Parameterization of a power regression model for aircraft wing mass using a two-criteria estimation method


Аuthors

Korytkin P. G.

Lomonosov Moscow State University, 1, Leninskie Gory, Moscow, 119991, Russia

e-mail: korytkinpg@my.msu.ru

Abstract

Improving the orientation accuracy of robotic systems remains an important task. Existing algorithm-based visual odometry solutions require manual parameter adjustment and are sensitive to illumination and color. A promising approach to improving existing algorithmic visual odometry systems is to integrate neural networks into the image processing stages of visual odometry. This paper examines the application of a neural network built using a combination of the MobileNet V2 and U-Net architectures for image feature extraction. Two datasets were prepared for training the neural network, consisting of four video recordings. First dataset consisted of color frames of the original image scaled to a resolution of 224x224x3; second one consisted of squares of a fixed resolution of 128x128x3 obtained from the original frame image. To obtain a feature map for the datasets, the SIFT algorithm was applied to video images to form a black-and-white map, where black indicates the absence of a feature and white indicates the presence of a feature. Four neural networks were trained on these datasets, two of which were trained on a dataset consisting of 64x64x3 and 128x128x3 resolution segments respectively for which the input images were color. One was trained on segments scaled to 64x64x3 resolution and converted to black-and-white format, and the other with color images of 224x224x3 resolution. The best result in terms of F1 Score was achieved by the neural network detector working with black-and-white 64x64x1 images. A simple visual odometry system was modified to use neural network-based feature detector for testing. A combined visual odometry system with neural network-based feature detector was tested on the KITTI dataset and compared with the original visual odometry system using the SIFT detector. The resulting software solution demonstrated its applicability. It was found that the neural network detector identifies a larger number of features than the SIFT detector. On one of the three KITTI routes used for testing, the neural network detector demonstrated superiority. On the other two routes, drift and error accumulation were detected due to oncoming traffic while the vehicle carrying the camera was stationary.

Keywords:

visual odometry, neural networks, image features, SIFT, MobileNet V2, U Net.

References

  1. Ermakov P. G., Gogolev A. A. Comparative analysis of information integration architectures of strapdown inertial navigation systems for unmanned aerial vehicles. Trudy MAI, 2021, no 117, http://mai.ru//upload/iblock/c31/xon4nnv6t4aum3wqzj4b7kfbsol369la/Ermakov_Gogolev_rus.pdf. DOI: 10.34759/trd-2021-117-11
  2. Maimone М. Autonomous Navigation Results from the Mars Exploration Rover (MER) Mission / M. Maimone, A. Johnson, Y Cheng, R. Willson, L. Matthies // Experimental Robotics IX, 2006. P. 3-12. DOI:10.1007/11552246_1 
  3. Antonov D. A., Zharkov M. V., Kuznetsov I. M., Tchernodoubov A. Y. Vehicle navigation system accuracy and noise immunity improvement techniques. Trudy MAI, 2016, no 90, http://mai.ru//upload/iblock/277/antonov_zharkov_kuznetsov_chernodubov_rus.pdf 
  4. D. Scaramuzza. Tutorial Visual Odometry / D. Scaramuzza, F. Fraundorfer // IEEE ROBOTICS & AUTOMATION MAGAZINE, 2011, Vol. 18, Iss. 4.
  5. D. Nister, O. Naroditsky and J. Bergen, "Visual odometry," Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., Washington, DC, USA, 2004, pp. I-I, doi: 10.1109/CVPR.2004.1315094.
  6. Teed, Zachary and Deng, Jia, “DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras”, Advances in neural information processing systems, 2021.
  7. Olivier Brochu Dufour and Abolfazl Mohebbi and Sofiane Achiche, An Attention-Based Deep Learning Architecture for Real-Time Monocular Visual Odometry: Applications to GPS-free Drone Navigation, 2024 https://arxiv.org/abs/2404.17745
  8. You Might Need to Say Goodbye to Affordable PCs; A Price Hike Storm Is Set to Hit in H2 2026 as Memory Shortages & Windows 10 EOL Collide: // https://wccftech.com/you-might-need-to-say-goodbye-to-affordable-pcs/ (дата обращения: 20.12.2025).
  9. L. Yu, E. Yang, B. Yang, Z. Fei and C. Niu, "A Robust Learned Feature-Based Visual Odometry System for UAV Pose Estimation in Challenging Indoor Environments," in IEEE Transactions on Instrumentation and Measurement, vol. 72, pp. 1-11, 2023, Art no. 5015411, doi: 10.1109/TIM.2023.3279458.
  10. MonoVO-python // Github URL: https://github.com/uoip/monoVO-python (дата обращения: 08.02.2025).
  11. Bradski, G. / The OpenCV Library. Dr. Dobb's Journal of Software Tools. 15.01.2008
  12. Doroshev A.S., Shelomanov D.A. Methodology for selecting hyperparameters of a neural network model in optical navigation tasks. Trudy MAI. 2025. No. 142. (In Russ.). URL: https://trudymai.ru/eng/published.php?ID=185106 
  13. Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, Hartwig Adam, MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications // arXiv.org, 2017, DOI:  https://doi.org/10.48550/arXiv.1704.04861 
  14. Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, Alan L. Yuille, DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs // arXiv.org, 2017 https://doi.org/10.48550/arXiv.1606.00915 
  15. Liang-Chieh Chen, George Papandreou, Florian Schroff, Hartwig Adam, Rethinking Atrous Convolution for Semantic Image Segmentation // arXiv.org, 2017 https://arxiv.org/abs/1706.05587 
  16. M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, L. Chen, MobileNetV2: Inverted Residuals and Linear Bottlenecks // The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 4510-4520, DOI: https://doi.org/10.48550/arXiv.1801.04381
  17. Andrew Howard, Mark Sandler and Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang, Yukun Zhu, Ruoming Pang, Vijay Vasudevan, Quoc V. Le, Hartwig Adam, Searching for MobileNetV3 // arXiv.org, 2019 https://arxiv.org/abs/1905.02244 
  18. Ronneberger, O., Fischer, P., Brox, T. (2015). U-Net: Convolutional Networks for Biomedical Image Segmentation. In: Navab, N., Hornegger, J., Wells, W., Frangi, A. (eds) Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. MICCAI 2015. Lecture Notes in Computer Science(), vol 9351. Springer, Cham. https://doi.org/10.1007/978-3-319-24574-4_28
  19. API Documentation // Tensorflow URL:  https://www.tensorflow.org/api_docs/ (дата обращения: 08.02.2025). 
  20. T. Ganegedara, Tensorflow In Action // MANNING SHELTER ISLAND., 2022
  21. Lowe, D.G. Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision 60, 91–110 (2004). https://doi.org/10.1023/B:VISI.0000029664.99615.94
  22. Taha AA, Hanbury A. Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool. BMC Med Imaging. 2015 Aug 12;15:29. doi: 10.1186/s12880-015-0068-x. PMID: 26263899; PMCID: PMC4533825.
  23. Geiger, P. Lenz and R. Urtasun, "Are we ready for autonomous driving? The KITTI vision benchmark suite," 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, RI, USA, 2012, pp. 3354-3361, DOI: 10.1109/CVPR.2012.6248074.


Download

mai.ru — informational site MAI

Copyright © 2000-2026 by MAI

 Вход