DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Kim, Soontae | - |
dc.contributor.advisor | 김순태 | - |
dc.contributor.author | Jang, Myeongjae | - |
dc.date.accessioned | 2019-09-04T02:46:25Z | - |
dc.date.available | 2019-09-04T02:46:25Z | - |
dc.date.issued | 2019 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=843537&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/267030 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전산학부, 2019.2,[iv, 52 p. :] | - |
dc.description.abstract | Small-sized Unmanned Aerial Vehicle (UAV) or drone becomes one of fast-growing techniques. Based on these social changes, a radar system for detecting and tracking a drone also becomes important. However, conventional radar systems are not efficient for drones. Characteristics of a drone signal affect to traditional detection algorithms, so they cause low detection accuracy. In this paper, we propose the detection algorithm which uses neural network-based machine learning technique and FMCW radar system. With neural networks, the proposed algorithm can distinguish changeable drone signal and others in real time. Therefore, the radar system can achieve higher accuracy than conventional radar systems. We experiment various detection processes of reflected drone signal with different neural networks, and implement our own appropriate neural network. We compare the proposed detection algorithm with a traditional detection algorithm based on reflected drone signal from FMCW radar system. Results shows that the proposed MLP-based detection algorithm makes 42.1% lower detection fault rate in average. It also achieves 35.7% higher detection accuracy in average with the first experiment. In the second experiment, it achieves 33.0% higher detection accuracy in average with 5MHz ADC sampling rate. For drone tracking, the proposed algorithm can track the drone beat frequency with the similar accuracy to the traditional algorithm. Moreover, the neural network-based detection algorithm takes about 45ms from receiving signal data stream as an input to returning a detection result. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Neural network▼amachine learning▼aFMCW radar▼adrone detection and tracking▼asignal processing | - |
dc.subject | 인공 신경망▼a기계 학습▼aFMCW 레이다▼a드론 탐지 및 추적▼a신호 처리 | - |
dc.title | Neural network-based FMCW radar system for detecting and tracking a drone | - |
dc.title.alternative | 소형 무인기 탐지 및 추적을 위한 인공 신경망 기반 FMCW 레이다 시스템 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전산학부, | - |
dc.contributor.alternativeauthor | 장명재 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.