• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.7
• /
• pp.279-286
• /
• 2021

Recently, We are carrying out a policy of physical distancing of at least 1m from each other to prevent the spreading of COVID-19 disease in public places. In this paper, we propose a method for measuring distances between people in real time and an automation system that recognizes objects that are within 1 meter of each other from stereo images acquired by drones or CCTVs according to the estimated distance. A problem with existing methods used to estimate distances between multiple objects is that they do not obtain three-dimensional information of objects using only one CCTV. his is because three-dimensional information is necessary to measure distances between people when they are right next to each other or overlap in two dimensional image. Furthermore, they use only the Bounding Box information to obtain the exact coordinates of human existence. Therefore, in this paper, to obtain the exact two-dimensional coordinate value in which a person exists, we extract a person's key point to detect the location, convert it to a three-dimensional coordinate value using Stereo Vision and Camera Calibration, and estimate the Euclidean distance between people. As a result of performing an experiment for estimating the accuracy of 3D coordinates and the distance between objects (persons), the average error within 0.098m was shown in the estimation of the distance between multiple people within 1m.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.6
• /
• pp.2689-2708
• /
• 2016

Multi-camera systems which integrate two or more low-cost digital cameras are adopted to reach higher ground coverage and improve the base-height ratio in low altitude remote sensing. To guarantee accurate multi-camera integration, the geometric relationship among cameras must be determined through platform calibration techniques. This paper proposed a combined two-step platform calibration method. In the first step, the static platform calibration was conducted based on the stable relative orientation constraint and convergent conditions among cameras in static environments. In the second step, a dynamic platform self-calibration approach was proposed based on not only tie points but also straight lines in order to correct the small change of the relative relationship among cameras during dynamic flight. Experiments based on the proposed two-step platform calibration method were carried out with terrestrial and aerial images from a multi-camera system combined with four consumer-grade digital cameras onboard an unmanned aerial vehicle. The experimental results have shown that the proposed platform calibration approach is able to compensate the varied relative relationship during flight, acquiring the mosaicing accuracy of virtual images smaller than 0.5pixel. The proposed approach can be extended for calibrating other low-cost multi-camera system without rigorously mechanical structure.

• Journal of Korea Multimedia Society
• /
• v.19 no.12
• /
• pp.1909-1918
• /
• 2016

This paper presents a unified framework for joint Convolutional Neural Network (CNN) based vehicle detection by leveraging multi-spectral image pairs. With the observation that under challenging environments such as night vision and limited light source, vehicle detection in a single color image can be more tractable by using additional far-infrared (FIR) image, we design joint CNN architecture for both RGB and FIR image pairs. We assume that a score map from joint CNN applied to overall image can be considered as confidence of vehicle existence. To deal with various scale ratios of vehicle candidates, multi-scale images are first generated scaling an image according to possible scale ratio of vehicles. The vehicle candidates are then detected on local maximal on each score maps. The generation of overlapped candidates is prevented with non-maximal suppression on multi-scale score maps. The experimental results show that our framework have superior performance than conventional methods with a joint framework of multi-spectral image pairs reducing false positive generated by conventional vehicle detection framework using only single color image.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.273-277
• /
• 2003

A general process of electronic assembly is composed of a series of geometric alignments and bonding/screwing processes. After assembly, the function is tested in a following process of inspection. However, assembly of micro-optic devices requires both processes to be performed in equipment. Coarse geometric alignment is made by using vision and optical function is improved by the following fine motion based on feedback of tunable laser interferometer. The general system is composed of a precision robot system for 3D assembly, a 3D vision guided system for geometric alignment and an optical feedback system with a tunable laser. In this study, we propose a new fast alignment algorithm of micro-optic devices for both of visual and optical alignments. The main goal is to find a fastest alignment process and algorithms with state-of-the-art technology. We propose a new approach with an optimal sequence of processes, a visual alignment algorithm and a search algorithm for an optimal optical alignment. A system is designed to show the effectiveness and efficiency of the proposed method.

• Journal of Broadcast Engineering
• /
• v.11 no.4 s.33
• /
• pp.533-540
• /
• 2006

In this study we propose to use a facial motion capture technique to track facial motions and expressions effectively by using the stereo vision module, which has two CMOS IMAGE SENSORS. In the proposed tracking algorithm, a center point tracking technique and correlation tracking technique, based on neural networks, were used. Experimental results show that the two tracking techniques using stereo vision motion capture are able to track general face expressions at a 95.6% and 99.6% success rate, for 15 frames and 30 frames, respectively. However, the tracking success rates(82.7%,99.1%) of the center point tracking technique was far higher than those(78.7%,92.7%) of the correlation tracking technique, when lips trembled.

• Journal of Institute of Control, Robotics and Systems
• /
• v.18 no.12
• /
• pp.1122-1131
• /
• 2012

This paper presents obstacle avoidance method for scout robot or industrial robot in unknown environment by using IR sensor and vision system. In the proposed method, robots share the information where the obstacles are located in real-time, thus the robots can choose the best path for obstacle avoidance. Using IR sensor and vision system, multiple robots efficiently evade the obstacles by the proposed cooperation method. No landmark is used at wall or floor in experiment environment. The obstacles don't have specific color or shape. To get the information of the obstacle, vision system extracts the obstacle coordinate by using an image labeling method. The information obtained by IR sensor is about the obstacle range and the locomotion direction to decide the optimal path for avoiding obstacle. The experiment was conducted in $7m{\times}7m$ indoor environment with two-wheeled mobile robots. It is shown that multiple robots efficiently move along the optimal path in cooperation with each other in the space where obstacles are located.

• Proceedings of the KIEE Conference
• /
• 1993.07a
• /
• pp.410-413
• /
• 1993

Chamferless peg-in-hole process of the cylindrical type parts using force/torque sensor and vision sensor is analyzed and simulated in this paper. Peg-in-hole process is classified to the normal mode (only position error) and tilted mode(position and orientation error). The tilted mode is sub-classified to the small and the big tilted mode according to the relative orientation error. Since the big tilted node happened very rare, most papers dealt with only the normal or the small tilted mode. But the most errors of the peg-in-hole process happened in the big tilted mode. This problem is analyzed and simulated in this paper using the force/torque sensor and vision senor. In the normal mode, fuzzy logic is introduced to combine the data of the force/torque sensor and vision sensor. Also the whole processing algorithms and simulations are presented.

• Journal of Korea Multimedia Society
• /
• v.19 no.10
• /
• pp.1747-1758
• /
• 2016

In this paper, we propose a new fast Hough transform to improve the processing time and line detection of Hough transform that is widely used in various vision systems. First, for the fast processing time, we reduce the number of features by using multi-statistical methods and also reduce the dimension of angle through six separate directions. Next, for improving the line detection, we effectively detect the lines of various directions by designing the line detection method which detects line in proportion to the number of features in six separate directions. The proposed method was evaluated with previous methods and obtained the excellent results. The processing time was improved in about 20% to 50% and line detection was performed better in various directions than conventional methods with experimental images.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.6 no.4
• /
• pp.249-254
• /
• 2013

In the area of computer vision, face recognition is being intensively researched. It is generally known that before a face is recognized it must be localized. Skin-color information is an important feature to segment skin-color regions. To extract skin-color regions the skin-color model based on multi-layer perceptron has been proposed. Extracted regions are analyzed to emphasize ellipsoidal regions. The results from this study show good accuracy for our vehicle driver face detection system.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.148-153
• /
• 2009

Few methods have dealt with segmenting multiple images with analogous content. Concurrent images of a scene and gathered images of a similar foreground are examples of these images, which we term consistent scene images. In this paper, we present a method to segment these images based on manual segmentation of one image, by iteratively propagating information via multi-level cues with adaptive confidence. The cues are classified as low-, mid-, and high- levels based on whether they pertain to pixels, patches, and shapes. Propagated cues are used to compute potentials in an MRF framework, and segmentation is done by energy minimization. Through this process, the proposed method attempts to maximize the amount of extracted information and maximize the consistency of segmentation. We demonstrate the effectiveness of the proposed method on several sets of consistent scene images and provide a comparison with results based only on mid-level cues [1].