검색 상세

Enhanced Radar False Alarm Mitigation in Low-RCS Target Detection Using Time-Varying Trajectories on Range-Doppler Diagrams with DCNN

  • 최지연 (서강대학교 일반대학원)

초록 (요약문)

This paper proposes two approaches for detecting low-radar cross section (RCS) targets using the time-varying characteristics in the range Doppler diagram. When low-RCS targets are in cluttered environments, it is not easy to detect them because of the trade-off between the probability of detection and FA rate depending on the detection threshold. To address this, a low detection threshold in the constant false alarm rate (CFAR) algorithm is employed to detect low- RCS targets in high-clutter environments. However, this process often results in the detection of a significant amount of clutter and noise. To overcome this issue, I propose a method to effectively suppress FA using three- dimensional deep convolutional neural networks (3D-DCNN). The 3D-DCNN observes the trajectory of an object over a certain period of time to determine whether the detected object is a target. This proposed algorithm effectively reduces FA while improving overall detection accuracy. However, when only the object's trajectory information is used, false detections can occur in the sidelobe region because the sidelobe area of the object aligns with the object's trajectory. To address this limitation, I propose conditional 3D-DCNN (C-3D-DCNN), which incorporates velocity information from the range-Doppler diagram as conditional input alongside target trajectory information. By dynamically adjusting the Gaussian filter based on the conditional input, the proposed method can effectively distinguish clutter, noise, and sidelobe regions from actual targets. This approach enhances the ability to suppress FA and ensures more accurate target identification in cluttered environments. Key words: conditional-three dimensional-deep convolutional neural networks (C-3D- DCNN), constant false alarm rate (CFAR), drone detection, false alarm (FA) suppression, frequency-modulated continuous wave (FMCW) radar, range-Doppler diagram, three dimensional-deep convolutional neural networks (3D-DCNN).

more

초록 (요약문)

본 학위 논문에서는 range-Doppler diagram의 시간 변화 특성을 이용해 저피탐체 표적을 탐지하는 두 가지 접근법을 제안합니다. 저피탐체 표적이 클러터가 복잡한 환경에 있는 경우, 탐지 임계값에 따라 탐지 확률과 오탐지율 간의 상충관계로 인해 탐지가 어려운 문제가 있다. 높은 클러터 환경에서 저피탐체를 탐지하기 위해서 constant false alarm rate (CFAR)의 낮은 탐지 임계값을 사용한다. 이 과정에서 클러터와 노이즈가 다수 탐지될 수 있다. 이를 해결하기 위해 three-dimensional deep convolutional neural networks (3D-DCNN)을 사용하여 오탐지를 효과적으로 억제하는 방법을 제안한다. 3D-DCNN은 일정 시간 동안 물체의 궤적을 관찰하여 탐지된 물체가 표적인지 여부를 판단하여 제안된 알고리즘은 오탐지를 효과적으로 억제하고 탐지 정확도를 향상시킨다. 물체의 궤적 정보만 이용하게 되면 물체의 사이드로브 영역도 물체의 궤적과 같기 때문에 사이드로브 영역에서의 오탐지가 발생하게 된다. 이를 해결하기 위해 conditional 3D-DCNN (C-3D-DCNN)을 제안한다. C-3D-DCNN은 표적 궤적 정보와 함께 range-Doppler diagram상의 속도 정보를 조건부 입력으로 사용한다. 조건부 입력에 따라 가우시안 필터를 동적으로 조정하여 클러터, 노이즈 및 사이드로브 영역을 표적과 구분하는 방법을 제안한다.

more

목차

1 Extraction of Range-Doppler Information using FMCW Radar 1
1.1 Range-Doppler Diagram Generation Using FMCW Radar 1
1.2 CA-CFAR Detection in Range-Doppler Diagram 3
2 Enhanced Radar False Alarm Mitigation in Low-RCS Target Detection Using Time- Varying Trajectories on Range-Doppler Diagrams with DCNN 5
2.1 Introduction 5
2.2 Simulation Training Data and Classification using 3D-DCNN 10
2.2.1 Simulation Data Generation 11
2.2.2 Classification Using 3D-DCNN 14
2.2.3 Preprocessing for the Whole Range-Doppler Diagram 16
2.2.4 Performance Verification Using the Whole Range-Doppler Diagram 21
2.2.5 Complexity Analysis 24
2.3 Detection of Drone Measured by FMCW Radar 24
2.3.1 Experimental Results Using TI FMCW Radar 24
2.3.2 Experimental Results using TORIS FMCW Radar 29
2.4 Conclusion 34
3 Reducing False Alarm in Target Sidelobes Using Time-Varying Trajectories in Range-Doppler Diagram with Conditional 3D-DCNN 35
3.1 Introduction 35
3.2 Structure of C-3D-DCNN 38
3.3 Simulation Data Generation and Training of C-3D-DCNN 43
3.4 Algorithm Verification using Simulation and Measurements 47
3.4.1 Preprocessing for the Entire Range-Doppler Diagram of C-3D-DCNN 47
3.4.2 Results Using Entire Range-Doppler Diagram Simulation Data 48
3.4.3 Drone Measurements using FMCW Radar 50
3.4.4 Experimental Results 50
3.5 Conclusion 54
Bibliography 56

more
반출 Meta View 목록