and uncertainties encountered during the constraint estimation. Several experimental results involving the motion of lifting a finger off an unknown object are presented.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.5
• /
• pp.2141-2155
• /
• 2020

In order to meet the requirements of background change, illumination variation, moving shadow interference and high accuracy in object detection of moving camera, and strive for real-time and high efficiency, this paper presents an object detection algorithm based on sparse approximation recursion and sparse coding migration in subspace. First, low-rank sparse decomposition is used to reduce the dimension of the data. Combining with dictionary sparse representation, the computational model is established by the recursive formula of sparse approximation with the video sequences taken as subspace sets. And the moving object is calculated by the background difference method, which effectively reduces the computational complexity and running time. According to the idea of sparse coding migration, the above operations are carried out in the down-sampling space to further reduce the requirements of computational complexity and memory storage, and this will be adapt to multi-scale target objects and overcome the impact of large anomaly areas. Finally, experiments are carried out on VDAO datasets containing 59 sets of videos. The experimental results show that the algorithm can detect moving object effectively in the moving camera with uniform speed, not only in terms of low computational complexity but also in terms of low storage requirements, so that our proposed algorithm is suitable for detection systems with high real-time requirements.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2002.10a
• /
• pp.219-224
• /
• 2002

This paper aims to derive a mathematical model of the dynamics of handling tasks in robot finger which stable grasping and manipulates a rigid object with some dexterity. Firstly, a set of differential equation describing dynamics of the manipulators and object together with geometric constraint of tight area-contacts is formulated by Lagrange's equation. Secondly, problems of controlling both the internal force and the rotation angle of the grasped object under the constraints of area-contacts of tight area-contacts are discussed. The effect of geometric constraints of area-contacts on motion of the overall system is analyzed and a method of computer simulation for overall system of differential-algebraic equations is presented. Thirdly, simulation results are shown and the effects of geometric constraints of area-contact is discussed. Finally, it is shown that even in the simplest case of dual single D.O.F manipulators there exists a sensory feedback from sensing data of the rotational angle of the object to command inputs to joint actuators and this feedback connection from sensing to action eventually realizes secure grasping of the object, provided that the object is of rectangular shape and motion is confined to a horizontal plane.

• Journal of Digital Contents Society
• /
• v.8 no.1
• /
• pp.85-91
• /
• 2007

This paper proposes a new hybrid anti-aliasing method for reducing aliasing appearing on an implicit surface using the point sampling. The hybrid anti-aliasing method is a method that finds differences in the values of pixel shown in Z-buffers for an implicit surface and thereby uses each of three types of Z-buffer. After determining the level of differences, it expresses a 3 dimensional object by using a multi Z-buffer if the level is severe, a double Z-buffer for a middle level, and only the original Z-buffer for a negligible difference. In comparison with the existing method in which multi Z-buffers have been entirely used for enhancing the anti-aliasing effect, the hybrid anti-aliasing method is an efficient method demonstrating an effect similar to the one using a multi Z-buffer while reducing the number of Z-buffers to be used.