• Annual Conference of KIPS
• /
• 2015.10a
• /
• pp.1442-1445
• /
• 2015

본 논문은 SAR 영상에서 객체를 추출하는 새로운 방법으로 특징점 기반 분할 히스토그램 기법을 제안한다. 제안하는 방법은 영상 히스토그램에서 객체로 추정될 수 있는 영역을 세밀하게 추출하기 위해 영상에서 특징점을 추출한 후, 특징점의 밝기를 기준으로 히스토그램을 분할한다. 분할 히스토그램이 배경과 객체 성분을 모두 포함하고 있을 경우 해당 영역의 혼합 확률밀도함수가 교차되는 임계점을 계산한다. 계산된 임계점을 기준으로 현재 영역이 전체 영상에서 차지하는 비율을 비교하여 배경과 객체 여부를 판단한다. 제안하는 방법은 무인 감시 정찰 시스템 등 다양한 응용 기술에 활용될 수 있을 것으로 기대한다.

• ETRI Journal
• /
• v.40 no.6
• /
• pp.788-793
• /
• 2018

3D reconstruction of urban architecture, land, and roads is an important part of building a "digital city." Unmanned aerial vehicles (UAVs) are gradually replacing other platforms, such as satellites and aircraft, in geographical image collection; the reason for this is not only lower cost and higher efficiency, but also higher data accuracy and a larger amount of obtained information. Recent 3D reconstruction algorithms have a high degree of automation, but their computation time is long and the reconstruction models may have many voids. This paper decomposes the object into multiple regional parallel reconstructions using the clustering principle, to reduce the computation time and improve the model quality. It is proposed to detect the planar area under low resolution, and then reduce the number of point clouds in the complex area.

• Journal of Korea Multimedia Society
• /
• v.25 no.8
• /
• pp.1049-1058
• /
• 2022

HAR(Human Action Recognition) such as anomaly and object detection has become a trend in research field(s) that focus on utilizing Artificial Intelligence (AI) methods to analyze patterns of human action in crime-ridden area(s), media services, and industrial facilities. Especially, in real-time system(s) using video streaming data, HAR has become a more important AI-based research field in application development and many different research fields using HAR have currently been developed and improved. In this paper, we propose and analyze a deep-learning-based HAR that provides more efficient scheme(s) using an intelligent AI models, such system can be applied to media services using RGB video streaming data usage without feature extraction pre-processing. For the method, we adopt Slowfast based on the Deep Neural Network(DNN) model under an open dataset(HMDB-51 or UCF101) for improvement in prediction accuracy.

• Journal of Broadcast Engineering
• /
• v.24 no.1
• /
• pp.67-76
• /
• 2019

In this paper, we propose an efficient object classification method based on semantic segmentation for multi-labeled image data. In addition to various pixel unit information and processing techniques such as color information, contour, contrast, and saturation included in image data, a detailed region in which each object is located is extracted as a meaningful unit and the experiment is conducted to reflect the result in the inference. We use a neural network that has been proven to perform well in image classification to understand which object is located where image data containing various class objects are located. Based on these researches, we aim to provide artificial intelligence services that can classify real-time detailed areas of complex images containing various objects in the future.

• Transactions on Semiconductor Engineering
• /
• v.2 no.1
• /
• pp.17-20
• /
• 2024

This paper presents a novel lightweight object detection model tailored for low-powered edge devices, addressing the limitations of traditional resource-intensive computer vision models. Our proposed detector, inspired by the Single Shot Detector (SSD), employs a compact yet robust network design. Crucially, it integrates an 'enhancer block' that significantly boosts its efficiency in detecting smaller objects. The model comprises two primary components: the Light_Block for efficient feature extraction using Depth-wise and Pointwise Convolution layers, and the Enhancer_Block for enhanced detection of tiny objects. Trained from scratch on the Udacity Annotated Dataset with image dimensions of 300x480, our model eschews the need for pre-trained classification weights. Weighing only 5.5MB with approximately 0.43M parameters, our detector achieved a mean average precision (mAP) of 27.7% and processed at 140 FPS, outperforming conventional models in both precision and efficiency. This research underscores the potential of lightweight designs in advancing object detection for edge devices without compromising accuracy.

• Journal of Broadcast Engineering
• /
• v.22 no.6
• /
• pp.713-723
• /
• 2017

Since the panoramic image can overcome the limitation of the viewing angle of the camera and have a wide field of view, it has been studied effectively in the fields of computer vision and stereo camera. In order to generate a panoramic image, stitching images taken by a plurality of general cameras instead of using a wide-angle camera, which is distorted, is widely used because it can reduce image distortion. The image stitching technique creates descriptors of feature points extracted from multiple images, compares the similarities of feature points, and links them together into one image. Each feature point has several hundreds of dimensions of information, and data processing time increases as more images are stitched. In particular, when a panorama is generated on the basis of an image photographed by a plurality of unspecified cameras with respect to an object, the extraction processing time of the overlapping feature points for similar images becomes longer. In this paper, we propose a preprocessing process to efficiently process stitching based on an image obtained from a number of unspecified cameras for one object or environment. In this way, the data processing time can be reduced by pre-grouping images based on camera sensor information and reducing the number of images to be stitched at one time. Later, stitching is done hierarchically to create one large panorama. Through the grouping preprocessing proposed in this paper, we confirmed that the stitching time for a large number of images is greatly reduced by experimental results.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.5
• /
• pp.83-93
• /
• 2013

Recently, the SURF algorithm, which is conjugated for object tracking system as part of many computer vision applications, is a well-known scale- and rotation-invariant feature detection algorithm. The SURF, due to its high computational complexity, there is essential to develop a hardware accelerator in order to be used on an IP in embedded environment. However, the SURF requires a huge local memory, causing many problems that increase the chip size and decrease the value of IP in ASIC and SoC system design. In this paper, we proposed a way to design a SURF algorithm in hardware with greatly reduced local memory by partitioning the algorithms into several Sub-IPs using external memory and a DMA. To justify validity of the proposed method, we developed an example of simplified object tracking algorithm. The execution speed of the hardware IP was about 31 frame/sec, the logic size was about 74Kgate in the 30nm technology with 81Kbytes local memory in the embedded system platform consisting of ARM Cortex-M0 processor, AMBA bus(AHB-lite and APB), DMA and a SDRAM controller. Hence, it can be used to the hardware IP of SoC Chip. If the image processing algorithm akin to SURF is applied to the method proposed in this paper, it is expected that it can implement an efficient hardware design for target application.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.4 no.1
• /
• pp.101-110
• /
• 2004

The robots that will be needed in the near future are human-friendly robots that are able to coexist with humans and support humans effectively. To realize this, humans and robots need to be in close proximity to each other as much as possible. Moreover, it is necessary for their interactions to occur naturally. It is desirable for a robot to carry out human following, as one of the human-affinitive movements. The human-following robot requires several techniques: the recognition of the moving objects, the feature extraction and visual tracking, and the trajectory generation for following a human stably. In this research, a predictable intelligent space is used in order to achieve these goals. An intelligent space is a 3-D environment in which many sensors and intelligent devices are distributed. Mobile robots exist in this space as physical agents providing humans with services. A mobile robot is controlled to follow a walking human using distributed intelligent sensors as stably and precisely as possible. The moving objects is assumed to be a point-object and projected onto an image plane to form a geometrical constraint equation that provides position data of the object based on the kinematics of the intelligent space. Uncertainties in the position estimation caused by the point-object assumption are compensated using the Kalman filter. To generate the shortest time trajectory to follow the walking human, the linear and angular velocities are estimated and utilized. The computer simulation and experimental results of estimating and following of the walking human with the mobile robot are presented.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.8
• /
• pp.1698-1706
• /
• 1997

In this paper, we envisioned a multimedia object recognition system processing and combinig information from all available sources, such as 2-D, 3-D, color and sound data. Out of the overall system, we proposed 3-D information extraction and object recognition methods. Firstly, surfaces are classified by z-gradient from the range data, surface features are extracted using the intersection of normal vectors. Also feature relationship such as intersection angle and distance is established between the surfaces. Secondly, recognition is accomplished by matching process which is improtant step in the image understanding systems. Matching process is very improtant procedures because of more general and more efficient method is needed in the field of multimedia sytem. Therefore, we focused the proposal of matching process and in this article, first of all, we deal with the matching process of the 3-D object. Similarity measures are calculated.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.2
• /
• pp.135-140
• /
• 2012

The mirror neuron system in the cerebrum, which are handled by visual information-based imitative learning. When we observe the observer's range of mirror neuron system, we can assume intention of performance through progress of neural activation as specific range, in include of partially hidden range. It is goal of our paper that imitative learning is applied to 3D vision-based intelligent system. We have experiment as stereo camera-based restoration about acquired 3D image our previous research Using Optical flow, unscented Kalman filter. At this point, 3D input image is sequential continuous image as including of partially hidden range. We used Hidden Markov Model to perform the intention recognition about performance as result of restoration-based hidden range. The dynamic inference function about sequential input data have compatible properties such as hand gesture recognition include of hidden range. In this paper, for proposed intention recognition, we already had a simulation about object outline and feature extraction in the previous research, we generated temporal continuous feature vector about feature extraction and when we apply to Hidden Markov Model, make a result of simulation about hand gesture classification according to intention pattern. We got the result of hand gesture classification as value of posterior probability, and proved the accuracy outstandingness through the result.