垂直起降无人机视觉助降技术研究

小型垂直起降无人机在野外飞行时，着陆区选择的准确性是影响无人机安全的关键问题。受载荷能力的限制，小型无人机一般只能携带光学载荷。本文利用光学载荷获取的图像信息，开展基于视觉的安全着陆区检测技术研究，创新地提出了一种利用初分割与再评估辅助卷积神经网络分类的无人机安全着陆区检测方法，提升了分类效率和准确率。具体研究内容可以概括为以下几个方面：

本文以机载相机获得的图像数据为基础，使用Canny边缘检测实现图像粗分割。首先根据检测到的边缘点的数量筛选平整的着陆区，使用对数变换操作改善图像光照不均匀的现象，提高阴影区障碍物的识别率。然后对边缘检测结果使用形态学滤波操作，筛选出具有合适大小和形状的平整候选着陆区。

本文将候选的平整着陆区分为安全和危险两类，从传统特征方法和深度学习方法两方面展开地形分类技术研究：结合不同的颜色特征和纹理特征作为多维特征，训练不同的机器学习分类器，分类精度最高可达86.1%；设计了一种深度可分离卷积与Inception结构相结合的卷积神经网络，实现候选着陆区的分类。通过训练学习和分类测试，分类准确率最高可达99.0%，与传统特征方法相比，可进一步提高分类准确度。

采用卷积神经网络进行分类后，复杂水面区域仍存在被错误分类为安全区域的情况。针对此问题，本文提出了一种基于特征点提取技术的着陆区安全性再评估算法。基于水面纹理特征较少的特性，根据特征点的数量采用双阈值机制来区分安全着陆区和水面，实验证明该算法对于误分类的水面识别率达到了95.4%，能够有效降低无人机降落在水面的风险。

最后，本文对安全着陆区检测总体算法流程进行设计，并在无人机实飞拍摄的视频数据集上对整体算法的效果和鲁棒性进行了测试。飞行高度较高时，仅使用卷积神经网络来初步评估边缘检测粗分割得到的区域是否安全，分类准确率为96.5%；当飞机准备着陆进入低空飞行阶段时，使用安全性再评估算法之后使得安全着陆区的识别准确率从89.4%提升到95.5%，并且对于光照，天气和运动模糊等多种干扰因素具有一定鲁棒性。

[1] 刘全波. 基于视觉的无人机自动着陆定位算法[D]. 天津大学, 2016.
[2] 袁俊. 基于无人机平台的目标跟踪和着陆位姿估计[D]. 哈尔滨工业大学, 2020.
[3] Lin Shanggang,Garratt Matthew-A.,Lambert Andrew-J.. Monocular vision-based real-time target recognition and tracking for autonomously landing an UAV in a cluttered shipboard environment[J]. Autonomous Robots, 2016, 41(4): 881-901.
[4] 单一. 基于视觉导航的四旋翼无人机自主着降控制研究[D]. 南京航空航天大学, 2018.
[5] Cabrera-ponce Aldrich-A.,Martinez-carranza Jose. A vision-based approach for autonomous landing[C]//IEEE: IEEE.
[6] Skoczylas Marcin,Walendziuk Wojciech. Multiple Camera Based Real-Time Locating System for Unmanned Air Vehicle[C]//IEEE: IEEE.
[7] 岳文斌. 旋翼无人机自主降落技术研究[D]. 山东大学, 2020.
[8] 康立鹏. 基于视觉导航的小型旋翼无人机目标识别自主降落系统研究[D]. 内蒙古工业大学, 2020.
[9] 贾配洋. 无人机高速移动降落技术研究[D]. 中国科学院大学(中国科学院国家空间科学中心), 2017.
[10] 步青. 无人机视觉智能控制相关技术研究[D]. 浙江工业大学, 2017.
[11] Cabrera-ponce Aldrich-A.,Martinez-carranza Jose. A vision-based approach for autonomous landing[J]. Crossref.
[12] Zhanpeng Gan,Huarong Xu,Yuanrong He, et al. Autonomous Landing Point Retrieval Algorithm for UAVs Based on 3D Environment Perception[C]//2021 Ieee 7th International Conference on Virtual Reality (icvr), 2021: 104-108.
[13] Vishnu R. Desaraju,Nathan Michael,Martin Humenberger, et al. Vision-based landing site evaluation and informed optimal trajectory generation toward autonomous rooftop landing[J]. Autonomous Robots, 2015, 39(3): 445-463.
[14] Michael Warren,Luis Mejias,Xilin Yang, et al. Field and Service Robotics: Results of the 9th International Conference[M]: Springer International Publishing, 2015: 167-181.
[15] Timo Hinzmann,Thomas Stastny,Cesar Cadena, et al. Free LSD: Prior-Free Visual Landing Site Detection for Autonomous Planes[J]. Ieee Robotics and Automation Letters, 2018, 3(3): 2545-2552.
[16] 彭灿兮. 基于视觉的四旋翼无人机无地标着陆方法研究[D]. 湖南科技大学, 2019.
[17] Shen Yufei,Rahman Zia-ur. An Automatic Computer-Aided Detection System for Aircraft Emergency Landing[C]//Infotech@Aerospace 2011, Reston, Virigina: American Institute of Aeronautics and Astronautics, 2011.
[18] Daniel Fitzgerald,Rodney Walker. Classification of Candidate Landing Sites for UAV Forced Landings[C]//Aiaa Guidance, Navigation, and Control Conference and Exhibit, 2005.
[19] Timothy Patterson,Sally McClean,Philip Morrow, et al. Timely autonomous identification of UAV safe landing zones[J]. Image and Vision Computing, 2014, 32(9): 568-578.
[20] Edward Carney,Lina Castano,Huan Xu. Determination of Safe Landing Zones for an Autonomous UAS using Elevation and Population Density Data[C]//Aiaa Scitech 2019 Forum, 2019.
[21] Sebastian Scherer,Lyle Chamberlain,Sanjiv Singh. Online Assessment of Landing Sites[C]//AIAA Infotech@ Aerospace 2010, 2010: 3358.
[22] Hsiu-Wen Cheng,Tsung-Lin Chen,Chung-Hao Tien. Learning-based risk assessment and motion estimation by vision for unmanned aerial vehicle landing in an unvisited area[J]. Journal of Electronic Imaging, 2019, 28(6).
[23] Cesetti A.,Frontoni E.,Mancini A. Autonomous safe landing of a vision guided helicopter[C]//IEEE: IEEE, 2010.
[24] J. P. Matos-Carvalho,Filipe Moutinho,Ana Beatriz Salvado, et al. Static and Dynamic Algorithms for Terrain Classification in UAV Aerial Imagery[J]. Remote Sensing, 2019, 11(21).
[25] Matos-carvalho J.-P.,Moutinho Filipe,Salvado Ana-Beatriz. Static and Dynamic Algorithms for Terrain Classification in UAV Aerial Imagery[J]. Remote Sensing, 11(21): 2501.
[26] Chihiro Kikumoto,Yoh Harimoto,Kazuki Isogaya, et al. Landing Site Detection for UAVs Based on CNNs Classification and Optical Flow from Monocular Camera Images[J]. Journal of Robotics and Mechatronics, 2021, 33(2): 292-300.
[27] 祝思君. 基于深度学习的无人机遥感图像目标识别方法研究[D], 2018.
[28] Lecun Y.,Bottou L.,Bengio Y., Gradient-based learning applied to document recognition[J]. Institute of Electrical and Electronics Engineers (IEEE)(11): 2278-2324.
[29] Alex Krizhevsky,I. Sutskever,G. Hinton. ImageNet Classification with Deep Convolutional Neural Networks[C]//Nips, 2012.
[30] K. Simonyan,A. Zisserman. Very Deep Convolutional Networks for Large-Scale Image Recognition[J]. Computer Science, 2014.
[31] Forrest and Han Song and Moskewicz Matthew and Ashraf Khalid and Dally William and Keutzer Kurt Iandola. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size[J].
[32] A. G. Howard,M. Zhu,B. Chen, et al. MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications[J], 2017.
[33] A. Howard M. Zhu A. Zhmoginov and L. -C. Chen M. Sandler. MobileNetV2: Inverted Residuals and Linear Bottlenecks[C]//2018 Ieee/cvf Conference on Computer Vision and Pattern Recognition, 2018: 4510-4520.
[34] Christian and Wei Liu and Yangqing Jia and Sermanet Pierre and Reed Scott and Anguelov Dragomir and Erhan Dumitru and Vanhoucke Vincent and Rabinovich Andrew Szegedy. Going deeper with convolutions[C]//2015 Ieee Conference on Computer Vision and Pattern Recognition (cvpr), 2015.
[35] Kaiming He,Jian Sun,Xiaoou Tang. Single Image Haze Removal Using Dark Channel Prior[J]. Ieee Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(12): 2341-2353.
[36] 李永健. 基于机器视觉的四旋冀无人机定点着陆系统设计与实现[D]. 华南理工大学, 2015.
[37] 李宇. 无人机视觉着陆关键技术研究[D], 2012.
[38] R. m. haralick K.-Shanmugam-and-I.-Dinstein. Textural Features for Image Classification[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1973, SMC-3: 610-621.
[39] Wang Zhi,Liu Chaoge,Cui Xiang. EvilModel: Hiding Malware Inside of Neural Network Models[C]//IEEE: IEEE.
[40] Nitish Srivastava,Geoffrey Hinton,Alex Krizhevsky, et al. Dropout: A Simple Way to Prevent Neural Networks from Overfitting[J]. Journal of Machine Learning Research, 2014, 15(56): 1929-1958.
[41] K. He,X. Zhang,S. Ren, et al. Deep Residual Learning for Image Recognition[J]. Ieee, 2016.
[42] Lowe D.G.. Object recognition from local scale-invariant features[C]//IEEE: IEEE.
[43] Herbert Bay,Andreas Ess,Tinne Tuytelaars, et al. Speeded-Up Robust Features (SURF)[J]. Computer Vision and Image Understanding, 2008, 110(3): 346-359.
[44] 庄曈. 单目视觉/惯性室内无人机自主导航算法研究[D], 2012.