• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 추계종합학술대회 논문집(4)
• /
• pp.33-36
• /
• 2000

To create a more realistic soccer game derived from TV images, we are developing an image synthesis system that generates 3D image sequence from TV images. We propose the method for the team and the pose recognition of players in TV images. The representation includes camera calibration method, team recognition method and pose recognition method. To find the location of a player on the field, a field model is constructed and a player's field position is transformed by 4-feature points. To recognize the team information of players, we compute RGB mean values and standard deviations of a player in TV images. Finally, to recognize pose of a player, this system computes the velocity and the ratio of player(height/width). Experimental results are included to evaluate the performance of the team and the pose recognition.

• ETRI Journal
• /
• 제35권6호
• /
• pp.949-959
• /
• 2013

Since the recent launch of Microsoft Xbox Kinect, research on 3D human pose estimation has attracted a lot of attention in the computer vision community. Kinect shows impressive estimation accuracy and real-time performance on massive graphics processing unit hardware. In this paper, we focus on further reducing the computation complexity of the existing state-of-the-art method to make the real-time 3D human pose estimation functionality applicable to devices with lower computing power. As a result, we propose two simple approaches to speed up the random-forest-based human pose estimation method. In the original algorithm, the random forest classifier is applied to all pixels of the segmented human depth image. We first use a multi-scale approach to reduce the number of such calculations. Second, the complexity of the random forest classification itself is decreased by the proposed cascade approach. Experiment results for real data show that our method is effective and works in real time (30 fps) without any parallelization efforts.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권6호
• /
• pp.3182-3198
• /
• 2019

Vision-based Human Pose Estimation has been considered as one of challenging research subjects due to problems including confounding background clutter, diversity of human appearances and illumination changes in scenes. To tackle these problems, we propose to use a new multi-stage convolution machine for estimating human pose. To provide better heatmap prediction of body joints, the proposed machine repeatedly produces multiple predictions according to stages with receptive field large enough for learning the long-range spatial relationship. And stages are composed of various modules according to their strategic purposes. Pyramid stacking module and dilation module are used to handle problem of human pose at multiple scales. Their multi-scale information from different receptive fields are fused with concatenation, which can catch more contextual information from different features. And spatial and channel information of a given input are converted to gating factors by squeezing the feature maps to a single numeric value based on its importance in order to give each of the network channels different weights. Compared with other ConvNet-based architectures, we demonstrated that our proposed architecture achieved higher accuracy on experiments using standard benchmarks of LSP and MPII pose datasets.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.734-739
• /
• 2003

A new method of estimating the pose of a mobile-task robot is developed based upon an active calibration scheme. The utility of a mobile-task robot is widely recognized, which is formed by the serial connection of a mobile robot and a task robot. For the control of the mobile robot, an absolute position sensor is necessary. This paper proposes an active calibration scheme to estimate the pose of a mobile robot that carries a task robot on the top. The active calibration scheme is to estimate a pose of the mobile robot using the relative position/orientation to a known object whose location, size, and shape are known a priori. Through the homogeneous transformation, the absolute position/orientation of the camera is calculated and that is propagated to getting the pose of a mobile robot. With the experiments in the corridor, the proposed active calibration scheme is verified experimentally.

• 로봇학회논문지
• /
• 제15권4호
• /
• pp.350-356
• /
• 2020

Workers have been replaced by mobile manipulators for factory automation in recent years. One of the typical tasks for automation is that a mobile manipulator moves to a target location and picks and places an object on the worktable. However, due to the pose estimation error of the mobile platform, the robot cannot reach the exact target position, which prevents the manipulator from being able to accurately pick and place the object on the worktable. In this study, we developed an automatic alignment system using a low-cost camera mounted on the end-effector of a collaborative robot. Camera calibration and pose estimation methods were also proposed for the automatic alignment system. This algorithm uses a markerboard composed of markers to calibrate the camera and then precisely estimate the camera pose. Experimental results demonstrate that the mobile manipulator can perform successful pick and place tasks on various conditions.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제2권2호
• /
• pp.120-133
• /
• 2008

This paper presents a novel approach for facial motion tracking and facial expression cloning to create a realistic facial animation of a 3D avatar. The exact head pose estimation and facial expression tracking are critical issues that must be solved when developing vision-based computer animation. In this paper, we deal with these two problems. The proposed approach consists of two phases: dynamic head pose estimation and facial expression cloning. The dynamic head pose estimation can robustly estimate a 3D head pose from input video images. Given an initial reference template of a face image and the corresponding 3D head pose, the full head motion is recovered by projecting a cylindrical head model onto the face image. It is possible to recover the head pose regardless of light variations and self-occlusion by updating the template dynamically. In the phase of synthesizing the facial expression, the variations of the major facial feature points of the face images are tracked by using optical flow and the variations are retargeted to the 3D face model. At the same time, we exploit the RBF (Radial Basis Function) to deform the local area of the face model around the major feature points. Consequently, facial expression synthesis is done by directly tracking the variations of the major feature points and indirectly estimating the variations of the regional feature points. From the experiments, we can prove that the proposed vision-based facial expression cloning method automatically estimates the 3D head pose and produces realistic 3D facial expressions in real time.

• 전자공학회논문지B
• /
• 제33B권1호
• /
• pp.15-24
• /
• 1996

본 논문은 로봇 매니퓰레이터의 엔드 이펙터(end-effector)에 부착된 스테레오 카메라를 사용하여 움직이는 물체의 초기자세나 이동에 관한 정보가 미지인 3차원 물체의 파지(grasping)를 위해서, 로봇 매니퓰레이터의 자세(위치 및 방위)제어에 관한 새로운 비주얼 서보잉(visual servoing)을 제안한다. 로봇 매니퓰레이터의 현재의 자세를 목표자세에 잘 추적하기 위해서 본 논문에서는 카메라 자세에 대한 대상물체의 자세변화와 이미지상의 특정점 변화를 기술하는 관계식인 이미지 Jacobian을 미분변환을 이용하여 구했으며, 로봇 매니퓰레이터의 제어를 위해서는 간단한 PD제어기를 사용하였다. 마지막으로 다양한 컴퓨터 시뮬레이션을 통하여 제안한 수법의 유효성을 확인했다.

• 센서학회지
• /
• 제33권5호
• /
• pp.391-398
• /
• 2024

Accurate positioning is essential for unmanned underwater vehicle (UUV) operations, particularly for long-term survey missions. To reduce the inherent positioning errors from the inertial navigation systems of UUVs, or dead reckoning, underwater terrain observations from sonar sensors are typically exploited. Within the framework of pose-graph optimization, we can generate submaps of the seafloor and use them to add loop-closure constraints to the pose graph by determining the best match between the submaps. However, this approach results in error accumulation in long-term operations because the quality of local submaps depends on the dead reckoning. Hence, we can adopt external acoustic positioning systems, such as an ultrashort baseline (USBL), to add global constraints to the existing pose graph. We assume that the acoustic transponder is installed on a UUV and that the acoustic transceiver is equipped in an unmanned surface vehicle trailing the UUV to maintain an acoustic connection between the vehicles. We simulate the terrain and USBL measurements as well as evaluate the performance of the UUV's pose estimation via online pose-graph optimization.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2001년도 ICCAS
• /
• pp.63.4-63
• /
• 2001

An x-ray vision can be a unique method to monitor and analyze the motion of mechanical parts in real time which are invisible from outside. Our problem is to identify the pose, i.e. the position and orientation of an object from x-ray projection images. It is assumed here that the x-ray imaging conditions that include the relative coordinates of the x-ray source and the image plane are predetermined and the object geometry is known. In this situation, an x-ray image of an object at a given pose can be estimated computationally by using a priori known x-ray projection image model. It is based on the assumption that a pose of an object can be determined uniquely to a given x-ray projection image. Thus, once we have the numerical model of x-ray imaging process, x-ray image of the known object at any pose could be estimated ...

• International journal of advanced smart convergence
• /
• 제7권2호
• /
• pp.95-100
• /
• 2018

In this paper, we propose a binarized multi-scale module to accelerate the speed of the pose estimating deep neural network. Recently, deep learning is also used for fine-tuned tasks such as pose estimation. One of the best performing pose estimation methods is based on the usage of two neural networks where one computes the heat maps of the body parts and the other computes the part affinity fields between the body parts. However, the convolution filtering with a large kernel filter takes much time in this model. To accelerate the speed in this model, we propose to change the large kernel filters with binarized multi-scale modules. The large receptive field is captured by the multi-scale structure which also prevents the dropdown of the accuracy in the binarized module. The computation cost and number of parameters becomes small which results in increased speed performance.