BEGIN:VCALENDAR VERSION:2.0 PRODID:-//pretalx//talks.staging.osgeo.org//foss4g-europe-2024-academic-tra ck//talk//S7VHTK BEGIN:VTIMEZONE TZID:EET BEGIN:STANDARD DTSTART:20000101T000000 RRULE:FREQ=YEARLY;BYMONTH=1;UNTIL=20001231T220000Z TZNAME:EET TZOFFSETFROM:+0200 TZOFFSETTO:+0200 END:STANDARD BEGIN:STANDARD DTSTART:20021027T050000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:EET TZOFFSETFROM:+0300 TZOFFSETTO:+0200 END:STANDARD BEGIN:DAYLIGHT DTSTART:20020331T040000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3 TZNAME:EEST TZOFFSETFROM:+0200 TZOFFSETTO:+0300 END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT UID:pretalx-foss4g-europe-2024-academic-track-S7VHTK@talks.staging.osgeo.or g DTSTART;TZID=EET:20240704T140000 DTEND;TZID=EET:20240704T143000 DESCRIPTION:#### Introduction\nAs part of the CO3D mission (Lebegue et al.\ , 2020)\, carried out in partnership with Airbus\, CNES is developing the image processing chain including the open source photogrammetry pipeline [ CARS](https://github.com/cnes/cars) (Youssefi et al.\, 2020). By acquiring land areas within two years\, providing 4 bands (Blue\, Green\, Red\, Nea r Infra Red) at 50 cm\, the objective is to produce a global Digital Surfa ce Model (DSM) with 1 m relative altimetric error (CE90) at 1 m ground sam pling distance (GSD) as target accuracy. The worldwide production of this 3D information will notably make a real contribution to the creation of di gital twins (Brunet et al.\, 2022). Satellite imagery provides global cove rage\, which unlocks the possibility to update the 3D model of any locatio n on Earth within a rapid time frame. However\, due to the smaller number of images or lower resolution than drone or aerial photography\, a denoisi ng step is necessary to extract relevant 3D information from satellite ima ges. This step smooths out features while retaining their edges that are s ometimes barely recognizable relative to the sensor resolution\, such as t he edges of small houses or the narrow gaps between them as our results sh ow.\n\n#### Geometrically guided and confidence-based point cloud denoisin g\n\nPoint cloud denoising is a topic widely studied in 3D reconstruction: several methods\, ranging from classical to deep learning-based have been designed over the past decades. In this article\, we propose a new method derived from bilateral filtering (Digne and de Franchis\, 2017) integrati ng new constraints. Our aim is to show how a priori knowledge can be used to guide denoising and\, above all\, to produce a denoised point cloud tha t is more consistent with the acquisition conditions or metrics obtained d uring correlation. \n\nThis new method takes into account two important co nstraints. The first is a geometric constraint. The input to the denoising step is a point cloud from photogrammetry resulting from matched points o n the sensor images. Our pipeline CARS derives lines of sight from theses matched points and\, the intersection of these lines give the target 3d po sitions. In our method\, when we denoise this point cloud\, the points are constrained to stay on their initial line of sight. This has two main adv antages: the associated color will remain consistent with the new position and points won't accumulate in certain spaces and create dataless areas.\ n\nThe second constraint comes from the correlator [PANDORA](https://githu b.com/cnes/pandora). The article (Sarrazin et al.\, 2021) describes a conf idence metric\, named ambiguity integral metric\, to assess the quality of the produced disparity map. This measurement determines the level of conf idence associated with each of the points. Each point is moved along the l ine of sight according to its confidence: the less confident the correlato r\, the more the point is moved while respecting the geometric constraint mentioned earlier. Appart from these two added major constraints\, our me thod still uses usual denoising parameters\, such as initial color and pos ition of each considered point regarding its neighborhood. Normal smoothin g is included to compensate correlation inaccuracy.\n\n#### Evaluation and applications\n\nEarly results are extremely promising. A visual compariso n of the mesh obtained before and after our proposed denoising step in a d ense urban area will be provided in the final article (Figure 1). This ill ustration shows that the regularization preserves fine elements and sharp edges and smooths out the flat features (roofs\, facades). Even if we cann ot yet guarantee that denoising will improve the accuracy of the 3D point cloud (or the DSM compared to the airborne LiDAR)\, this verification will be the subject of future work which will be described in the full paper\, we can already affirm that the proposed denoising filter significantly im proves rendering and realism. In fact\, this denoising makes it possible t o enhance roof sections that are barely visible in the denoised point clou d\, thus facilitating the building reconstruction stage for the generation of 3D city models (CityGML). In order to evaluate the quality of the 3D r econstruction on a larger scale\, we plan to use [Lidar HD®](https://geos ervices.ign.fr/lidarhd). This freely distributed data contains 10 points p er m² and includes a semantic label for each point\, allowing for a class -specific quality assessment according to building\, vegetation or ground. We are currently benchmarking state of the art solutions according to met rics that reflect how fine elements are missed in the absence of geometric and confidence constraints.\n\n#### Perspectives\n\nIn future work\, we w ould like to see the potential of adding the constraints proposed in the p aper to other denoising methods\, find out whether it is possible to do th is using deep learning techniques. In addition to comparisons with ground truth\, we would also like to prove that denoising makes it easier to rec onstruct 3D city models\, for example by showing that we can increase the level of detail even with very high resolution satellites such as Pleiades HR. Finally\, with a view to using 3D as a digital twin\, this denoising could be a tool for simplifying 3D models according to specific simulation s. We would therefore like to begin a parameterisation study to quantify t he trade-off between simplicity and quality. DTSTAMP:20260601T212420Z LOCATION:Omicum SUMMARY:Geometrically guided and confidence-based denoising - David Youssef i URL:https://talks.staging.osgeo.org/foss4g-europe-2024-academic-track/talk/ S7VHTK/ END:VEVENT END:VCALENDAR