• 한국통신학회논문지
• /
• 제40권9호
• /
• pp.1828-1836
• /
• 2015

객체 추적 알고리즘들은 객체 인식 결과를 이용한 관심영역 설정을 통해 영상 전체에 대한 연산이 수행되는 것을 방지하여 연산량을 줄일 수 있다. 따라서 객체 인식 알고리즘의 정확한 객체 검출은 객체 추적에서 매우 중요한 과정이다. 고정된 카메라를 기반으로 하여 이동하는 객체를 검출 하는 방법으로 배경 차 알고리즘이 널리 사용되어왔고 많은 연구에 의해 배경 모델링 방법이 개선되면서 배경 차 알고리즘의 성능이 개선되었으나 여전히 정확하지 못한 배경 모델링에 의한 객체 오검출의 문제를 가진다. 이에 본 논문에서는 제스쳐 인식에 주로 사용되는 모션 히스토리 이미지 알고리즘을 배경 차 알고리즘과 융합하여 기존의 배경 차 알고리즘이 가지는 문제점을 극복할 수 있는 다중 이동객체 검출 알고리즘을 제안한다. 제안하는 알고리즘은 융합 과정 추가로 수행시간이 다소 길어지나 실시간성을 만족하며 기존의 배경 차 알고리즘에 비해 높은 정확도를 가짐을 실험을 통해 확인하였다.

• 정보학연구
• /
• 제10권4호
• /
• pp.27-39
• /
• 2007

Object tracking in a real time environment is one of challenging subjects in computer vision area during past couple of years. This paper proposes a method of object detection and tracking using adaptive background estimation in real time environment. To obtain a stable and adaptive background, we combine 3-frame differential method and running average single gaussian background model. Using this background model, we can successfully detect moving objects while minimizing false moving objects caused by noise. In the tracking phase, we propose a matching criteria where the weight of position and inner brightness distribution can be controlled by the size of objects. Also, we adopt a Kalman Filter to overcome the occlusion of tracked objects. By experiments, we can successfully detect and track objects in real time environment.

• 제어로봇시스템학회논문지
• /
• 제19권1호
• /
• pp.65-71
• /
• 2013

In this work, new smart phone based moving target detection is proposed. In order to implement the task, methods of real time image transmission from network camera, motion detecting algorithm and its effective implementation are also addressed. The network camera transfers image data by MJPEG format which contains various information such as data and IP address, and the smart phone separates the image data received through a WiFi module. Later, the image data is converted to a Bitmap image format, and with the help of the embedded OpenCV library on a smart phone and algorithm, it was found that the moving object was identified effectively in terms of real time monitoring and detection.

• 한국HCI학회:학술대회논문집
• /
• 한국HCI학회 2006년도 학술대회 1부
• /
• pp.812-817
• /
• 2006

This paper presents an image-space algorithm to real-time collision detection, which is run completely by GPU. For a single object or for multiple objects with no collision, the front and back faces appear alternately along the view direction. However, such alternation is violated when objects collide. Based on these observations, the algorithm has been devised, and the implementation utilizes the state-of-the-art functionalities of GPU such as framebuffer objects(FBO), vertex buffer object(VBO) and occlusion query. The experimental results show the feasibility of GPU-intensive collision detection and its performance gain in real-time applications such as 3D games.

• 한국항행학회논문지
• /
• 제27권5호
• /
• pp.685-689
• /
• 2023

The proposed system in the study aims to detect forest fires in real-time stream data received from the drone-camera. Recently, the number of wildfires has been increasing, and also the large scaled wildfires are frequent more and more. In order to prevent forest fire damage, many experiments using the drone camera and vision analysis are actively conducted, however there were many challenges, such as network speed, pre-processing, and model performance, to detect forest fires from real-time streaming data of the flying drone. Therefore, this study applied image data processing works to capture five good image frames for vision analysis from whole streaming data and then developed the object detection model based on YOLO_v2. As the result, the classification model performance of forest fire images reached upto 93% of accuracy, and the field test for the model verification detected the forest fire with about 70% accuracy.

• 로봇학회논문지
• /
• 제17권3호
• /
• pp.255-263
• /
• 2022

3D object detection pipelines often incorporate RGB-based object detection methods such as YOLO, which detects the object classes and bounding boxes from the RGB image. However, in complex environments where objects are heavily cluttered, bounding box approaches may show degraded performance due to the overlapping bounding boxes. Mask based methods such as Mask R-CNN can handle such situation better thanks to their detailed object masks, but they require much longer time for data preparation compared to bounding box-based approaches. In this paper, we present a 3D object recognition pipeline which uses either the YOLO or Mask R-CNN real-time object detection algorithm, K-nearest clustering algorithm, mask reduction algorithm and finally Principal Component Analysis (PCA) alg orithm to efficiently detect 3D poses of objects in a complex environment. Furthermore, we also present an improved YOLO based 3D object detection algorithm that uses a prioritized heightmap clustering algorithm to handle overlapping bounding boxes. The suggested algorithms have successfully been used at the Artificial-Intelligence Robot Challenge (ARC) 2021 competition with excellent results.

• Journal of information and communication convergence engineering
• /
• 제18권3호
• /
• pp.147-154
• /
• 2020

In this study, we created a system that can determine the number of vehicles entering and departing a charging station in real time for solving waiting time problems during refueling. Accordingly, we use the You Only Look Once object detection algorithm to detect and count the number of vehicles in the charging station and send the data to the Firebase Realtime Database. The result is shown using an Android application that provides a map function with the Kakao Maps API at the user interface side. Our system has an accuracy of 91% and an average response time of 3.1 s. Therefore, this system can be used by drivers to determine the availability of a charging station and to identify the charging station with the least waiting time for charging their vehicle.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2001년도 합동 추계학술대회 논문집 정보 및 제어부문
• /
• pp.365-368
• /
• 2001

This work describes a real-time method, based on motion energy detection, for detecting and tracking moving object in the consecutive image sequences. The motion of moving objects is detected by taking the difference of the two consecutive image frames. In addition an edge information of the current image is utilized in order to further increase the accuracy of detection. We can track the moving objects continuously by detecting the motion of objects from the sequence of image frames. A prototype system has been implemented using a TI TMS320C6201 EVM fixed-point DSP board, which can successfully track a moving human in real-time.

• 지능정보연구
• /
• 제28권1호
• /
• pp.89-106
• /
• 2022

최근 영상 데이터의 급증으로 이를 효과적으로 처리하기 위해 객체 탐지 및 추적, 행동 인식, 표정 인식, 재식별(Re-ID)과 같은 다양한 컴퓨터비전 기술에 대한 수요도 급증했다. 그러나 객체 탐지 및 추적 기술은 객체의 영상 촬영 장소 이탈과 재등장, 오클루전(Occlusion) 등과 같이 성능을 저하시키는 많은 어려움을 안고 있다. 이에 따라 객체 탐지 및 추적 모델을 근간으로 하는 행동 및 표정 인식 모델 또한 객체별 데이터 추출에 난항을 겪는다. 또한 다양한 모델을 활용한 딥러닝 아키텍처는 병목과 최적화 부족으로 성능 저하를 겪는다. 본 연구에서는 YOLOv5기반 DeepSORT 객체추적 모델, SlowFast 기반 행동 인식 모델, Torchreid 기반 재식별 모델, 그리고 AWS Rekognition의 표정 인식 모델을 활용한 영상 분석 시스템에 단일 연결 계층적 군집화(Single-linkage Hierarchical Clustering)를 활용한 재식별(Re-ID) 기법과 GPU의 메모리 스루풋(Throughput)을 극대화하는 처리 기법을 적용한 행동 및 표정 검출용 영상 분석 시스템을 제안한다. 본 연구에서 제안한 시스템은 간단한 메트릭을 사용하는 재식별 모델의 성능보다 높은 정확도와 실시간에 가까운 처리 성능을 가지며, 객체의 영상 촬영 장소 이탈과 재등장, 오클루전 등에 의한 추적 실패를 방지하고 영상 내 객체별 행동 및 표정 인식 결과를 동일 객체에 지속적으로 연동하여 영상을 효율적으로 분석할 수 있다.

• ETRI Journal
• /
• 제45권5호
• /
• pp.836-846
• /
• 2023

Mobile robots are used in modern life; however, object recognition is still insufficient to realize robot navigation in crowded environments. Mobile robots must rapidly and accurately recognize the movements and shapes of pedestrians to navigate safely in pedestrian-rich spaces. This study proposes real-time, accurate, three-dimensional (3D) multi-pedestrian detection and tracking using a 3D light detection and ranging (LiDAR) point cloud in crowded environments. The pedestrian detection quickly segments a sparse 3D point cloud into individual pedestrians using a lightweight convolutional autoencoder and connected-component algorithm. The multi-pedestrian tracking identifies the same pedestrians considering motion and appearance cues in continuing frames. In addition, it estimates pedestrians' dynamic movements with various patterns by adaptively mixing heterogeneous motion models. We evaluate the computational speed and accuracy of each module using the KITTI dataset. We demonstrate that our integrated system, which rapidly and accurately recognizes pedestrian movement and appearance using a sparse 3D LiDAR, is applicable for robot navigation in crowded spaces.