• 한국통신학회논문지
• /
• 제39C권10호
• /
• pp.957-966
• /
• 2014

Video Stabilization For Vehicular Applications Is An Important Method Of Removing Unwanted Shaky Motions From Unstable Videos. In This Paper, An Improved Video Stabilization Method With Image Stitching Has Been Proposed. Scale Invariant Feature Transform (Sift) Matching Is Used To Calculate The New Position Of The Points In Next Frame. Image Stitching Is Done In Every Frame To Get Stabilized Frames To Provide Stable Video As Well As A Better Understanding Of The Previous Frame'S Position And Show The Surrounding Objects Together. The Computational Complexity Of Sift (Scale-Invariant Feature Transform) Is Reduced By Reducing The Sift Descriptors Size And Resticting The Number Of Keypints To Be Extracted. Also, A Modified Matching Procedure Is Proposed To Improve The Accuracy Of The Stabilization.

• 전자공학회논문지B
• /
• 제33B권6호
• /
• pp.198-207
• /
• 1996

A PSRI (position, scale, and rotation invariant ) feature extraction and automatic target recognition system using a cooperative network and an MLP is proposed. We can extract position invarient features by obtaining the target center using the projection and the moment in preprocessing stage. The scale and rotation invariant features are extracted from the contour projection of the number of edge pixels on each of the concentric circles, which is input to the cooperative network. By extracting the representative PSRI features form the features and their differentiations using max-net and min-net, we can rdduce the number of input neurons of the MLP, and make the resulted automatic target recognition system less sensitive to input variances. Experiments are conduted on various complex images which are shifted, rotated, or scaled, and the results show that the proposed system is very efficient for PSRI feature extractions and automatic target recognitions.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2008년도 하계종합학술대회
• /
• pp.987-988
• /
• 2008

In this paper, we propose a rotation invariant face detection method using Haar-like feature variation. Previous approaches using rectangular features can be calculated very fast. But rectangular features is weak in rotated face. Rotated Haar-like features can get high accuracy, but the performance is slow because it can't use the integral image. Our method vary Haar-like features keeping rectangular. this method makes the performance a bit slow, but gives better accuracy.

• 한국ITS학회 논문지
• /
• 제8권1호
• /
• pp.127-135
• /
• 2009

This paper presents a new technique for corner shape based object retrieval from a database. The proposed feature matrix consists of values obtained through a neighborhood operation of detected corners. This results in a significant small size feature matrix compared to the algorithms using color features and thus is computationally very efficient. The corners have been extracted by finding the intersections of the detected lines found using Hough transform. As the affine transformations preserve the co-linearity of points on a line and their intersection properties, the resulting corner features for image retrieval are robust to affine transformations. Furthermore, the corner features are invariant to noise. It is considered that the proposed algorithm will produce good results in combination with other algorithms in a way of incremental verification for similarity.

• 로봇학회논문지
• /
• 제14권1호
• /
• pp.8-13
• /
• 2019

Visual place recognition is widely researched area in robotics, as it is one of the elemental requirements for autonomous navigation, simultaneous localization and mapping for mobile robots. However, place recognition in changing environment is a challenging problem since a same place look different according to the time, weather, and seasons. This paper presents a feature extraction method using a deep convolutional auto-encoder to recognize places under severe appearance changes. Given database and query image sequences from different environments, the convolutional auto-encoder is trained to predict the images of the desired environment. The training process is performed by minimizing the loss function between the predicted image and the desired image. After finishing the training process, the encoding part of the structure transforms an input image to a low dimensional latent representation, and it can be used as a condition-invariant feature for recognizing places in changing environment. Experiments were conducted to prove the effective of the proposed method, and the results showed that our method outperformed than existing methods.

• 정보처리학회논문지B
• /
• 제17B권5호
• /
• pp.333-344
• /
• 2010

본 논문은 동일한 대상물을 촬영한 영상을 합성하여 파노라마 영상을 생성하는 방법에 대하여 설명한다. 디지털 카메라의 보급으로 파노라마 영상에 대한 관심이 높아지면서 다양한 방법의 파노라마 영상의 제작 방법이 연구되고 있다. 본 논문에서는 크기 및 회전 불변 특징점을 활용하여 파노라마 영상을 합성하는 방법에 대해서 제안한다. 먼저, 입력 영상들에 대해서 특징점을 추출하고, RANSAC 알고리즘을 통해 추출된 특징점을 정합한다. 정합점을 이용하여 투영 변환식을 모델링하고, 모델링된 변환식을 통하여 영상을 정렬하여 파노라마 영상을 생성한다. 제안한 알고리즘은 SURF 특징점 추출 알고리즘을 적용하여 영상의 크기 및 회전 등의 기하학적 변형에 강인하며, 처리 속도도 향상하였다. 실험에서는 기존 Harris corner 검출기나 SIFT 알고리즘을 통해 검출한 특징과 제안한 알고리즘에서 사용된 SURF 알고리즘을 비교 분석 하였고, $640{\times}480$ 크기의 영상을 이용하여 제안한 알고리즘을 통해 파노라마 영상을 합성하였다. 그 결과 파노라마 영상의 합성에 소요되는 시간은 평균0.4초로 나타났고, 기존 알고리즘에 비하여 효율적인 것으로 나타났다.

• 한국컴퓨터정보학회논문지
• /
• 제10권3호
• /
• pp.173-182
• /
• 2005

본 논문에서는 항해 시 위치인식에 사용하기 위하여 카메라의 뷰포인트에 무관한 랜드마크를 인식하는 방법을 제안한다. 기존연구에서 사용된 특징들은 카메라의 뷰포인트에 따라 변하고 이에따른 정보 양의 증가로 위치확인을 위한 시각적인 랜드마크의 추출이 어렵다. 본 논문에서 제안된 방법은 특징 추출단계, 학습과 인식단계, 정합단계의 삼단계로 구성된다. 특징 추출단계에서는 영상의 관심영역을 설정, 이 영역 안에서 코너점을 추출하는데, 추출 시 작은 고유값의 통계적 분석을 통해 보다 정확하고 잡음에 강한 특징을 추출하는 방법을 제안한다. 학습 및 인식단계에서는 5개의 특징점으로 구성된 특징모델이 뷰포인트에 무관한 특징점인지를 검사하여 강건 특징모델을 구성한다. 정합단계에서는 시간 복잡도를 줄이고 정확한 대응점을 산출하기 위하여 유사도 평가함수와 Graham 탐색방법을 이용한 정합 방법을 제안한다. 실험에서는 다양한 실내영상을 가지고 제안한 방법과 기존방법을 비교 분석함으로써 제안한 방법의 우수함을 보였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.2393-2396
• /
• 2005

We present a method to detect on-road vehicle using geometric invariant of feature points on side planes of the vehicle. The vehicles are assumed into a set of planes and the invariant from motion information of features on the plane segments the plane from the theory that a geometric invariant value defined by five points on a plane is preserved under a projective transform. Harris corners as a salient image point are used to give motion information with the normalized correlation centered at these points. We define a probabilistic criterion to test the similarity of invariant values between sequential frames. Experimental results using images of real road scenes are presented.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2006년도 하계종합학술대회
• /
• pp.321-322
• /
• 2006

In this paper, we propose a method that extracts invariant iris codes from user's iris pattern in order to apply these codes to a new cryptographic construct called fuzzy vault. The fuzzy vault, proposed by Juels and Sudan, has been used to manage cryptographic key safely by merging with biometrics. Generally, iris data has intra-variation of iris pattern according to sensed environmental changes, but cryptography requires correctness. Therefore, to combine iris data and fuzzy vault, we have to extract an invariant iris feature from iris pattern. In this paper, we obtain invariant iris codes by clustering iris features extracted by independent component analysis(ICA) transform. From experimental results, we proved that the iris codes extracted by our method are invariant to sensed environmental changes and can be used in fuzzy vault.

• 대한전기학회논문지:전력기술부문A
• /
• 제48권1호
• /
• pp.76-81
• /
• 1999

The degree of relative similarity between 2D patterns is obtained using Open-Ball Scheme. Open-Ball Scheme employs a method of transforming the geometrical information on 3D objects or 2D patterns into the features to measure the relative similarity for object(patten) recognition, with invariance on scale, rotation, and translation. The feature of an object is used to obtain the relative similarity and mapped into [0, 1] the interval of real line. For decades, Moment-Invariant Method has been used as one of the excellent methods for pattern classification and object recognition. Open-Ball Scheme uses the geometrical structure of patterns while Moment Invariant Method uses the statistical characteristics. Open-Ball Scheme is compared to Moment Invariant Method with respect to the way that it interprets two-dimensional patten classification, especially the paradigms are compared by the degree of closeness to human's intuitive understanding. Finally the effectiveness of the proposed Open-Ball Scheme is illustrated through simulations.