DBF 차량용 레이더를 위한 다중 표적의 정확한 각도 추정 연구

이성현(Seong-Hyeon Lee),최인오(In-Oh Choi),김경태(Kyung-Tae Kim) 한국전자파학회 2015 한국전자파학회논문지 Vol.26 No.9

차량용 레이더(automotive radar)는 크기, 무게, 비용 등의 문제로 적은 수의 수신 채널을 가진 안테나를 사용한다. 이때, 스펙트럼 추정 기법 중 하나인 relaxation(RELAX) 기법을 이용하면, 표적의 방위각뿐만 아니라, 빔 내에 존재하는 표적의 개수 추정 또한, 가능하다. 하지만, 차량용 레이더의 적은 수신 채널 개수로 인하여, 정밀하지 않은 각도 영역에서의 스펙트럼 추정이 이루어지며, 이는 부정확한 방위각 추정 성능을 야기한다. 이를 개선하기 위해 본 논문에서는 minimum length description(MDL) 기준을 통한 표적 개수의 추정과 결부한 root-multiple signal classification(root-MUSIC) 기법을 이용하여 표적의 정확한 방위각을 추정하도록 제안한다. 시뮬레이션에서는 세 점 표적들이 안테나 빔 내에 존재할 때, 제안한 방법이 더 나은 방위각 추정 결과를 도출할 수 있음을 보인다. In order to satisfy several conditions with respect to size, weight, and costs, automotive radars use an antenna consisting of a small number of receiving channels. If RELAX technique is applied to the automotive radars, angles of targets located in antenna beam can be estimated as well as the number of the targets. However, a small number of receiving channels in the antenna leads to inaccurate spectral estimation in angle domain, which in turn degrades performance of RELAX technique. Therefore, in this study, root-MUSIC technique coupled with MDL criterion is introduced to decide accurate angles of targets in antenna beam. In simulations, we show superior performance of proposed scheme using simulation results when three point targets are located in antenna beam.

가설검정 및 주값분해를 이용한 개선된 거리측면도 식별 기법

최인오(In-O Choi),정주호(Joo-Ho Jung),김경태(Kyung-Tae Kim),박상홍(Sang-Hong Park) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.1

Range profile (RP) is an one-dimensional (1D) distribution of the scatterer of a target, projected onto the radar line-of-sight. Because RP provides the information of the 1D distribution of the target regardless of the weather and day-night conditions and automated algorithms can be designed for the recognition of RP, it can be used for the recognition purpose very effectively. However, due to the severe fluctuation of RP depending on the aspect angle and the operating frequency, a large amount of training database is required and in addition, a compression method of feature vectors is required to remove redundancies and separate targets in the feature space. This paper proposes an efficient compression method to solve these problems. The proposed method is a combination of the statistical hypothesis testing and the principal component analysis applied to the central moment of RP. In classification simulations using the targets modeled by using the CAD data of the real target, the proposed method provided higher classification results with further compressed data than did the existing method.

기동특성을 사용한 우주표적들 간의 변별 성능 분석

최인오(In-O Choi),강기봉(Ki-Bong Kang),김시호(Si-Ho Kim),정주호(Joo-Ho Jung),김경태(Kyung-Tae Kim),박상홍(Sang-Hong Park) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.11

Detection and tracking of ballistic missile can be possible when the discrimination between space-targets is possible. Since representative space-targets such as warhead and decoy have similar shapes, radar crosssection (RCS) and high resolution range profile are not suitable for discrimination. In this paper, an effective discrimination method based on the ballistic factor is proposed. The proposed method uses Kalman-filter to track a real-time maneuvering trajectory and extracts the ballistic factor from an estimated trajectory at altitudes less than 100km. Since the ballistic factor of warhead is generally larger than decoys, effective discrimination is conducted. In simulations using the RCS values predicted by the physical optics in which targets are engaged in high-maneuver and micro-motion, the proposed method efficiently discriminated targets.

다중각도 거리측면도를 이용한 효율적인 차량식별 연구

최인오(In O Choi),박상홍(Sang-Hong Park) 한국정보기술학회 2013 한국정보기술학회논문지 Vol.11 No.6

Multi-aspect range profile (RP) obtained using a radar can provide more improved automobile identification than the single-aspect RP in terms of the limited factor such as a narrow band. In this paper, we conducted a research on the efficiency of multi-aspect about improved factors to classify the automobile compared with single-aspect using compressed useful features based on principle component analysis (PCA) and a simple nearest neighbor classifier. In simulations using five automobiles composed of 3D scatterers, Multi-aspect of Narrow band lower than 100MHz were provided better classification ratios than single-aspect in factors of SNR, threshold, observation angle and order of PCA.

시간-주파수 영상을 이용한 효과적인 미사일 미세운동 변수 추출 방법

최인오(In-O Choi),김시호(Si-Ho Kim),정주호(Joo-Ho Jung),김경태(Kyung-Tae Kim),박상홍(Sang-Hong Park) 한국전자파학회 2016 한국전자파학회논문지 Vol.27 No.6

현대 전장에서 널리 사용되는 미사일 표적은 작은 레이다 단면적 및 빠른 기동 특성으로 인하여 탐지가 매우 힘들다. 특히 미사일과 유사한 운동 변수로 기동하는 기만체의 존재 때문에 표적의 병진운동 변수들이 아닌 다른 형태의 특성벡터를 추출할 필요가 있다. 본 연구에서는 성공적인 미사일 표적의 식별을 위하여 미세운동을 하는 표적의 미세운동변수 및 산란점 추출을 위한 효과적인 기법을 제안한다. 제안된 기법에서는 모델링 된 미세 도플러 함수와 신호처리 된 수신신호의 시간-주파수 영상과의 정합을 비용함수로 사용하여 미세운동변수들과 산란점들의 위치를 추출하였다. 점 산란원으로 구성된 표적을 이용한 시뮬레이션 결과, 미세운동 및 산란점들의 변수가 정확하게 추출되었다. It is very difficult to intercept the missiles because of the small radar cross-section and the high maneuverability. In addition, due to the decoy with the similar motion parameters, additional features other than those of the translation motion parameters need to be developed. In this paper, for the successful recognition of missiles, we propose an efficient method to extract micro-motion parameters and scatterers of the missile engaged in the micro motion. The proposed method extracts motion parameters and scatterers by using the matching score between the modeled micro-Doppler function and the time-frequency binary image as a cost function. Simulation results using a target composed of the point scatterer show the parameters and the scatterers were accurately extracted.

거리측면도를 이용한 효율적인 차량식별 요인 및 레이더 규격 연구

최인오(In-O Choi),주문갑(Moon-G Joo),박상홍(Sang-Hong Park) 한국정보기술학회 2012 한국정보기술학회논문지 Vol.10 No.7

In case the chirp radar signal of long pulse is used, the high resolution range profile (RP) compressed through matched-filtering can be obtained. The RP is very useful for anti-collision of automobiles and the traffic control because it possesses each unique property according to various cars regardless of weather, or the day and night conditions. In this paper, we conducted a research on the classification factor of automobile RPs using useful features and a nearest neighbor classifier and the required radar specification. In simulations using five automobiles composed of 3D scatterers, classification ratios higher than 80% were achieved when the bandwidth was more than 200MHz, RP threshold higher than 3% of the maximum value, the order of PCA larger than 3 and the observation angle between ?50° and 50°

1차원 중심 모멘트를 이용한 효과적인 ISAR 영상 식별

최인오(In-O Choi),김경태(Kyung-Tae Kim),정주호(Joo-Ho Jung),박상홍(Sang-Hong Park) 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.1

ISAR(The inverse synthetic aperture radar) image which is a two dimensional distribution of the target reflectivity is an effective method to recognize an unknown target. However, automated recognition of ISAR image is very difficult because the image varies significantly depending on the aspect angle between the radar and the target. This paper proposes an efficient method to classify the ISAR image. The proposed method achieves invariance to scale, translation and rotation, small dimensionality, and high discriminative information on the target by a combined use of polar mapping, 1D central moments, and principal component analysis. In simulations using five aircraft models composed of 3D scatterers, the proposed method yielded high classifications with considerably little amount of training database compared with the existing polar mapping classifier.

KOMSAR의 실측데이터를 이용한 표적 식별

최인오 ( In-o Choi ),박상홍 ( Sang-hong Park ) 한국정보처리학회 2013 한국정보처리학회 학술대회논문집 Vol.20 No.1

본 논문에서는 국방과학연구소에서 운용하는 KOMSAR(Korea Miniature Synthetic Aperture Radar)장비로 측정한 실제 항공기의 데이터를 이용하여 효율적인 표적식별을 수행하였다. 표적식별과정은 수신된 모든 데이터에 대하여 거리측면도를 구한 다음 4개의 표적으로 분리한 후, 효과적인 특성벡터를 구성하여 nearest neighbor(NN) 구분기로 표적식별 성능을 수행하였다. 표적식별수행 결과, 높은 구분성능으로 구분이 가능하였다.

칼만 필터를 이용한 효율적인 ISAR Cross-Range Scaling Algorithm

김민(Min Kim),최인오(In-O Choi),정주호(Joo-Ho Jung),김경태(Kyung-Tae Kim),박상홍(Sang-Hong Park) 한국정보기술학회 2018 한국정보기술학회논문지 Vol.16 No.3

In this paper, an efficient method is proposed for the cross-range scaling of the inverse synthetic aperture radar (ISAR) image. To estimate the rotation angle (RA) between two ISAR images, polar mapping was performed based on the center of rotation formed at a specific frequency using a two-dimensional Fourier transform. Because the rotational component of the spectrum is represented by translational motion in the polar coordinate, RA can be estimated by the correlation. The estimated RA from the correlation of the 2D spectra generated during the coherent processing interval is optimized by the Kalman filter. Furthermore, the least square estimation minimizes the estimation error. In simulations using the Mig-25 model consisting of isotropic point scattering sources, ISAR images were properly rescaled in (range, cross-range) domain for various signal-to-noise ratios.

실제 측정된 레이더 반사 신호의 표적식별 연구

최인오(In-O Choi),박상홍(Sang-Hong Park) 한국정보기술학회 2014 한국정보기술학회논문지 Vol.12 No.6

KOMSAR supervised by Agency for Defence Development had been developed domestically as a useful radar equipment which can be applied regardless of the day and night conditions. There were admirable evaluations about military application of KOMSAR manufactured by the pure domestic technology, and both it’s merit and practicality were proved by means of a number of tests. In this paper, we conducted the efficient target recognition by using the returned radar signals of the Boeing747 aircraft measured by KOMSAR. Using the high resolution range profile (HRRP) obtained by matched-filtering, we constructed the database of the four targets by using the four sets of the independent HRRPs from different aspect-angles. In experiments using useful features and classifier, multi-aspect classification ratios higher than 99% were obtained by using the threshold higher than 0.1% of the maximum value.