Multi-Task Learning From Fixed-Wing Uav Images for 2d/3d City Modelling

Bayanlou, Mohammad R.; Khoshboresh-Masouleh, Mehdi

doi:10.5194/isprs-archives-xliv-m-3-2021-1-2021

Cited by 8 publications

(3 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…So, the representative and diverse task-oriented dataset is necessary part of machine learning study. A number of datasets exists (Bastani and Madden, 2021, Mohanty, 2018, Bayanlou and Khoshboresh-Masouleh, 2021, that are available for scientific community interested in semantic segmentation and image vectorization problems.…”

Section: Introductionmentioning

confidence: 99%

Multi-sensor Data Analysis for Aerial Image Semantic Segmentation and Vectorization

Knyaz,

Kniaz,

Zheltov

et al. 2024

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. One of the urgent and constantly in demand problems is updating maps. Maps, representing geo-information in vector form, have undoubted advantages in compactness and ”readability” compared to aerial photographs. The issue of maps actuality is critically important for rational urban planning, precision farming, the relevance of the cadastre and other geospatial applications. Various sources of data are used for maps updating, with aerial imagery being the main and rich source of information. Automatic processing of aerial photographs makes it possible to efficiently extract vector information, providing operational monitoring and accounting for changes that have appeared. The presented study addresses the problem of multi sensor information fusion in order to obtain accurate vector information. We use aerial images as a main data source and additionally the data of laser scanning and ground survey to increase performance of automatic image semantic segmentation and vectorization. The proposed framework is demonstrated on the task of forest monitoring.

show abstract

Section: Introductionmentioning

confidence: 99%

Multi-sensor Data Analysis for Aerial Image Semantic Segmentation and Vectorization

Knyaz,

Kniaz,

Zheltov

et al. 2024

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…After different tests and analysis of the esults, has been shown that it is possible to replace the "classic method" of surveying via UAV platform, thath use of GCPs (Ground Control Points), with a more immediate and less expensive scanning method based on direct "geotag" on board the drone (McMahon et al, 2021;Bayanlou & Khoshboresh-Masouleh, 2021).…”

Section: Introductionmentioning

confidence: 99%

UAS-LC-GNSS: Precision Surveying with a Low-Cost GNSS System for Commercial Drones

Memmolo¹,

Famiglietti²,

Moschillo³

et al. 2023

ROL

View full text Add to dashboard Cite

INGV (National Institute of Geophysics and Volcanology, Italy) is one of the institutions that studies and monitors the geophysical phenomena (earthquakes, volcanic eruptions, landslides, etc.) that occur on Earth. During these events, it is essential to carry out a large, detailed and fast map of the affected areas. If we think of the difficulties encountered during the mapping of the fault sources for the 2016 earthquake in central Italy, we can understand how the UAS (Unmanned Aircraft System) can be a valid "low-cost" alternative to the traditional methods of surveys. These devices, thanks to precision instrumentation such as GNSS (Global Navigation Satellite Systems) receivers and IMU (Inertial Measurement Unit) control units, allow a detailed reconstruction of the investigated areas, especially for small-scale analysis applications. These aircraft are based on multiple technologies and show great investigative capabilities, therefore they must be considered as complete systems. Starting from these concepts, we have developed a low-cost RTK/PPK (Real Time Kinematic/Post Processing Kinematic) GNSS survey system on "commercial" UAVs (Unmanned Aerial Vehicles), i.e. professional drones that are not created to be modified. We have demonstrated how the integration of a GNSS RTK/PPK module on commercial UAVs makes the system efficient for the reconstruction of a highly detailed and precise DEM (Digital Elevation Model), without using GCP (Ground Control Point), allowing to make precision measurements in areas that are difficult to explore and investigate. Indeed, the altimetric trends of the PPK processing without GCPs are perfectly comparable with those of the PVs (Verification Points) deriving from RTK analysis and sshow small acceptable deviations. The height differences between PVs measurements and those deriving from the DEM in the same planar coordinates vary between a minimum of 1 cm and a maximum of 7.8 cm. Based on these results, we can state that the precision mapping with a drone equipped with an on-board GNSS module does not differ much from the technique that involves measuring GCP on the ground, in reality, it is comparable in terms of errors, even on the more difficult field of altitudes.

show abstract

“…Therefore, it is very important to construct a deep learning model suitable for datasets of different sources (or even for different tasks). For example, Bayanlou et al 8 . proposed a common framework and developed a new dataset based on the SAMA-VTOL dataset from a fixed-wing UAV for the performance assessment of multitask learning methods.…”

Section: Introductionmentioning

confidence: 99%

Unsupervised domain adaptation using modified cycle generative adversarial network for aerial image classification

Ren

Jia

Yue

et al. 2022

J. Appl. Rem. Sens.

View full text Add to dashboard Cite

.The unsupervised domain adaptation (UDA) of aerial image classification is of great significance in the traffic management and public safety applications. Due to the differences in the distribution of source and target domain datasets from different sensors and cities, and the acquisition difficulty and high annotation cost of target domain datasets, existing domain adaptive methods have difficulty obtaining high accuracy and stability. In this work, a modified cycle generative adversarial network (CycleGAN) (MGAN) based on the multiscale residual block (MSRB) and multipooling coordinate attention (MPCA) modules is proposed to generate fake images, which is the most critical stage of the UDA method. In the MGAN, the MSRB module can extract the multiscale information of the image, and the MPCA module can fuse the position and channel information in the image. These two modules better extract the features of edges, regions and semantic classes, which makes the generated fake images better retain the information of the source domain and possess the style of the target domain. Otherwise, the Deeplabv3+ model is used to obtain a pretrained and final segmentation model in the first and third stages. In the experiment, the proposed method is validated on the aerial images of the Potsdam and Vaihingen datasets in the International Society for Photogrammetry and Remote Sensing (ISPRS) two-dimensional semantic labeling benchmark and improves overall accuracy from 52% to 60% on the segmentation of the target domain, compared with the existing optimal method; it also achieves an accuracy of 77% for buildings. The results demonstrate that the proposed method can achieve a higher accuracy and stability.

show abstract

Multi-Task Learning From Fixed-Wing Uav Images for 2d/3d City Modelling

Cited by 8 publications

References 17 publications

Multi-sensor Data Analysis for Aerial Image Semantic Segmentation and Vectorization

Multi-sensor Data Analysis for Aerial Image Semantic Segmentation and Vectorization

UAS-LC-GNSS: Precision Surveying with a Low-Cost GNSS System for Commercial Drones

Unsupervised domain adaptation using modified cycle generative adversarial network for aerial image classification

Contact Info

Product

Resources

About