Object Tracking using Adaptive Template Matching

Chantara, Wisarut,Mun, Ji-Hun,Shin, Dong-Won,Ho, Yo-Sung The Institute of Electronics and Information Engin 2015 IEIE Transactions on Smart Processing & Computing Vol.4 No.1

Template matching is used for many applications in image processing. One of the most researched topics is object tracking. Normalized Cross Correlation (NCC) is the basic statistical approach to match images. NCC is used for template matching or pattern recognition. A template can be considered from a reference image, and an image from a scene can be considered as a source image. The objective is to establish the correspondence between the reference and source images. The matching gives a measure of the degree of similarity between the image and the template. A problem with NCC is its high computational cost and occasional mismatching. To deal with this problem, this paper presents an algorithm based on the Sum of Squared Difference (SSD) and an adaptive template matching to enhance the quality of the template matching in object tracking. The SSD provides low computational cost, while the adaptive template matching increases the accuracy matching. The experimental results showed that the proposed algorithm is quite efficient for image matching. The effectiveness of this method is demonstrated by several situations in the results section.

Joint Template Matching Algorithm for Associated Multi-object Detection

( Jianbin Xie ),( Tong Liu ),( Zhangyong Chen ),( Zhaowen Zhuang ) 한국인터넷정보학회 2012 KSII Transactions on Internet and Information Syst Vol.6 No.1

A joint template matching algorithm is proposed in this paper to reduce the high rate of miss-detection and false-alarm caused by the traditional template matching algorithm during the process of multi-object detection. The proposed algorithm can reduce the influence on each object by matching all objects together according to the correlation information among different objects. Moreover, the rate of miss-detection and false-alarm in the process of single-template matching is also reduced based on the algorithm. In this paper, firstly, joint template is created from the information of relative positions among different objects. Then, matching criterion according to normalized cross correlation is generated for multi-object matching. Finally, the proposed algorithm is applied to the detection of watermarks in bill. The experiments show that the proposed algorithm has lower miss-detection and false-alarm rate comparing to the traditional NCC algorithm during the process of multi-object detection.

Improvement of Template Matching for Distance Measurement System Based on Image Sensors

Akio Kita,Yoshinobu Hagiwara,Yongwoon Choi,Kazuhiro Watanabe 제어로봇시스템학회 2012 제어로봇시스템학회 국제학술대회 논문집 Vol.2012 No.10

This paper describes an improved template matching for a distance measurement system based on image sensors using a target. Authors have developed the distance measurement system that consists of two movable image sensors for automatic berthing of ships. The measurement system measures a distance by detecting and tracking the target. We aim at applying the distance measurement system for automatic relative positioning system between a ship and a FPSO. In this application, the shape of the target captured in images is deformed by their relative positions and attitudes and then measurements error is increased due to the target deformation. For solving the deformation, we propose an improved template matching robust to target deformation. Improved template matching is able to detect the deformed target with template database which is created by image conversion with perspective projection of a reference template. By using improved template matching, the distance measurement error is decreased. From the experimental results performed in miniature and indoor environments, we confirmed that the measurement error of the relative distance is decreased by using the database.

영상 인식을 위한 2차원 자동 변형 템플릿 매칭

한영모(Young-Mo Han) 한국산학기술학회 2019 한국산학기술학회논문지 Vol.20 No.9

영상 인식을 위한 한 방법으로 템플릿 매칭이 있다. 기존의 템플릿 매칭에서는 주어진 매칭 영상 내에서 템플릿의 2차원 이동 변위를 바꿔가면서 블록 매칭 알고리즘(BMA)을 수행한다. 이 블록 매칭 알고리즘 수행 중에 템플릿의 크기와 모양은 바뀌지 않는다. 그리고 각각의 2차원 이동변위에 해당하는 블록에서 유사성 척도(similarity measure)로 계산된 매칭 에러 값을 비교하여 대상 체의 위치를 결정한다. 2차원 이동변위만 고려하기 때문에 템플릿과 매칭 영상에서 대상 체의 크기와 방향이 일치하지 않으면 성공률이 떨어진다. 반면 본 논문의 경우는 템플릿의 2차원 방향과 크기를 조정하는 변수를 새로이 추가하고 각각의 2차원 이동 변위에 해당하는 블록에서 이 변수의 최적 값이 자동으로 계산된다. 이렇게 계산된 최적 값을 사용하여, 각 블록에 최적인 템플릿으로 자동 변형된다. 그리고 자동 변형된 템플릿을 기준으로 각 블록의 매칭 에러 값이 계산된다. 이렇게 방향과 크기 차이가 보정된 각 블록의 매칭 에러 값들을 비교하여 대상 체의 위치를 결정한다. 따라서 방향과 크기 차이에 대해 좀 더 안정적인 결과 값을 얻을 수 있다. 사용의 편의를 위해서, 알고리즘을 템플릿 영상 외에 추가의 정보, 예를 들면, 거리정보를 필요로 하지 않는 닫힌 형태로 설계하는 데 주력한다. One method for image recognition is template matching. In conventional template matching, the block matching algorithm (BMA) is performed while changing the two-dimensional translational displacement of the template within a given matching image. The template size and shape do not change during the BMA. Since only two-dimensional translational displacement is considered, the success rate decreases if the size and direction of the object do not match in the template and the matching image. In this paper, a variable is added to adjust the two-dimensional direction and size of the template, and the optimal value of the variable is automatically calculated in the block corresponding to each two-dimensional translational displacement. Using the calculated optimal value, the template is automatically transformed into an optimal template for each block. The matching error value of each block is then calculated based on the automatically deformed template. Therefore, a more stable result can be obtained for the difference in direction and size. For ease of use, this study focuses on designing the algorithm in a closed form that does not require additional information beyond the template image, such as distance information.

Proposal of a Method to Update Template Images for Holistic Matching According to the Width of Character

Michiaki TOYODA,Kazuhisa OBA,Tetsuya YANAGIMOTO,Takatomo MORI,Kazuo HEMMI 제어로봇시스템학회 2009 제어로봇시스템학회 국제학술대회 논문집 Vol.2009 No.8

In this paper, we propose a method to update numeral template images according to the width of characters for holistic matching. The purpose of our study is to develop a method that discriminates between correct patterns and incorrect patterns by matching input images against template images. We applied this method to handwritten class marks of book labels to assist librarians in Japanese libraries to find misshelved books. In experiments, initial numeral template images are gathered randomly from isolated characters. We confirmed that the proposed method is useful to increase the variation of the width of template images.

객체 인식을 위한 SIFT 템플릿 생성 및 매칭 방법

김상철(Sangchul Kim),김소현(Sohyun Kim),김유진(Yoojin Kim),류재석(Jaesug Ryu),낭종호(Jongho Nang) 한국정보과학회 2014 정보과학회 컴퓨팅의 실제 논문지 Vol.20 No.6

최근 카메라를 장착한 하드웨어 기기의 보급으로 영상에서의 객체 인식을 기반으로 한 응용이 늘어나고 있다. 매칭 기반 객체 인식 방법인 SIFT(Scale Invariant Feature Transform)의 경우 1:1 매칭 기반의 방법으로써 비교하고자 하는 대상이 완전히 같은 것이 아닐 경우에 객체의 인식률이 현저히 나빠지는 단점이 존재한다. 이를 해결하기 위해서 해당 객체를 대표하는 템플릿 피쳐를 준비한다면 객체의 인식률이 좋아질 것이다. 이러한 템플릿 피쳐를 제공하기 위하여 샘플 이미지의 피쳐를 클러스터링하여 대표 클러스터를 템플릿 피쳐로 선정한다. 이 때 정확도를 높이기 위하여 TF/IDF 방법을 적용하였으며 인식을 위한 Adaptive Threshold를 적용한 방법을 제안하였다. 실험을 통하여 단순한 SIFT 매칭 방법이나 BoVW방법에 비하여 interclass similarity가 낮은 객체들의 경우 성능이 높게 나옴을 알 수 있었다. Object recognition-based applications have been more required because widespread use of mobile devices equipped with camera has rapidly grown. Object recognition rate of SIFT(Scale Invariant Feature Transform) matching has disadvantage when detecting same objects but slightly different in detail except for scale and rotation because SIFT is one-to-one matching-based object recognition. To solve this problem, we propose SIFT template feature which represents features extracted from an object in many pictures. We perform clustering in order to make SIFT template feature then each centroid of clusters is selected as template features. We conduct TF/IDF likely manner while clustering and Adaptive Threshold matching to elevate recognition accuracy. Experimental results using SIFT template based matching show that object recognition rate of proposed method is more higher than original SIFT matching and BoVW(Bag of Visual Word) using SIFT in such case objects, which has low interclass similarity like brand logo and road sign, are recognized.

심층학습과 템플릿 매칭을 이용한 신속하고 자동화된 지진 분석 체계: 2022년 10월 29일 규모 4.1 괴산 지진 사례 연구

신동훈,서근주,김성진,김보현,홍윤택,변아현 대한지질학회 2023 지질학회지 Vol.59 No.2

As various seismological tasks with deep learning have been implemented, it has been reported that fundamental seismic analysis processes were replaced with deep learning techniques while minimizing the intervention of skilled analysts. In addition, the template matching method using the GPU (Graphic Processing Unit) architecture has enabled effective detection of smaller earthquakes than previously known. In this study, we analyzed a magnitude 4.1 earthquake that occurred in Goesan-gun, South Korea, around 8:27:49 on October 29, 2022, and foreshocks and aftershocks of this earthquake in an automated way using deep learning and template matching techniques. Using deep learning techniques, seismograms from permanent seismic stations within 50 km of the epicenter were used to create an initial earthquake catalog for 11 days from October 21 to October 31, 2022. It was found that a total of 50 events, including 25 events published by the Korea Meteorological Administration, occurred from the morning of October 29th. The focal mechanisms of 10 events, including the mainshock of magnitude 4.1, were automatically determined, and all of them were characterized by a strike-slip faulting mechanism. By cross-correlating waveforms of template events, it was identified that more than 500 micro-earthquakes near the hypocenter occurred for few days from October 29, 2022. These tasks took only about 70 minutes to generate an initial earthquake catalog of 11-day seismograms and about 10 minutes for template matching. To investigate seismic activity in this region, template matching with the same templates was carried out with data from March 18, 2020 to October 20, 2022. The result shows that more than four micro-earthquakes have occurred since May 2020. Therefore, this study suggests that the automatic seismic analysis procedure using deep le t information about earthquakes in the early stages.

Recent Advances in Vision-Based Terrain Relative Navigation in Space Robotic Landing Missions

Mohomad Aqeel Abdhul Rahuman,이규만 제어로봇시스템학회 2024 제어로봇시스템학회 국내학술대회 논문집 Vol.2024 No.7

Planetary descent vehicles operating in distant celestial bodies face challenges in navigation, lacking robust data sources such as GPS commonly available in low Earth orbits. Traditional inertial navigation systems have historically led to a significant accumulation of large inertial navigation errors, often several kilometers off target because of prolonged descent times. The terrain relative navigation (TRN), has emerged as a solution to navigate without reliance on Earth-based data, utilizing observations of the surrounding terrain to provide essential updates, reducing landing errors significantly. This article provides an overview of recent advancements in TRN approaches within space exploration, covering a range of methods including template matching and pattern matching. This review finds template matching and pattern matching as prominent methods in TRN, noting a growing trend of integrating deep learning methodologies combined with other techniques. This integration marks a significant progression in TRN approaches, indicating an increasing adoption of advanced strategies that blend traditional and modern technologies to improve navigation accuracy and reliability.

객체 추적 알고리즘 기반 비전 검사를 통한 압출부 타겟 검출 방법

황정원,이건영,이의철 국제차세대융합기술학회 2023 차세대융합기술학회논문지 Vol.7 No.9

공장 자동화 분야에서 자동 검사 시스템은 기존의 수작업 검사 작업을 대체하여 효율성과 생산성을 향상시키며, 제조업체들은 생산 과정의 품질 관리를 강화하고 불량률을 감소시키는 경제적 이점을 얻을 수 있다. 본연구에서는 공정 라인의 압출부 타겟 검출, 즉 압출부의 정확한 위치를 식별하는 것을 목표로 한다. 일반 RGB 카메라를 사용하여 실제 공정 라인에서 이미지 데이터를 획득하였고, 템플릿 매칭(Template Matching) 알고리즘을활용하여 압출부의 위치를 검출하였다. 평가 지표로는 IoU(Intersection over Union) 임계값이 주어진 경우에 대한 평균 정밀도(Average Precision)를 사용하였다. 본 연구에서 제안한 템플릿 매칭 기반의 압출부 타겟 검출 방법은 기존의 비전 검사 시스템에 비해 경제적이고 간편한 솔루션을 제공하며, 공정의 생산성을 향상시키고 제품의품질 유지에 도움이 될 것으로 기대된다. In the field of factory automation, an automated inspection system replaces manual inspection tasks, leading to improved efficiency and productivity. Manufacturing companies can enhance quality control in the production process and achieve economic benefits by reducing defect rates. This research focuses on the detection of extrusion target positions in a production line. Real-world image data was acquired using a standard RGB camera, and a template matching algorithm was employed to detect the positions of extrusion targets. The evaluation metric used was the average precision for a given Intersection over Union (IoU) threshold. The proposed template matching-based approach for extrusion target detection offers an economical and straightforward solution compared to traditional vision inspection systems. It is expected to enhance production efficiency and contribute to maintaining product quality in the manufacturing process.

형판정합기반 영상 정규화를 통한 고유얼굴(Eigenface)알고리즘 성능개선 방법

최영규,신현금,장경식 한국기술교육대학교 2003 論文集 Vol.10 No.1

A new approach for face recognition is proposed. Our method adopts the Eigenface algorithm as the main classifier, but improves performance by normalizing input images based on template matching technique. Firstly, the two eye regions are evaluated by template matching with a set of ordinary eye templates. The scale and rotation factors are estimated based on the distance and angle between left and right eyes, and we generate a normalized face of the input image, and finally, it is provided as the input of the Eigenface algorithm. Since, Eigenface is a good recognition method but is vulnerable to image variations such as rotation and illumination conditions, our face normalization approach could be very effective.