• Journal of Broadcast Engineering
• /
• v.26 no.6
• /
• pp.738-747
• /
• 2021

Today, with the improvement of deep learning technology, computer vision areas such as image classification, object detection, object segmentation, and object tracking have shown remarkable improvements. Various applications such as intelligent surveillance, robots, Internet of Things, and autonomous vehicles in combination with deep learning technology are being applied to actual industries. Accordingly, the requirement of an efficient compression method for video data is necessary for machine consumption as well as for human consumption. In this paper, we propose an object-based compression of thermal infrared images for machine vision. The input image is divided into object and background parts based on the object detection results to achieve efficient image compression and high neural network performance. The separated images are encoded in different compression ratios. The experimental result shows that the proposed method has superior compression efficiency with a maximum BD-rate value of -19.83% to the whole image compression done with VVC.

• The Journal of the Convergence on Culture Technology
• /
• v.4 no.3
• /
• pp.209-212
• /
• 2018

Anisotropic conductive film(ACF) bonding is widely used for making fine interconnections between two different materials where soldering is not easily applicable. There are three constraints for the successful implementation of ACF bonding. A bonding contact should be pressed by a hot head with the right pressure and temperature for a pre-defined curing time. In this paper, a method for ACF bonding based on machine vision system is proposed and verified through some experiments. The system calculates the position and orientation of printed circuit boards(PCBs) on a bonding table and estimates the optimal hitting point where the hot head should be applied. Experimental results show that the proposed system achieves better adhesive strength by providing head flatness over contact surfaces.

• Journal of Electrical Engineering and Technology
• /
• v.8 no.5
• /
• pp.1194-1201
• /
• 2013

This paper introduces a vision-based automatic system (VAMS) for non-contact measurement of morphometric characteristics of flatfish, such as total length (TL), body width (BW), height (H), and weight (W). The H and W are simply measured by a laser displacement and a load cell, respectively. The TL and BW are measured by a proposed morphological image processing algorithm. The proposed algorithm cans measurement, when the tail of flatfish is deformed, and when it is randomly oriented. In the experiment, the average and maximum measurement errors were recorded, and standard deviations and coefficients of variation (CVs) for the measurements were calculated. From those results, when flatfish the TL measurements had an average of 266.844 mm, a standard deviation of 0.351 mm, a CV of 0.131%, and a maximum error of 0.87 mm with straightened flatfish ($TL_A$ : 267 mm, $BW_A$ : 141 mm), and when flatfish of different sizes were measured, the errors in the TL and BW measurements were both about 0.2 %. Using a single conveyor, the VAMS can process up to 900 fishes per hour. Moreover, it can measure morphometric characteristics of flatfish with a TL of up to 500 mm.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.5 no.2
• /
• pp.74-80
• /
• 2012

In this paper, we developed the profile analysis-based machine vision system for inspecting assembly parts in the industrial field. Implemented system composed of triple set of camera: one was used for acquiring slant image; other is required to acquire a top image; the other was used for side image. After obtaining parts which have gray scale image, threshold value was calculated by analyzing the profile of the image. Experimental results showed that proposed algorithm have a good performance for detecting fault parts and for classifying each parts as well.

• Journal of the Korean Society for Precision Engineering
• /
• v.31 no.3
• /
• pp.261-268
• /
• 2014

Precision screws have a wide range of industrial applications such as electrical and automotive products. To produce screw threads with high precision, not only high precision manufacturing technology but also reliable measurement technology is required. Machine vision systems have been used in the automatic inspection of screw threads based on backlight illumination, which cannot detect defects on the thread surface. Recently, an omnidirectional inspection system for screw threads was developed to obtain $360^{\circ}$ images of screws, based on front light illumination. In this study, the illumination design for the omnidirectional inspection system was modified by adding a light shield to improve the image uniformity. Optical simulation for various shield designs was performed to analyze image uniformity of the obtained images. The simulation results were analyzed statistically using response surface method, from which optical performance of the omnidirectional inspection system could be optimized in terms of image quality and uniformity.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.396-398
• /
• 2022

The endoscopic pathological findings of gastrointestinal tract (GIT) are important in the early diagnosis of colorectal cancer. Deep learning based on convolutional nueral network (CNN) has been implemented to solve the subjective analysis problem and to increase the performance of early detection of pathological findings. However, the desired performance is yet to be achieved and CNNs are computationally complex. To solve these problems, in this paper, we propose a vision transformer based endoscopic pathological findings classification for the early detection of colorectal cancer. Publicly available endoscopic images with three pathological findings, including esophagitis, polyps, and ulcerative colitis, each with 1000 images were used. Using our approach, we have achieved a test accuracy of 98% in classifying the three pathological findings.

• Proceedings of the KSME Conference
• /
• 2008.11a
• /
• pp.768-772
• /
• 2008

Various scanning systems have been studied in many industrial areas to acquire a range data or to reconstruct an explicit 3D model. Currently optical technology has been used widely by virtue of noncontactness and high-accuracy. In this paper, we describe a 3D laser scanning system developped to reconstruct the 3D surface of a large-scale object such as a curved-plate of ship-hull. Our scanning system comprises of 4ch-parallel laser vision modules using a triangulation technique. For multi laser vision, calibration method based on least square technique is applied. In global scanning, an effective method without solving difficulty of matching problem among the scanning results of each camera is presented. Also minimal image processing algorithm and robot-based calibration technique are applied. A prototype had been implemented for testing.

• Journal of Ocean Engineering and Technology
• /
• v.30 no.5
• /
• pp.426-430
• /
• 2016

Recently, many unmanned surface vehicles (USVs) have been developed and researched for various fields such as the military, environment, and robotics. In order to perform purpose specific tasks, common autonomous navigation technologies are needed. Obstacle avoidance is important for safe autonomous navigation. This paper describes a vector field histogram+ (VFH+) based obstacle avoidance method that uses the monocular vision of an unmanned surface vehicle. After creating a polar histogram using VFH+, an open space without the histogram is selected in the moving direction. Instead of distance sensor data, monocular vision data are used for make the polar histogram, which includes obstacle information. An object on the water is recognized as an obstacle because this method is for USV. The results of a simulation with sea images showed that we can verify a change in the moving direction according to the position of objects.

• Structural Monitoring and Maintenance
• /
• v.6 no.2
• /
• pp.161-171
• /
• 2019

There have been a few checking frameworks intended to ensure and improve the nature of their regular habitat. The greater part of these frameworks are constrained in their capacities. In this paper, the insightful checking framework intended for debacle help and administrations has been exhibited. The ideal administrations, necessities and coming about plan proposition have been indicated. This has prompted a framework that depends fundamentally on ecological examination so as to offer consideration and security administrations to give the self-governance of indigenous habitats. In this sense, ecological acknowledgment is considered, where, in light of past work, novel commitments have been made to help include based and PC vision situations. This epic PC vision procedure utilized as notice framework for avalanche identification depends on changes in the normal landscape. The multi-criteria basic leadership strategy is used to incorporate slope data and the level of variety of the highlights. The reproduction consequences of highlight point discovery are shown in highlight guide coordinating toward discover steady and coordinating component focuses and effectively identified utilizing these two systems, by examining the variety in the distinguished highlights and the element coordinating.

• The Journal of Korea Robotics Society
• /
• v.3 no.1
• /
• pp.33-41
• /
• 2008

Based on object recognition technology, we present a new global localization method for robot navigation. For doing this, we model any indoor environment using the following visual cues with a stereo camera; view-based image features for object recognition and those 3D positions for object pose estimation. Also, we use the depth information at the horizontal centerline in image where optical axis passes through, which is similar to the data of the 2D laser range finder. Therefore, we can build a hybrid local node for a topological map that is composed of an indoor environment metric map and an object location map. Based on such modeling, we suggest a coarse-to-fine strategy for estimating the global localization of a mobile robot. The coarse pose is obtained by means of object recognition and SVD based least-squares fitting, and then its refined pose is estimated with a particle filtering algorithm. With real experiments, we show that the proposed method can be an effective vision- based global localization algorithm.