Fotogramm-meetria rakendamine pöördprojekteerimises
Laen...
Kuupäev
2025
Kättesaadav alates
11.09.2025
Autorid
Ajakirja pealkiri
Ajakirja ISSN
Köite pealkiri
Kirjastaja
Eesti Maaülikool
Abstrakt
Eesmärk on hinnata fotogramm-meetrilise meetodi sobivust keeruka geomeetriaga objektide kolmemõõtmeliseks rekonstrueerimiseks pöördprojekteerimise kontekstis. Töö võrdleb selle meetodiga loodud mudelite täpsust nendega, mis on loodud professionaalsete 3D-skanneritega (Nikon ModelMaker ja Revopoint RANGE). Samuti uuritakse, kui täpselt suudavad erinevad fotogramm-meetria tarkvarad (Meshroom, RealityCapture, 3DF Zephyr, Agisoft Metashape) luua keeruka geomeetriaga objekti 3D mudeli erinevate fotokogude põhjal.
Töö käigus jäädvustati 200 fotot rahakassast, kasutades Samsung Galaxy S10 nutitelefoni. Kolmemõõtmeliste mudelite loomiseks koostati kolmes mahus pildikogumid (50, 100 ja 200 fotot), mida töödeldi nelja erineva tarkvaraga ja võrreldi referentsmudelitega tarkvaras CloudCompare.
Tulemused näitasid, et fotogramm-meetria suudab piisava fotode hulga ja sobiva tarkvara kasutamisel pakkuda võrreldavat geomeetrilist täpsust professionaalsete 3D-skanneritega. RealityCapture ja Agisoft Metashape andsid parima kvaliteediga mudelid. Mudelite täpsus paranes oluliselt 200 foto kasutamisel võrreldes väiksemate pildikogumitega. Tulemustel on praktiline rakendatavus madalama eelarvega pöördprojekteerimise projektides ning edasised uuringud võiksid keskenduda mudeli skaleerimise ja objekti omaduste mõjule modelleerimise täpsuses. Samuti võiks edasistes töödes uurida suuremate pildikogumite kasutamise mõju mudeli detailsusele ja geomeetrilisele täpsusele.
The objective is to evaluate the suitability of the photogrammetric method for three-dimensional reconstruction of objects with complex geometry in the context of reverse engineering. The study compares the accuracy of models created using this method to those produced with professional 3D scanners (Nikon ModelMaker and Revopoint RANGE). The aim was to assess how accurately different photogrammetry software solutions (Meshroom, RealityCapture, 3DF Zephyr, Agisoft Metashape) can generate a 3D model of a complex object based on different sets of photographs. In the course of the work, 200 photos of a piggy bank were captured using a Samsung Galaxy S10 smartphone. To create 3D models, image sets of three different sizes (50, 100, and 200 photos) were compiled, processed with four different software tools, and compared to reference models using CloudCompare software. The results showed that photogrammetry, with a sufficient number of photos and appropriate software, can provide geometric accuracy comparable to professional 3D scanners. RealityCapture and Agisoft Metashape produced the highest quality models. Accuracy improved significantly when using 200 photos compared to smaller photo sets. The results have practical applicability in lower-budget reverse engineering projects, and future research could focus on the impact of model scaling and object characteristics on modeling accuracy. Additionally, further studies could explore the effect of even larger image sets on the level of detail and geometric accuracy of the models.
The objective is to evaluate the suitability of the photogrammetric method for three-dimensional reconstruction of objects with complex geometry in the context of reverse engineering. The study compares the accuracy of models created using this method to those produced with professional 3D scanners (Nikon ModelMaker and Revopoint RANGE). The aim was to assess how accurately different photogrammetry software solutions (Meshroom, RealityCapture, 3DF Zephyr, Agisoft Metashape) can generate a 3D model of a complex object based on different sets of photographs. In the course of the work, 200 photos of a piggy bank were captured using a Samsung Galaxy S10 smartphone. To create 3D models, image sets of three different sizes (50, 100, and 200 photos) were compiled, processed with four different software tools, and compared to reference models using CloudCompare software. The results showed that photogrammetry, with a sufficient number of photos and appropriate software, can provide geometric accuracy comparable to professional 3D scanners. RealityCapture and Agisoft Metashape produced the highest quality models. Accuracy improved significantly when using 200 photos compared to smaller photo sets. The results have practical applicability in lower-budget reverse engineering projects, and future research could focus on the impact of model scaling and object characteristics on modeling accuracy. Additionally, further studies could explore the effect of even larger image sets on the level of detail and geometric accuracy of the models.
Kirjeldus
Bakalaureusetöö
Tehnika ja tehnoloogia õppekaval
Märksõnad
bakalaureusetööd, 3D-modelleerimine, punktipilv, skaneerimine, võrdlusanalüüs