Earticle

Rumination in cattle is closely related to their health, which makes the automatic monitoring of rumination an important part of smart pasture operations. However, manual monitoring of cattle rumination is laborious and wearable sensors are often harmful to animals. Thus, we propose a computer vision-based method to automatically identify multi-object cattle rumination, and to calculate the rumination time and number of chews for each cow. The heads of the cattle in the video were initially tracked with a multi-object tracking algorithm, which combined the You Only Look Once (YOLO) algorithm with the kernelized correlation filter (KCF). Images of the head of each cow were saved at a fixed size, and numbered. Then, a rumination recognition algorithm was constructed with parameters obtained using the frame difference method, and rumination time and number of chews were calculated. The rumination recognition algorithm was used to analyze the head image of each cow to automatically detect multi-object cattle rumination. To verify the feasibility of this method, the algorithm was tested on multi-object cattle rumination videos, and the results were compared with the results produced by human observation. The experimental results showed that the average error in rumination time was 5.902% and the average error in the number of chews was 8.126%. The rumination identification and calculation of rumination information only need to be performed by computers automatically with no manual intervention. It could provide a new contactless rumination identification method for multi-cattle, which provided technical support for smart pasture.

To deal with the problems of occlusion, pose variations and illumination changes in the object tracking system, a regression model weighted multi-templates mean-shift (MS) algorithm is proposed in this paper. Target templates and occlusion templates are extracted to compose a multi-templates set. Then, the MS algorithm is applied to the multi-templates set for obtaining the candidate areas. Moreover, a regression model is trained to estimate the Bhattacharyya coefficients between the templates and candidate areas. Finally, the geometric center of the tracked areas is considered as the object's position. The proposed algorithm is evaluated on several classical videos. The experimental results show that the regression model weighted multi-templates MS algorithm can track an object accurately in terms of occlusion, illumination changes and pose variations.

Multi-object tracking techniques are receiving increasing attention due to the growing demand of autonomous driving. Recently, the performance of multi-object tracking has been improved significantly thank to deep learning technique. Most of multi-object tracking methods based on deep learning, however, are highly prone to frequent tracking losses and track-ID switching in case of limited viewpoint and occluded objects. To alleviate this problem, we propose a multi-camera Collaborate Multi Object Tracking (CMOT) method which performs online association of multiple tracked vehicles from stereo vision camera. CMOT not only provides global tracking IDs between multiple cameras but also helps reduce the problem of ID switching compared with the conventional multi-object tracking based on single camera. It can, therefore, improve the overall performance of multi-vehicle tracking compared to each individual camera. We demonstrate the multi-object tracking performance of the proposed method using stereo images of the KITTI dataset.

실사 및 그래픽 기반 가상현실 콘텐츠는 360도 영상을 기반으로 하며, 시청자의 의도나 자동 추천 기능을 통한 뷰포트 추출이 필수적이다. 본 논문은 360도 영상에서 다중 객체 추적 기반의 뷰포트 추출 시스템을 설계하고, 다중 뷰포트 추출에 필요한 병렬화된 연산 구조를 제안한다. 360도 영상에서 뷰포트 추출 과정을 ERP 좌표의 3D 구 표면 좌표 변환과 3D 구 표면 좌표의 뷰포트 내 2D 좌표 변환 과정을 순서대로 픽셀 단위의 스레드로 구성하여 연산을 병렬화하였다. 제안 구조는 항공 360도 영상 시퀀스들에 대하여 최대 30개의 뷰포트 추출 과정에 대한 연산 시간이 평가되었으며, 뷰포트 수에 정비례하는 CPU 기반 연산 시간에 비해 최대 5240배 가속화됨을 확인하였다. ERP 프레임 I/O 시간을 줄일 수 있는 고속의 I/O나 메모리 버퍼를 사용 시 뷰포트 추출 시간을 7.82배 추가 가속화가 가능하다. 제안하는 뷰포트 추출 병렬화 구조는 360도 비디오나 가상현실 콘텐츠들에 대한 동시 다중 접속 서비스나 사용자별 영상 요약 서비스 등에 활용될 수 있다.

Realistic and graphics-based virtual reality content is based on 360-degree videos, and viewport extraction through the viewer's intention or automatic recommendation function is essential. This paper designs a viewport extraction system based on multiple object tracking in 360-degree videos and proposes a parallel computing structure necessary for multiple viewport extraction. The viewport extraction process in 360-degree videos is parallelized by composing pixel-wise threads, through 3D spherical surface coordinate transformation from ERP coordinates and 2D coordinate transformation of 3D spherical surface coordinates within the viewport. The proposed structure evaluated the computation time for up to 30 viewport extraction processes in aerial 360-degree video sequences and confirmed up to 5240 times acceleration compared to the CPU-based computation time proportional to the number of viewports. When using high-speed I/O or memory buffers that can reduce ERP frame I/O time, viewport extraction time can be further accelerated by 7.82 times. The proposed parallelized viewport extraction structure can be applied to simultaneous multi-access services for 360-degree videos or virtual reality contents and video summarization services for individual users.

본 논문에서는 엣지컴퓨팅 기반 객체검출 및 Pan-Tilt-Zoom(PTZ) 제어를 활용한 다중객체의 순차추적 방법을 제 안한다. 제안하는 방법은 순찰단계와 순차추적단계로 나누어 PTZ 제어를 수행한다. 먼저 순찰 단계에서 카메라의 화각을 기준점으로 이동시킨 후 사전 정의된 이동 패턴에 따라 순찰 기동을 반복 수행한다. 순찰 기동 중 입력되는 모든 영상 프레임에서 사전 정의된 이상상황의 발생여부를 확인하기 위하여 객체검출 및 식별자 부여를 수행한다. 이상상황이 발생중이라고 판단되면 제안하는 방법은 순차추적단계로 진입하여 사전에 정의된 규칙과 현재 상황이 반영된 객체별 우선 추적 순위를 설정한다. 이 순위에 따라 제안하는 방법은 각 객체를 집중 촬영하기 위한 PTZ 제 어를 순차적으로 수행한다. 순차추적이 모두 끝나면 제안하는 방법은 이상상황을 처음 인지한 화각으로 돌아간 뒤 여전히 이상상황이 존재한다면 순차추적을 반복하고, 그렇지 않다면 순찰단계로 돌아오게 된다. 제안하는 방법의 성 능을 평가하기 위해 스쿨존에서의 돌발상황을 즉시 인지하는 시나리오에 맞춰 실험을 설계하였다. 실험결과, 제안하는 방법은 0.951의 F1 Score로 이상상황을 판별하면서 24 frame per seconds의 처리 속도를 보이며 그 유효성을 입증하였다.

In this paper, a method for sequential tracking of multi-object using edge computing-based object detection and Pan-Tilt-Zoom(PTZ) control is proposed. The proposed method performs PTZ control in two procedures, which are patrol stage and sequential tracking stage. In the patrol stage, the proposed method firstly moves the camera’s angle of view to the reference point then follows the predefined moving procedure for sequential patrol. During the patrol, object detection and identifier assignment are conducted on inputted video frames to recognize incident of the predefined abnormal situations. In case the abnormal situation is on going, the proposed method switches the mode into sequential tracking stage, and by the predefined rules and current situations, assigns the tracking priorities. Followed by the priority, the proposed method conducts PTZ controls in sequential order to capture each objects related to the incident. After all the sequential tracking is over, the proposed method returns back to the angle of view at which the initial abnormal situation was recognized, and then repeats the sequential tracking if the abnormal situation still exists, and returns to the patrol stage otherwise. In order to evaluate the performance of the proposed method, an experiment was designed according to the scenario of immediately recognizing traffic accidents in the school zone. As a result of the experiment, the proposed method proved its effectiveness by showing an F1 score of 0.95 for discriminating abnormal situations with showing a processing speed of 24 frames per seconds.

본 연구에서는 다중 물체 추적 기술 (Multi-Object Tracking)과 공간 데이터베이스를 이용한 실내 환경에서의 객체 위치 데이터 수집 및 분석 시스템을 제안한다. 기존의 다중 물체 추적의 결과는 영상 내 3차원 공간에 위치한 객체의 좌표를 2차원 좌표로 반환하여 객체 위치 분석에 직접적으로 사용하기 어려웠으나 평면 투영 (Ground Projection)을 통해 2차원 목표 지도상 좌표로 변환하여 이를 해결하고, 공간 데이터베이스의 쿼리를 이용하여 대상 공간에서의 객체 위치 분석에 활용될 수 있도록 하였다.

애플사의 아이폰 탄생과 함께한 스마트폰의 발전 속도는 매우 짧은 주기로 새로운 기술과 개념을 쏟아내고 있다. 이러한 기술의 발전 속도는 ‘무어의 법칙1)’과 같이 점점 빨라지고 있으며, 스마트폰 역시 인간생활의 많은 부 분을 빠르게 변화 시키고 있다. 특히 스마트폰을 이용한 오브젝트 제어 기술은 로봇, 홈오토매이션, 스마트오브젝트 등 많은 분야에서 활용되고 있다. 하지만, 현재 스마트폰의 오브젝트 제어 기술에는 다중제어의 한계성을 가지고 있 다. 본 연구는 스마트폰을 이용한 다중오브젝트 제어를 위해 블루투스모듈과 지그비모듈의 혼합 이용을 제안하며 그 것을 위해 필요한 어플리케이션의 디자인 요소와 지그비 통신을 위한 방법을 제시한다. 이를 통해 현재 스마트폰에 내장된 무선통신기술중 하나인 블루트스모듈을 이용해 다중 오브젝트제어를 가능하게 한다. 이는 스마트오브젝트 시 대에 스마트폰을 활용할 수 있는 새로운 방법의 제안으로 디자인 영역확장의 시도라 할 수 있을 것이다.

Since the advent of Apple’s iPhone, the smartphone industry has been producing new technologies and concepts at an accelerated pace. The speed of progress in this sector is exponentially increasing in accordance with Moore’s Law, and smartphones are rapidly changing various aspects of human life. Especially, object control technologies using smartphones are being utilized in various sectors, including robots, home automation, and smart objects. However, the current smartphone object control technology is limited in terms of multicontrol. This study proposes the combined usage of the Bluetooth and Zigbee Modules for multiple object control using smartphones, and presents the necessary application design properties and the methodology for Zigbee communication. The study is an attempt at a territorial expansion of design, as a proposal of new methods for utilizing smartphones in the age of smart objects.

본 논문에서는 별도의 고사양 카메라 또는 센서를 사용하지 않고, 단일 RGB 카메라와 경량화된 자세추 정 모델만으로 3차원 다중 객체 추정 방법에 대하여 제안하였다. 이를 위하여 자세추정 모델로부터 추정한 2D 관 절 키포인트 좌표값으로 검출 박스를 추정하였으며, 검출 박스의 jitter 오류를 보완하고자 OneEuroFilter를 사용 하였다. 검출된 박스의 중심점을 사용하여 객체의 좌우 이동을 추적하였으며, 박스의 크기와 높이 그리고 비율을 사용한 깊이 추정을 통하여 객체 추적 알고리즘을 개발하였다. 또한, 다중 객체 추적을 위하여 객체가 검출된 프 레임을 ID 값으로 하여 전후 프레임간 검출된 관절 키포인트의 유사도를 검사하여 ID 매칭을 수행하였다.

This paper proposed a 3D MOT method with only a single RGB camera and a lightweight pose estimation model without using a separate high-end camera or sensor. To this end, the detection box was estimated from the 2D joint keypoint coordinate estimated from the pose estimation model and OneEuroFilter was used to compensate for the jitter error in the detection box. The center point of the detected box was used to track the left and right movement of the object, and an object tracking algorithm was developed through depth estimation using the size, height, and ratio of the box. In addition, for multiple object tracking, ID matching was performed by examining the similarity of joint key points detected between current and previous frames using the frame in which the object was detected as an ID value.

To analyze the applications of all types of mechanical systems, general purpose analysis programs have been developed and commercialized. However, it is customary to develop and use customized programs even though they sometimes require more work than a general purpose program. A customized program is simplified to adapt to a particular application from the beginning, is designed for small computers, and developed with hardware-in-the-loop in mind so it can be applied effectively. By adding design knowledge and bundling know-how to an analysis program, analysis time can be reduced. And because an analysis has to work in conjunction with other analysis programs, a proprietary program that the user can easily modify can be useful. In this thesis, a multi-body dynamics analysis system is presented using one of the most useful programming techniques, object-oriented concept. The object-oriented concept defines a problem from the physical world as an abstract object, an abstract model. The object becomes encapsulated with the data and method. Simulation is performed using the object's interface. It is then possible for the user and the developer to modify and upgrade the program without having particular knowledge of the analysis program. The method presented in this thesis has the following advantages. Since the mechanical components of the multi-body system converts independent modeling into a class, the modification, exchange, distribution, and reuse of elements are increased. It becomes easier to employ a new analysis method and interface with other S/W and H/W systems. To employ a new analysis method, there is no need to modify elements of the main solver and the Library. In addition, information can be communicated to each object through messaging. It makes the modeling of new elements easier using inheritance. When developing a S/W for the computer simulation of physical system, it is reasonable to use object-oriented modeling. Also, for multi-body dynamics analysis, it is possible to develop a solver that is user-oriented.

인터넷과 웹 기술이 모바일 장치 중심으로 발전하면서 이미지 데이터는 사람, 텍스트, 공간 등 다양한 유형의 민감 정보를 담고 있다. 이러한 특성과 더불어 SNS 사용이 증가하면서 온라인 상의 개인정보가 노출되고 악용되는 피해 규 모가 커지고 있다. 그러나 개인정보보호를 위한 다중 유형 객체 탐지 기반의 비식별화 기술에 관한 연구는 미흡한 상황 이다. 이에 본 논문은 기존의 단일 유형 객체 탐지 모델을 병렬적으로 이용하여 다중 유형의 객체를 탐지 및 비식별화 하는 인공지능 모델을 제안한다. Cutmix 기법을 통해 사람과 텍스트 객체가 함께 존재하는 이미지를 생성하여 학습 데이터로 구성하고, 사람과 텍스트라는 다른 특징을 가진 객체에 대한 탐지 및 비식별화를 수행하였다. 제안하는 모델은 두 가지 객체가 동시에 존재할 때 0.724의 precision과 0.745의 mAP@.5 를 달성한다. 또한, 비식별화 수행 후 전체 객체 에 대해 mAP@.5 가 0.224로, 0.4 이상의 감소폭을 보였다.

As the Internet and web technology develop around mobile devices, image data contains various types of sensitive information such as people, text, and space. In addition to these characteristics, as the use of SNS increases, the amount of damage caused by exposure and abuse of personal information online is increasing. However, research on de-identification technology based on multi-type object detection for personal information protection is insufficient. Therefore, this paper proposes an artificial intelligence model that detects and de-identifies multiple types of objects using existing single-type object detection models in parallel. Through cutmix, an image in which person and text objects exist together are created and composed of training data, and detection and de-identification of objects with different characteristics of person and text was performed. The proposed model achieves a precision of 0.724 and mAP@.5 of 0.745 when two objects are present at the same time. In addition, after de-identification, mAP@.5 was 0.224 for all objects, showing a decrease of 0.4 or more.

본 논문에서는 레이저스캐너만으로 이루어진 감지 시스템을 이용하여 도로 위에 있는 객체의 위치를 추정하고 분류 하는 알고리즘을 제안한다. 각각의 레이저 스캐너에서 획득한 데이터는 그리드 맵을 사용하여 데이터를 융합하였으며, 팽창 연산과 레이블링 방법을 사용하여 측정 오차를 보정하였다. 추출한 객체의 정보(길이, 폭)를 입력으로 사용한 퍼지 방법을 통해 객체를 보행자, 자전거, 차량으로 분류하였으며, 이러한 방법은 레이저스캐너로만 이루어진 감지 시스템의 정확도를 증가시켰다. 또한 본 논문에서는 실제 도로 환경에서 몇 가지 시나리오를 설정하여 실험을 하였다. 실험을 통 해 감지 시스템이 객체를 정확히 분류하는지, GPS-RTK 장비를 사용하여 획득한 위치 정보와 비교하여 객체의 위치 정 보를 정확히 추정하는지 검증하였다.

This paper proposes the on-road object detection and classification algorithm by using a detection system consisting of only laser scanners. Each sensor data acquired by the laser scanner is fused with a grid map and the measurement error and spot spaces are corrected using a labeling method and dilation operation. Fuzzy method which uses the object information (length, width) as input parameters can classify the objects such as a pedestrian, bicycle and vehicle. In this way, the accuracy of the detection system is increased. Through experiments for some scenarios in the real road environment, the performance of the proposed detection and classification system for the actual objects is demonstrated through the comparison with the actual information acquired by GPS-RTK.

본 논문에서는 금속 환경에서 사용이 가능한 RFID 국제 규격인 910MHz 대역에서 최상의 성능을 갖는 미앤더 형태의 마이크로스트립 패치 안테나를 설계하였다. 안테나에 부착되는 상용 태그 칩과 정합을 위해 사각 형태의 급전부를 본체에 연결하였으며, 사각 형태의 급전부를 본체 안에 위치하였고, 안테나의 복수부인 본체를 효과적으로 축소하기 위하여 패치의 소자를 증가시켜 다중의 미앤더 형태로 설계하였다. 제안된 안테나는 Case 1, Case 2, Case 3로 각각 3 종류의 안테나를 제작하였으며, 안테나의 크기와 접은 횟수에 따른 대역폭, 효율, 인식 거리 등의 특성을 비교․분석하였다. 측정결과 Case 3 안테나의 성능이 가장 좋은 것으로 나타났다. 또한 안테나의 크기와 미앤더 형태로 접은 횟수에 따른 효율 및 이득 특성 변화가 안테나의 인식 거리에 큰 영향을 미치는 것을 확인하였다.

In this paper, we present a design of meander type microstrip patch antenna which has the best quality in RFID approved international standard of 910MHz broadband, usable in metal environment. In order to be fit into the commercial tag chip on the antenna, a square shaped feeder is installed on the main body as well as in the main body. In addition, a multi meander type patching element was designed in order to reduce the main body effectively. Three antennae, Cases 1, 2 and 3, were designed and compared in the areas of broad band width, effectiveness or recognition distance according to their sizes and number of folds. The measurement result of Case 3 was determined to be the best. It was confirmed that mostly the efficiency and characteristic gain change due to antenna size and number of folds in meander type influenced the antenna recognition distance

본 연구에서는 코사인 유사도 기반 객체 추적 알고리즘을 통합한 개선된 행동 인식 방법을 제안한다. 제 안된 시스템은 Faster R-CNN을 사용한 객체 검출, HRNet을 이용한 포즈 추정, 그리고 PoseC3D를 활용한 행동 분류를 기반으로 한다. 핵심 개선 사항은 PoseC3D의 출력에 코사인 유사도 기반 추적 알고리즘을 적용하여 연속 된 프레임에서 각 개별 객체를 일관되게 식별하고 추적하는 것이다. 이를 통해 다양한 인물이 등장하는 야구 영상 에서 투수를 인식할 수 있도록 하였다. 실험을 진행한 결과, 한 사람이 등장하는 야구 영상에서는 97%의 정확도 를 나타내었으며, 시각화를 통해 타자와 투수, (외)야수 등 각 인물을 잘 인식하고 있음을 확인할 수 있었다. 이어 서 다중 인물 중 투수를 인식하여 영상을 추출하는 실험을 진행한 결과, 92%의 정확도로 투수를 인식하였음을 확인할 수 있었다.

In this study, we propose the improved action recognition method that integrates a cosine similarity-based object tracking algorithm. The proposed system is based on object detection using Faster R-CNN, pose estimation using HRNet, and action classification using PoseC3D. The key improvement lies in applying the cosine similarity-based tracking algorithm to the output of PoseC3D, allowing consistent identification and tracking of individual objects across consecutive frames. As a result, the proposed method enables the recognition of the pitcher in baseball videos including multiple objects. Experimental results showed an accuracy of 97% in baseball videos with a single individual. Further experiments on baseball videos with multiple individuals demonstrated a 92% accuracy in recognizing the pitcher.

In Korea, the percentage of multi-unit dwelling among all housing reached to 63% in 2013. Thus, multi-unit dwelling seems to be an universal type of dwelling, in Korea. By the way, in recent, floor noises causes many problems between multi-unit dwellers. A multi-unit dweller committed even a murder for the noises through floor. Of course, the fundamental way to solve floor-noise problems is to force constructors to do floor-noise- proofing work. However, for various reasons, including cost, etc., the preferential way to solve floor-noise problems is to make resort to endurance and consideration between multi-unit dwellers. Nevertheless, many multi-unit dwellers seems to have an outdated way of thinking to use as they please because of their own houses. As a result, they seldom make a concession, and reject a compromise. With this critical mind, this Article is made on three levels. First, this Article begins, in Part Ⅱ, with the examination of requirement for establishment and scope of section for exclusive ownership. Second, Part Ⅲ examines the significance of the sectional ownership and the requirements to establish the sectional ownership. Finally, Part Ⅳ shows how a multi-unit dweller should use his or her section for exclusive ownership, based on neighboring relations.

We have carried out research on high-speed crowd simulation effects through marker-based multi-object augmentation as the basis for a crowd scene processing system that can produce natural movement of a crowd with less resources and operation. When applied to 3D AR images, this method is expected to achieve high efficiency with less mobile resources.

This study addresses the issue of apple detection and segmentation, which plays a crucial role in agricultural automation systems, by employing multi-object detection using machine learning. A Support Vector Machine (SVM) model was used to accurately distinguish apples from leaves, with apple pixels classified as red and leaf pixels as green. The performance of the SVM model was evaluated using various metrics. Key evaluation metrics included IoU (Intersection over Union), Precision, Recall, and mAP (mean Average Precision). The results showed an IoU of 0.48, a Precision of 0.51, a Recall of 0.90, and an mAP of 0.48. Consequently, the SVM model exhibited a high recall rate, successfully detecting most apples, but also had a high false-positive rate due to its low precision. In the future, the need for models that can simultaneously handle real-time processing and accurate boundary recognition is emerging, which could address a critical issue in agricultural automation systems.

Aiming at object occlusions and appearance changes and other issues under complex environment, a new tracking method of multi-cue integration particle filter based on blocking is proposed. Based on fragment the method using color, texture and edge cues integration constructs particle weights updating scheme, and according to the probabilities of which varied information belongs to the object or to the background region, distinguishes appearance changes and occlusions, then the local template updating is applied and the stability of tracking is improved. In order to verify the performance and effectiveness of this method, the experiments use the indoor and outdoor test videos which contained the light changes, occlusions, and the appearance changes of objects. The experimental results showed that the method had strong robustness and high tracking accuracy.