學術產出-Theses
Article View/Open
Publication Export
-
題名 無人機區域偵查及目標物件定位策略之技術研究
Technical Research on UAV Area Search and Object Geolocation Strategy作者 劉益誠
Liu, Yi-Chen貢獻者 劉吉軒
Liu, Jyi-Shane
劉益誠
Liu, Yi-Chen關鍵詞 智慧無人機
區域偵查
平面視覺
距離預測
三角測量
地理座標投影
區域覆蓋路徑規劃
無人機定位策略
特徵比對
Smart UAV
Area Search
2D vision
Distance Estimation
Triangulation
Geographic Coordinate Projection
Coverage Path Planning
UAV Geolocation Strategy
Feature Matching日期 2023 上傳時間 1-Sep-2023 15:24:38 (UTC+8) 摘要 無人機在現今產業發展快速,因其相較於需人員搭載的飛行器有更低的製造成本、更高的機動性而且減少了駕駛人員傷亡的風險而大量應用於過往需要人力的工作上。無人機發展初期主要應用於軍事用途上,但隨著技術逐漸商用化無人機逐漸在民生用途上取得大量的發展,包含在工業、農業、電影拍攝甚至是競技娛樂都出現了無人機的應用技術。除了無人機硬體本身的發展外,影像處理的技術發展使無人機能在更多應用場景發揮價值,尤其是平面視覺的影像處理技術使無人機能夠以平面視覺的相機進行更多的任務,特別是對於重量有限制而無法搭載大量感測器的微型無人機。對於微型無人機來說平面相機、GPS、指北針與高度計是常備的感測器,因此本研究對於偵查區域內目標物件定位任務以微型無人機常備的感測器發展三項定位策略來達成不同任務環境下的目標偵查與定位。其中,多點平均定位策略以單目視覺定位模組為主,搭配平面視覺影像及感測器數據來達成對地上目標物件的定位,並利用了無人機執行任務的連續性對感測器資料進行校正進而提升定位結果的可靠度。為了降低感測器的依賴程度,本研究以三角計算的方式發展三角測量定位策略,成功降低感測器的依賴度以及高度對於定位準確度的影響。影像比對定位策略則是以特徵比對技術為基礎來達成純影像的定位任務,使無人機在感測器失效的任務環境下仍能夠達成定位任務。
With the rapid development of unmanned aerial vehicle technology and it’shigh mobility, low risk for drone pilot.Unmanned Aerial Vehicle have been usedin a variety of applications.In early stage, UAV was mainly used for militarypurposes.But, with UAV technology became more and more prevalent, UAV widelyapplied on Manufacturing industry, agriculture and film industry.Beside the UAVtechnology, the development of image processing also improve development ofUAV application.And 2D-vision-based image processing was important especiallyfor micro UAV because of it’s weight limit.For micro UAV, the commonly equipped sensors are a camera, GPS, compassand altimeter. Therefore, this research develop three geolocation strategies usingthe sensors commonly found on micro drones to achieve target detection andpositioning in different mission environments. Among them, the multi-pointaveraging geolocation strategy focuses on the monocular visual positioning module,using plane visual images and sensor data to locate ground target.This strategyalso calibrate the sensor data to improve the reliability of the positioning results.To reduce sensor dependency, a triangulation geolocation strategy was developedusing trigonometric calculations, successfully reducing the reliance on sensors andmitigating the impact of altitude on positioning accuracy. The image matchinggeolocation strategy is based on feature matching techniques to achieve pure imagebasedpositioning tasks, enabling drones to perform positioning tasks even inmission environments where sensors may fail.參考文獻 [1] Robot operating system (ros) https://www.ros.org/.[2] Ros-mobile http://wiki.ros.org/ros-mobile.[3] Rodney Brooks. A robust layered control system for a mobile robot. IEEE Journal onRobotics and Automation, 2(1):14–23, 1986.[4] John Canny. A computational approach to edge detection. IEEE Transactions on PatternAnalysis and Machine Intelligence, 8(6):679–698, 1986.[5] D. Comaniciu and P. Meer. Mean shift: a robust approach toward feature space analysis.IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(5):603–619, 2002.[6] Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabinovich. Superpoint: Self-supervised interest point detection and description. Computer Vision and PatternRecognition(CVPR), 2018.[7] Yoav Gabriely and Elon Rimon. Spanning-tree based coverage of continuous areas by amobile robot. International Conference on Robotics and Automation, 2:1927–1933, 2001.[8] Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley,Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks.NeurIPS, 27, 2014.[9] Fatih Gökçe, Göktürk Üçoluk, Erol ̧Sahin, and Sinan Kalkan. Vision-based detectionand distance estimation of micro unmanned aerial vehicles. Sensors, 15(9):23805–23846,2015.[10] Elder M. Hemerly. Automatic georeferencing of images acquired by uav’s. InternationalJournal of Automation and Computing, 11(347–352), 2014.[11] Luc Van Gool Herbert Bay, Tinne Tuytelaars. Surf: Speeded up robust features. ComputerVision–ECCV, 3951, 2006.[12] Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A. Efros. Image-to-image translationwith conditional adversarial networks. International Journal of Computer Vision, 2018.[13] Charles F. F. Karney. Algorithms for geodesics. Journal of Geodesy, 87:43–55, 2013.[14] Ragab Khalil. The accuracy of gis tools for transforming assumed total station surveys toreal world coordinates. Geographic Information System, 5(486-491), 2013.[15] Arturo De la Escalera and Jose María Armingol. Automatic chessboard detection forintrinsic and extrinsic camera parameter calibration. Sensors, 10(3)(2027-2044), 2010.[16] Chaozhen Lan, Wanjie Lu, Junming Yu, and Qing Xu. Deep learning algorithm for featurematching of cross modality remote sensing images. Acta Geodaetica et CartographicaSinica, 50(2):14–23, 2021.[17] Zuoyue Li, Jan Dirk Wegner, and Aurelien Lucchi. Topological map extraction fromoverhead images. International Conference on Computer Vision, 2019.[18] Xiao Ling, Yongjun Zhang, Jinxin Xiong, Xu Huang, and Zhipeng Chen. An imagematching algorithm integrating global srtm and image segmentation for multi-sourcesatellite imagery. Remote Sensing, 8, 2016.[19] D.G Lowe. Distinctive image features from scale-invariant keypoints. InternationalJournal of Computer Vision, 60(91–110), 2004.[20] L. H. Nam, L. Huang, X. J. Li, and J. F. Xu. An approach for coverage path planning foruavs. IEEE 14th International Workshop on Advanced Motion Control, (411-416), 2016.[21] Donggeun Oh and Junghee Han. Smart search system of autonomous flight uavs fordisaster rescue. Sensors, 21(20), 2021.[22] Edwin Olson. Apriltag: A robust and flexible visual fiducial system. InternationalConference on Robotics and Automation, 2011.[23] Parrot. Anafi https://www.parrot.com/en/drones/anafi.[24] Parrot. Bebop2 https://www.parrot.com/en/drones.[25] Shashikant Prasad. pix2pix gan for generating maps given satellite images usingpytorch, https:// medium.com/ @skpd/ pix2pix-gan-for-generating-map-given-satellite-images-using-pytorch-6e50c318673a.[26] Wahyu Rahmaniar, Wen-June Wang, Wahyu Caesarendra, Adam Glowacz, KrzysztofOprz ̨edkiewicz, Maciej Sułowicz, and Muhammad Irfan. Distance measurementof unmanned aerial vehicles using vision-based systems in unknown environments.Electronics, 10(14), 2021.[27] Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. Orb: an efficientalternative to sift or surf. Proceedings of the IEEE International Conference on ComputerVision, (2564-2571), 2011.[28] Sajid Saleem, Abdul Bais, and Robert Sablatnig. Towards feature points based imagematching between satellite imagery and aerial photographs of agriculture land. Computersand Electronics in Agriculture, 126:12–20, 2016.[29] Paul-Edouard Sarlin, Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabinovich.Superglue: Learning feature matching with graph neural networks. Computer Vision andPattern Recognition(CVPR), 2020.[30] Chris Simpson. Behavior trees for ai: How they work, https://www.gamedeveloper.com/programming/behavior-trees-for-ai-how-they-work.[31] Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. Loftr:Detector-free local feature matching with transformers. Computer Vision and PatternRecognition(CVPR), 2021.[32] Dengqing Tang, Tianjiang Hu, Zhaowei Ma, Lincheng Shen, and Chongyu Pan. Apriltagarray-aided extrinsic calibration of camera–laser multi-sensor system. Robotics andBiomimetics, 3(13), 2016.[33] Jinbiao Yuan, Zhenbao Liu, Yeda Lian, Lulu Chen, Qiang An, Lina Wang, and Bodi Ma.Global optimization of uav area coverage path planning based on good point set and geneticalgorithm. Aerospace, 9(2), 2022.[34] Xiaoyue Zhao, Fangling Pu, Zhihang Wang, Hongyu Chen, and Zhaozhuo Xu. Detection,tracking, and geolocation of moving vehicle from uav using monocular camera. IEEEAccess, 7:101160–101170, 2019.[35] 佐翼科技. Dx30-w1 https://www.droxotech.com/.[36] 內政部國土測繪中心. 國土測繪圖資服務雲 https://maps.nlsc.gov.tw.[37] 擎壤科技. Eg2 https://www.earthgen.com.tw/eg2. 描述 碩士
國立政治大學
資訊科學系
110753136資料來源 http://thesis.lib.nccu.edu.tw/record/#G0110753136 資料類型 thesis dc.contributor.advisor 劉吉軒 zh_TW dc.contributor.advisor Liu, Jyi-Shane en_US dc.contributor.author (Authors) 劉益誠 zh_TW dc.contributor.author (Authors) Liu, Yi-Chen en_US dc.creator (作者) 劉益誠 zh_TW dc.creator (作者) Liu, Yi-Chen en_US dc.date (日期) 2023 en_US dc.date.accessioned 1-Sep-2023 15:24:38 (UTC+8) - dc.date.available 1-Sep-2023 15:24:38 (UTC+8) - dc.date.issued (上傳時間) 1-Sep-2023 15:24:38 (UTC+8) - dc.identifier (Other Identifiers) G0110753136 en_US dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/147033 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 資訊科學系 zh_TW dc.description (描述) 110753136 zh_TW dc.description.abstract (摘要) 無人機在現今產業發展快速,因其相較於需人員搭載的飛行器有更低的製造成本、更高的機動性而且減少了駕駛人員傷亡的風險而大量應用於過往需要人力的工作上。無人機發展初期主要應用於軍事用途上,但隨著技術逐漸商用化無人機逐漸在民生用途上取得大量的發展,包含在工業、農業、電影拍攝甚至是競技娛樂都出現了無人機的應用技術。除了無人機硬體本身的發展外,影像處理的技術發展使無人機能在更多應用場景發揮價值,尤其是平面視覺的影像處理技術使無人機能夠以平面視覺的相機進行更多的任務,特別是對於重量有限制而無法搭載大量感測器的微型無人機。對於微型無人機來說平面相機、GPS、指北針與高度計是常備的感測器,因此本研究對於偵查區域內目標物件定位任務以微型無人機常備的感測器發展三項定位策略來達成不同任務環境下的目標偵查與定位。其中,多點平均定位策略以單目視覺定位模組為主,搭配平面視覺影像及感測器數據來達成對地上目標物件的定位,並利用了無人機執行任務的連續性對感測器資料進行校正進而提升定位結果的可靠度。為了降低感測器的依賴程度,本研究以三角計算的方式發展三角測量定位策略,成功降低感測器的依賴度以及高度對於定位準確度的影響。影像比對定位策略則是以特徵比對技術為基礎來達成純影像的定位任務,使無人機在感測器失效的任務環境下仍能夠達成定位任務。 zh_TW dc.description.abstract (摘要) With the rapid development of unmanned aerial vehicle technology and it’shigh mobility, low risk for drone pilot.Unmanned Aerial Vehicle have been usedin a variety of applications.In early stage, UAV was mainly used for militarypurposes.But, with UAV technology became more and more prevalent, UAV widelyapplied on Manufacturing industry, agriculture and film industry.Beside the UAVtechnology, the development of image processing also improve development ofUAV application.And 2D-vision-based image processing was important especiallyfor micro UAV because of it’s weight limit.For micro UAV, the commonly equipped sensors are a camera, GPS, compassand altimeter. Therefore, this research develop three geolocation strategies usingthe sensors commonly found on micro drones to achieve target detection andpositioning in different mission environments. Among them, the multi-pointaveraging geolocation strategy focuses on the monocular visual positioning module,using plane visual images and sensor data to locate ground target.This strategyalso calibrate the sensor data to improve the reliability of the positioning results.To reduce sensor dependency, a triangulation geolocation strategy was developedusing trigonometric calculations, successfully reducing the reliance on sensors andmitigating the impact of altitude on positioning accuracy. The image matchinggeolocation strategy is based on feature matching techniques to achieve pure imagebasedpositioning tasks, enabling drones to perform positioning tasks even inmission environments where sensors may fail. en_US dc.description.tableofcontents 第一章 緒論 1第一節 研究背景 1第二節 研究動機 2第三節 研究目的 4第二章 文獻探討 6第一節 無人機區域偵查與定位 6第二節 區域覆蓋路徑規劃 7第三節 距離估算 7第四節 地理對位 8第五節 影像特徵比對 10第三章 技術架構 12第一節 定位策略設計 12第二節 路徑規劃 14第三節 目標偵測 15第四節 多點平均定位策略 21一、單目視覺定位模組 21二、感測器資訊與影像之映射 24三、GPS 座標推算 25四、校正模組與結果計算 26第五節 三角測量定位策略 28第六節 影像比對定位策略 30一、衛星圖資 30二、特徵比對 31三、區域搜尋 31四、定位模組 36第七節 無人機控制 38一、ROS 節點設計 8二、行為樹設計 38三、無人機控制機制介面 40第四章 實驗及評估 42第一節 評估指標 42第二節 實驗設計 43第三節 實驗結果 46第五章 結論 53第一節 結果分析 53第二節 發展方向 54參考文獻 58 zh_TW dc.format.extent 15973369 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0110753136 en_US dc.subject (關鍵詞) 智慧無人機 zh_TW dc.subject (關鍵詞) 區域偵查 zh_TW dc.subject (關鍵詞) 平面視覺 zh_TW dc.subject (關鍵詞) 距離預測 zh_TW dc.subject (關鍵詞) 三角測量 zh_TW dc.subject (關鍵詞) 地理座標投影 zh_TW dc.subject (關鍵詞) 區域覆蓋路徑規劃 zh_TW dc.subject (關鍵詞) 無人機定位策略 zh_TW dc.subject (關鍵詞) 特徵比對 zh_TW dc.subject (關鍵詞) Smart UAV en_US dc.subject (關鍵詞) Area Search en_US dc.subject (關鍵詞) 2D vision en_US dc.subject (關鍵詞) Distance Estimation en_US dc.subject (關鍵詞) Triangulation en_US dc.subject (關鍵詞) Geographic Coordinate Projection en_US dc.subject (關鍵詞) Coverage Path Planning en_US dc.subject (關鍵詞) UAV Geolocation Strategy en_US dc.subject (關鍵詞) Feature Matching en_US dc.title (題名) 無人機區域偵查及目標物件定位策略之技術研究 zh_TW dc.title (題名) Technical Research on UAV Area Search and Object Geolocation Strategy en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) [1] Robot operating system (ros) https://www.ros.org/.[2] Ros-mobile http://wiki.ros.org/ros-mobile.[3] Rodney Brooks. A robust layered control system for a mobile robot. IEEE Journal onRobotics and Automation, 2(1):14–23, 1986.[4] John Canny. A computational approach to edge detection. IEEE Transactions on PatternAnalysis and Machine Intelligence, 8(6):679–698, 1986.[5] D. Comaniciu and P. Meer. Mean shift: a robust approach toward feature space analysis.IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(5):603–619, 2002.[6] Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabinovich. Superpoint: Self-supervised interest point detection and description. Computer Vision and PatternRecognition(CVPR), 2018.[7] Yoav Gabriely and Elon Rimon. Spanning-tree based coverage of continuous areas by amobile robot. International Conference on Robotics and Automation, 2:1927–1933, 2001.[8] Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley,Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial networks.NeurIPS, 27, 2014.[9] Fatih Gökçe, Göktürk Üçoluk, Erol ̧Sahin, and Sinan Kalkan. Vision-based detectionand distance estimation of micro unmanned aerial vehicles. Sensors, 15(9):23805–23846,2015.[10] Elder M. Hemerly. Automatic georeferencing of images acquired by uav’s. InternationalJournal of Automation and Computing, 11(347–352), 2014.[11] Luc Van Gool Herbert Bay, Tinne Tuytelaars. Surf: Speeded up robust features. ComputerVision–ECCV, 3951, 2006.[12] Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A. Efros. Image-to-image translationwith conditional adversarial networks. International Journal of Computer Vision, 2018.[13] Charles F. F. Karney. Algorithms for geodesics. Journal of Geodesy, 87:43–55, 2013.[14] Ragab Khalil. The accuracy of gis tools for transforming assumed total station surveys toreal world coordinates. Geographic Information System, 5(486-491), 2013.[15] Arturo De la Escalera and Jose María Armingol. Automatic chessboard detection forintrinsic and extrinsic camera parameter calibration. Sensors, 10(3)(2027-2044), 2010.[16] Chaozhen Lan, Wanjie Lu, Junming Yu, and Qing Xu. Deep learning algorithm for featurematching of cross modality remote sensing images. Acta Geodaetica et CartographicaSinica, 50(2):14–23, 2021.[17] Zuoyue Li, Jan Dirk Wegner, and Aurelien Lucchi. Topological map extraction fromoverhead images. International Conference on Computer Vision, 2019.[18] Xiao Ling, Yongjun Zhang, Jinxin Xiong, Xu Huang, and Zhipeng Chen. An imagematching algorithm integrating global srtm and image segmentation for multi-sourcesatellite imagery. Remote Sensing, 8, 2016.[19] D.G Lowe. Distinctive image features from scale-invariant keypoints. InternationalJournal of Computer Vision, 60(91–110), 2004.[20] L. H. Nam, L. Huang, X. J. Li, and J. F. Xu. An approach for coverage path planning foruavs. IEEE 14th International Workshop on Advanced Motion Control, (411-416), 2016.[21] Donggeun Oh and Junghee Han. Smart search system of autonomous flight uavs fordisaster rescue. Sensors, 21(20), 2021.[22] Edwin Olson. Apriltag: A robust and flexible visual fiducial system. InternationalConference on Robotics and Automation, 2011.[23] Parrot. Anafi https://www.parrot.com/en/drones/anafi.[24] Parrot. Bebop2 https://www.parrot.com/en/drones.[25] Shashikant Prasad. pix2pix gan for generating maps given satellite images usingpytorch, https:// medium.com/ @skpd/ pix2pix-gan-for-generating-map-given-satellite-images-using-pytorch-6e50c318673a.[26] Wahyu Rahmaniar, Wen-June Wang, Wahyu Caesarendra, Adam Glowacz, KrzysztofOprz ̨edkiewicz, Maciej Sułowicz, and Muhammad Irfan. Distance measurementof unmanned aerial vehicles using vision-based systems in unknown environments.Electronics, 10(14), 2021.[27] Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. Orb: an efficientalternative to sift or surf. Proceedings of the IEEE International Conference on ComputerVision, (2564-2571), 2011.[28] Sajid Saleem, Abdul Bais, and Robert Sablatnig. Towards feature points based imagematching between satellite imagery and aerial photographs of agriculture land. Computersand Electronics in Agriculture, 126:12–20, 2016.[29] Paul-Edouard Sarlin, Daniel DeTone, Tomasz Malisiewicz, and Andrew Rabinovich.Superglue: Learning feature matching with graph neural networks. Computer Vision andPattern Recognition(CVPR), 2020.[30] Chris Simpson. Behavior trees for ai: How they work, https://www.gamedeveloper.com/programming/behavior-trees-for-ai-how-they-work.[31] Jiaming Sun, Zehong Shen, Yuang Wang, Hujun Bao, and Xiaowei Zhou. Loftr:Detector-free local feature matching with transformers. Computer Vision and PatternRecognition(CVPR), 2021.[32] Dengqing Tang, Tianjiang Hu, Zhaowei Ma, Lincheng Shen, and Chongyu Pan. Apriltagarray-aided extrinsic calibration of camera–laser multi-sensor system. Robotics andBiomimetics, 3(13), 2016.[33] Jinbiao Yuan, Zhenbao Liu, Yeda Lian, Lulu Chen, Qiang An, Lina Wang, and Bodi Ma.Global optimization of uav area coverage path planning based on good point set and geneticalgorithm. Aerospace, 9(2), 2022.[34] Xiaoyue Zhao, Fangling Pu, Zhihang Wang, Hongyu Chen, and Zhaozhuo Xu. Detection,tracking, and geolocation of moving vehicle from uav using monocular camera. IEEEAccess, 7:101160–101170, 2019.[35] 佐翼科技. Dx30-w1 https://www.droxotech.com/.[36] 內政部國土測繪中心. 國土測繪圖資服務雲 https://maps.nlsc.gov.tw.[37] 擎壤科技. Eg2 https://www.earthgen.com.tw/eg2. zh_TW