• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.2
• /
• pp.143-148
• /
• 2014

Particle filter algorithm has been proven very successful for non-linear and non-Gaussian estimation problem and thus it has been widely used for object tracking for video signals. If the object moves significantly, particle filter needs very large number of particles to track object and this results high computational cost. In this paper, modified particle filter by adopting motion vector is proposed for tracking vehicle in low frame rate(LPR) video input, which the object moving significantly and randomly between consecutive frames. In the proposed algorithm, motion vector is applied in selection and observe step. The experimental result shows that the proposed particle filter can track vehicle successfully in the case when previous one fails. And it also shows the propose method increases the precision of tracking.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.19 no.1
• /
• pp.1-8
• /
• 2024

Robotic arms have been widely utilized in various labor-intensive industries such as manufacturing, agriculture, and food services, contributing to increasing productivity. In the development of industrial robotic arms, camera sensors have many advantages due to their cost-effectiveness and small sizes. However, estimating object positions is a challenging problem, and it critically affects to the robustness of object manipulation functions. This paper proposes a method for estimating the 3D positions of objects, and it is applied to a pick-and-place task. A deep learning model is utilized to detect 2D bounding boxes in the image plane, and the pinhole camera model is employed to compute the object positions. To improve the robustness of measuring the 3D positions of objects, we analyze the effect of lens distortion and introduce a false positive filtering process. Experiments were conducted on a real-world scenario for moving medicine bottles by using a camera-based manipulator. Experimental results demonstrated that the distortion removal and false positive filtering are effective to improve the position estimation precision and the manipulation success rate.

• The Journal of the Korea Contents Association
• /
• v.6 no.2
• /
• pp.44-50
• /
• 2006

This study has proposed a method of content-based image retrieval using object region in order to overcome disadvantages of existing color histogram methods. The existing color histogram methods have a weak point of reducing accuracy, because these have both a quantization error and an absence of spatial information. In order to overcome this problem, we convert a color information to a HSV space, quantize hue factor being pure color information, and calculate histogram. And then we use hue for retrieval feature that is robust in brightness, movement, and rotation. To solve the problem of the absence of spatial information, we select object region in terms of color feature and region correlation. And we use both the edge and the DC in the selected region for retrieving. As a result of experiment with 1,000 natural color images, the proposed method shows better precision and recall than the existing methods.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.6 no.5
• /
• pp.367-372
• /
• 2005

In this paper, we proposed a method of content-based image retrieval using variable region. Content-based image retrieval uses color histogram for the most part. But the existing color histogram methods have a disadvantage that it reduces accuracy because of quantization error and absence of spatial information. In order to overcome this, we convert color information to HSV space, quantize hue factor being pure color information, and calculate histogram of the factor. On the other hand, to solve the problem of the absence of spatial information, we select object region in consideration of color feature and region correlation. It maintains the size of region in the selected object region. But non-object region is integrated in one region. After of selection variable region, we retrieve using color feature. As the result of experimentation, the proposed method improves 10$\%$ in average of precision.

• Journal of Institute of Control, Robotics and Systems
• /
• v.8 no.4
• /
• pp.327-332
• /
• 2002

The main problems of computational complexity in object tracking are definition of objects, segmentations and identifications in non-structured environments with erratic movements and collisions of objects. The object's information as a region that corresponds to objects without discriminating among objects are considered. This paper describes the algorithm that, automatically and efficiently, recognizes and keeps tracks of interest-regions selected by users in video or camera image sequences. The block-based feature matching method is used for the region tracking. This matching process considers only dominant feature points such as corners and curved-edges without requiring a pre-defined model of objects. Experimental results show that the proposed method provides above 96% precision for correct region matching and real-time process even when the objects undergo scaling and 3-dimen-sional movements In successive image sequences.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.118.2-118
• /
• 2001

This research describes a development of laser distance sensor with precise resolution even in the case that the object surface has some curvature. There are typical two methods in measuring the distance by using laser light source, so called time of flight and optic-triangular methods. Both methods have an advantage and a disadvantage each other. In general, the time of flight method produces wide range of the measurement, but low accuracy. The other method is vice versa. In this research, the optic-triangular methods with one-dimensional CCD cell are proposed to obtain the precise distance measure from the sensor the surface of the curved object ...

• Journal of Sensor Science and Technology
• /
• v.30 no.5
• /
• pp.290-294
• /
• 2021

Experts have designed popular and successful model architectures, which, however, were not the optimal option for different scenarios. Despite the remarkable performances achieved by deep neural networks, manually designed networks for classification tasks are the backbone of object detection. One major challenge is the ImageNet pre-training of the search space representation; moreover, the searched network incurs huge computational cost. Therefore, to overcome the obstacle of the pre-training process, we introduce a network adaptation technique using a pre-trained backbone model tested on ImageNet. The adaptation method can efficiently adapt the manually designed network on ImageNet to the new object-detection task. Neural architecture search (NAS) is adopted to adapt the architecture of the network. The adaptation is conducted on the MobileNetV2 network. The proposed NAS is tested using SSDLite detector. The results demonstrate increased performance compared to existing network architecture in terms of search cost, total number of adder arithmetics (Madds), and mean Average Precision(mAP). The total computational cost of the proposed NAS is much less than that of the State Of The Art (SOTA) NAS method.

• Journal of the Korean Society for Precision Engineering
• /
• v.21 no.7
• /
• pp.185-194
• /
• 2004

In order to analyze the motion errors of the aerostatic stage, it is necessary to consider the influence of the moment variation occurred inside the pads. In this paper, a motion error analysis method utilizing the transfer functions on the reaction force and moment is proposed, and general characteristics of the transfer functions are discussed. Calculated motion errors by the proposed method show good agreement with the ones calculated by Multi fad Method, which is considered the entire table as an analysis object. Also, by the introduction of the transfer function of motion errors, which represent the relationship between the spatial frequency components of the rail form error and motion errors, motional characteristics of the porous aerostatic stage can be generalized. In detail, the influence of the spatial frequencies is analyzed qualitatively, and the patterns of the insensitive frequencies which almost do not affect the linear motion error or angular motion error according to the rail length ratio and the number of the pad are verified. The relationship between the moment variation occurred inside the pads and the motion errors is also verified together.

• The Journal of the Acoustical Society of Korea
• /
• v.37 no.2
• /
• pp.118-128
• /
• 2018

In this paper, we have studied how to search an underwater object by learning the image generated by the side scan sonar in the convolution neural network. In the method of human side analysis of the side scan image or the image, the convolution neural network algorithm can enhance the efficiency of the analysis. The image data of the side scan sonar used in the experiment is the public data of NSWC (Naval Surface Warfare Center) and consists of four kinds of synthetic underwater objects. The convolutional neural network algorithm is based on Faster R-CNN (Region based Convolutional Neural Networks) learning based on region of interest and the details of the neural network are self-organized to fit the data we have. The results of the study were compared with a precision-recall curve, and we investigated the applicability of underwater object detection in convolution neural networks by examining the effect of change of region of interest assigned to sonar image data on detection performance.

• Journal of Korean Society on Water Environment
• /
• v.37 no.4
• /
• pp.275-285
• /
• 2021

Algal bloom is an important issue in maintaining the safety of the drinking water supply system. Fast detection and classification of algae images are essential for the management of algal blooms. Conventional visual identification using a microscope is a labor-intensive and time-consuming method that often requires several hours to several days in order to obtain analysis results from field water samples. In recent decades, various deep learning algorithms have been developed and widely used in object detection studies. YOLO is a state-of-the-art deep learning algorithm. In this study the third version of the YOLO algorithm, namely, YOLOv3, was used to develop an algae image detection model. YOLOv3 is one of the most representative one-stage object detection algorithms with faster inference time, which is an important benefit of YOLO. A total of 1,114 algae images for 30 genera collected by microscope were used to develop the YOLOv3 algae image detection model. The algae images were divided into four groups with five, 10, 20, and 30 genera for training and testing the model. The mean average precision (mAP) was 81, 70, 52, and 41 for data sets with five, 10, 20, and 30 genera, respectively. The precision was higher than 0.8 for all four image groups. These results show the practical applicability of the deep learning algorithm, YOLOv3, for algae image detection.