This work investigates the process of automating reconstruction of buildings from video imagery. New metrics were developed to detect the least blurred images in a sequence for further processing. Phase correlation for point matching was investigated and new metrics were developed to identify successful matches. Direct relative orientation algorithms were investigated in-depth. A significant finding was a new 6-point algorithm which outperformed previously published algorithms for a number of calibrated camera and target geometries. The development of the new metrics and the outcomes from the comprehensive investigations conducted have contributed to a better understanding of the challenging problem of automatically reconstructing 3D objects from image sequences.

Till följd av den utveckling som skett hos digitalkameror och programvaror för bildbehandling de senaste åren har intresset för fotobaserad skanning (FBS) vuxit. I denna studie utvärderades FBS som ett alternativ till terrester laserskanning (TLS) för statisk deformationsmätning och relaterade till kavitationsskador på vattenturbinkolvar där volym och maxdjup mäts. En stor del av syftet var att det skulle göras med simpla förutsättningar så att metoden blir enkel att utföra och tillämpbar framförallt i industrimiljöer. Simulerade deformationer skapades på en bräda i trä med en plan omgivande yta och på en boll i kork för att också undersöka en välvd yta. Fotograferingen utfördes med icke-professionella kameror, en digitalkamera och en smartphonekamera. Objekten som undersöktes hade en matt textur, bilderna togs med stor övertäckning (80–100%) och ljusförhållandena var goda för att undvika skuggning och blänk i bilderna. Det som utvärderades i studien var hur många kodade markörer som bör användas som konnektionspunkter för att förbättra bildmatchningen, optimalt antal skalstockar för skalsättning och vilken kameramodell och självkalibreringsmodell som gav den lägsta mätosäkerheten. Bildmatchningen och skapandet av 3D modellerna utfördes i programvaran Agisoft Photoscan. Analysen av modellerna utfördes i programvarorna 3DReshaper och CloudCompare. Resultaten utvärderades mot en skannad referensmodell framtagen med en ROMER Absolute Arm som har möjlighet att uppvisa mätosäkerheter på hundradels mm och kunde därmed anses som “sant” värde i denna studie. Resultatet visade att de kodade markörerna inte var nödvändiga som konnektionspunkter för bildmatchningen. Det visade även att oavsett kamera, självkalibreringsmodell och antal skalstockar (4, 3 & 2) låg alla modeller under den tillåtna avvikelsen i volym och maxdjup som normen för kavitationsskador på vattenturbinkolvar föreskriver (IEC 60609-1, från Swedish Standards Institute). Normen anger att mätosäkerheten på metoden som används inte får överskrida ±15% från det sanna värdet. Den här metoden har här visat sig vara tillämpbar och kan ersätta TLS för dokumentation och avgjutning för uppmätning av statiska deformationer när kraven på en relativ mätosäkerhet är ±5% av det sanna värdet. / As a result of the development of digital cameras and image processing software in recent years, the interest in close-range photogrammetry (CRP) has grown. In this study, CRP was evaluated as an alternative method to terrestrial laser scanning (TLS) for static deformation measurements and it was related to cavitation damage on water turbine pistons where the volume and maximum depth is measured. A large part of the purpose of the study was that it should be done with simple conditions so that the method will be easy to carry out and applies especially in industrial environments. Simulated deformations were created on a wooden board with a flat surrounding surface, and on a ball made of cork. The shooting was done with non-professional cameras, a digital camera and a smartphone camera. The objects that was examined had a matte texture, the images were taken with a large overlap (80–100%) and the lighting conditions were good enough to avoid shading and glare in the pictures. What was evaluated in the study was how many coded markers that should be used to improve image alignment, what the optimal number of scale bars for scaling the model is, and which camera model and self-calibration model that gave the lowest measurement uncertainty. The alignment of the images and creation of the 3D-models of the deformations were performed in the software Agisoft Photoscan. The analysis of the models was performed in the softwares 3DReshaper and CloudCompare. The results were evaluated against a scanned reference model developed with a ROMER Absolute Arm which has the ability to display measurement uncertainties on hundreds of millimetres and could thus be considered as a “true” value in this study. The result showed that the coded markers were not necessary in the image alignment procedure. It also showed that, regardless of which camera, self- calibration model and number of scale bars (4, 3 & 2), all the models created were below the permitted deviation in volume and maximum depth as the standard for cavitation damage on water turbine pistons (IEC 60609-1, from the Swedish Standards Institute) prescribes. The standard states that the measurement uncertainty of the method used must not exceed ±15% from the true value. This method has been found to be applicable and can replace TLS for documentation and volume measurement with a temporary filler for static deformations when the requirements for a relative measurement uncertainty are ±5% from the true value.

Kubhörnsreflektorer används som måltavlor med kontinuerlig och identifierbar reflekterad signalstyrka vid fjärranalys, bland annat för tekniken ”interferometric synthetic aperture radar” [InSAR]. Kubhörnsreflektorer tillämpas exempelvis för bevakning av sättningar i jordytan och kalibrering av [SAR]-system (”synthetic aperture radar”). Hur starkt en kubhörnsreflektor reflekterar satellitsignaler anges med ”radar cross section” [RCS], som minskar vid deformationer såsom avvikelse från ortogonalitet mellan reflektorplåtar, buktighet och ytoregelbundenheter. Därmed är det viktigt att kunna mäta och analysera sådana deformationer. Studiens syfte var att undersöka hur väl fotobaserad skanning [FBS] och terrester laserskanning [TLS] kan användas för att göra deformationsmätningar på kubhörnsreflektorer. En problematik med kubhörnsreflektorer är att ytorna vanligtvis är reflekterande och texturlösa.  Skanningen genomfördes i fältmiljö och FBS gjordes med en systemkamera. FBS-tekniken som användes i studien är baserad på Structure-from-Motion [SfM], vilket automatiserar bildmatchning och 3D-modellering. TLS utfördes med en Leica C10 på kort avstånd, cirka 2 m, från kubhörnsreflektorerna. Insamlade punktmoln segmenterades till separata punktmoln motsvarande de enskilda reflektorplåtarna och referensplan skapades för dessa. Referensplanen användes för att mäta vinklar mellan reflektorplåtar i alla punktmoln, med uppmätta avvikelser från ortogonalitet på 0–0,8°. Buktighet mättes som avstånd mellan plåtarnas punktmoln och referensplan och varierade mycket mellan de två reflektorernas sidor och mellan TLS och FBS, i ett spann från 0 till 6 mm. Ytoregelbundenheter i form av popnitar med storlek 0,6 mm kunde mätas i FBS-punktmoln. Mätosäkerheten var generellt något lägre för deformationsmätningar utifrån TLS jämfört med FBS i studien. Både TLS och FBS har begränsningar vid skanning av kubhörnsreflektorer på grund av reflektorernas ytegenskaper. För FBS kan dessa problem minskas med åtgärder i fält, såsom extra fokuspunkter och artificiell yttextur. TLS-resultat påverkades av infallsvinkeln mot reflektorplåtarna vid skanningen, eftersom en stor infallsvinkel leder till få returer och för liten infallsvinkel riskerar att leda till returer med hög intensitet (och felaktig position). Uppmätt deformation i studien skulle motsvara som mest en förlust på strax över en fjärdedel av det maximala RCS-värdet för den studerade reflektortypen. Den största RCS-förlusten i den här studien berodde på uppmätt buktighet i bottenplåten, i kontrast med att RCS-värdet enligt tidigare studier anses mer känsligt för avvikelse från ortogonalitet mellan reflektorplåtar. / Corner reflectors are used as targets with a continuous and identifiable reflected signal in remote sensing, commonly used with interferometric synthetic aperture radar [InSAR]. Corner reflectors are applied for monitoring crustal changes and calibrating synthetic aperture radar [SAR]-systems. The strength of the reflected radar signal is measured with radar cross section [RCS]. The RCS decreases if the reflector has deformations, such as deviation from orthogonality of the reflector plates, the plate curvature and surface irregularities. Therefore, it is important to be able to measure and analyse these kinds of deformations. The aim of this study was to examine how well close-range photogrammetry [CRP] and terrestrial laser scanning [TLS] can be used to measure deformations of corner reflectors. A problematic aspect of corner reflectors are their surfaces, that usually are reflective and textureless.  Scanning was conducted in a field environment and CRP was performed with a digital camera. The CRP-technique used in this study is based on Structure-from-Motion [SfM], which automates the image matching and 3D-modeling. TLS was done with a Leica C10 at short range from the corner reflector, about 2 m. The point clouds were segmented into separate point clouds for each reflector plate and reference planes were fitted to them. The reference planes were used to measure angles between reflector plates, with measured deviations from orthogonality between 0-0,8°. Plate curvature was measured as the distance from the point cloud to the reference plane and varied between the reflector sides and between TLS and CRP, in an interval from 0 to 6 mm. Surface irregularities in the shape of pop rivets, 0,6 mm in size, could be measured in the CRP-point clouds. Measurement uncertainties were generally lower in measurements based on TLS compared to CRP. Both TLS and CRP have limitations when scanning corner reflectors, caused by surface properties of the corner reflector. These problems can be reduced for CRP with certain field measures, such as extra focus points and artificial surface texture. The TLS results were affected by the incident angle while scanning, since a large incident angle leads to few return pulses and a too small incident angle may lead to returns with high intensity (and incorrect position). Measured deformation in this study would be equivalent to a reduction of RCS slightly above one fourth of the maximum RCS-value for the studied corner reflector type. In contrast to earlier studies, which say that RCS is most sensitive to lack of orthogonality between the plates, the largest reduction of RCS in this study was caused by the measured plate curvature of the bottom plate.

Cette thèse s'intéresse au recalage de données issues de capteurs 3D et multispectraux pour l'étude du patrimoine.Lorsque l'on étudie ce type d'objet, il y a souvent peu de points saillants naturels entre ces jeux de données complémentaires. Par ailleurs, l'utilisation de mires optiques est proscrite.Notre problème est donc de recaler des données multimodales sans points caractéristiques.Nous avons développé une méthode de recalage basé sur le suivi des systèmes d'acquisition en utilisant des techniques issues de la photogrammétrie.Des simulations nous ont permis d'évaluer la précision de la méthode dans trois configurations qui représentent des cas typiques dans l'étude d'objets du patrimoine.Ces simulations ont montré que l'on peut atteindre une précision du suivi de 0.020 mm spatialement et 0.100 mrad angulairement en utilisant quatre caméras 5 Mpx lorsque l'on numérise une zone de 400 mm x 700 mm.La précision finale du recalage repose sur le succès d'une série de calibrations optiques et géométriques, ainsi que sur leur stabilité pour la durée du processus d'acquisition.Plusieurs tests ont permis d'évaluer la précision du suivi et du recalage de plusieurs jeux de données indépendants; d'abord seulement 3D, puis 3D et multispecrales.Enfin, nous avons étendu notre méthode d'estimation de la réflectance à partir des données multispectrales lorsque celles-ci sont recalées sur un modèle 3D. / The concern and interest of this PhD thesis is the registration of featureless 3D and multispectral datasets describing cultural heritage objects.In this context, there are few natural salient features between the complementary datasets, and the use of targets is generally proscribed.We thus develop a technique based on the photogrammetric tracking of the acquisition systems in use.A series of simulations was performed to evaluate the accuracy of our method in three configurations chosen to represent a variety of cultural heritage objects.These simulations show that we can achieve a spatial tracking accuracy of 0.020 mm and an angular accuracy of 0.100 mrad using four 5 Mpx cameras when digitizing an area of 400 mm x 700 mm. The accuracy of the final registration relies on the success of a series of optical and geometrical calibrations and their stability for the duration of the full acquisition process.The accuracy of the tracking and registration was extensively tested in laboratory settings. We first evaluated the potential for multiview 3D registration. Then, the method was used for to project of multispectral images on 3D models.Finally, we used the registered data to improve the reflectance estimation from the multispectral datasets

點雲（point cloud）為以大量三維坐標描述地表實際情形的資料形式，其中包含其三維坐標及相關屬性。通常點雲資料取得方式為光達測量，其以單一波段雷射光束掃描獲取資料，以光達獲取點雲，常面臨掃描時間差、缺乏多波段資訊、可靠邊緣線及角點資訊、大量離散點雲又缺乏語意資訊（semantic information）難以直接判讀及缺乏多餘觀測量等問題。 攝影測量藉由感測反射自太陽光或地物本身放射之能量，可記錄為二維多光譜影像，透過地物在不同光譜範圍表現之特性，可輔助分類，改善分類成果。若匹配多張高重疊率的多波段影像，可以獲取包含多波段資訊且位於明顯特徵點上的點雲，提供光達以外的點雲資料來源。 傳統空中三角測量平差解算地物點坐標及產製數值表面模型（Digital Surface Model, DSM）時，多採用可見光影像為主；而目前常見之高空間解析度數值航照影像，除了記錄可見光波段之外，亦可蒐集近紅外光波段影像。但較少採用近紅外光波段影像，以求解地物點坐標及建立DSM。 因此本研究利用多波段影像所蘊含的豐富光譜資訊，以取像方式簡易及低限制條件的近景攝影測量方式，匹配多張可見光、近紅外光及紅外彩色影像，分別建立可見光、近紅外光及紅外彩色之DSM，其目的在於探討加入近紅外光波段後，所產生的近紅外光及紅外彩色DSM，和可見光DSM之異同；並比較該DSM是否更能突顯植被區。 研究顯示，以可見光點雲為檢核資料，計算近紅外光與紅外彩色點雲的均方根誤差為其距離門檻值之相對檢核方法，可獲得約21%的點雲增加率；然而使用近紅外光或紅外彩色影像，即使能增加點雲資料量，但對於增加可見光影像未能匹配的資料方面，其效果仍屬有限。 / Point cloud represents the surface as mass 3D coordinates and attributes. Generally, these data are usually collected by LIDAR (LIght Detection And Ranging), which acquires data through single band laser scanning. But the data collected by LIDAR could face problems, such as scanning process is not instantaneous, lack of multispectral information, breaklines, corners, semantic information and redundancies. However, photogrammetry record the electromagnetic energy reflected or emitted from the surface as 2D multispectral images, via ground features with different characteristics differ in spectrum, it can be classified more efficiently and precisely. By matching multiple high overlapping multispectral images, point cloud including multispectral information and locating on obvious feature points can be acquired. This provides another point cloud source aparting from LIDAR. In most studies, visible light (VIS) images are used primarily, while calculating ground point coordinates and generating digital surface models （DSM） through aerotriangulation. Although nowadays, high spatial resolution digital aerial images can acquire not only VIS channel, but also near infrared (NIR) channel as well. But there is lack of research doing the former procedures by using NIR images. Therefore, this research focuses on the rich spectral information in multispectral images, by using easy image collection and low restriction close range photogrammetry method. It matches several VIS, NIR and color infrared (CIR) images, and generate DSMs respectively. The purpose is to analyze the difference between VIS, NIR and CIR data sets, and whether it can emphasize the vegetation area, after adding NIR channel in DSM generation. The result shows that by using relative check points between NIR, CIR data with VIS one. First, VIS point cloud was set as check point data, then, the RMSE (Root Mean Square Error) of NIR and CIR point cloud was calculated as distance threshold. Its data increment is 21% ca. However, the point cloud data amount can be increased, by matching NIR and CIR images. But the effect of increasing data, which was not being matched from VIS images are limited.

This diploma thesis deals with the design and implementation of appropriate procedure photogrammetry processing of technical documentation of selected historical artifacts gauges. During processing was calibrated camera, historical gauges were photographed and metrological documented and model viualization was created.

<p dir="ltr">Accurate 3D point clouds are crucial in various fields, and the advancement of software algorithms has facilitated the reconstruction of 3D models from high-quality images. Notably, both single-camera and multi-camera systems have gained popularity in obtaining these images. While single-camera setups offer simplicity and cost-effectiveness, multi-camera systems provide a broader field of view and improved coverage. However, a crucial gap persists, a lack of direct comparison and comprehensive analysis regarding the quality of point clouds acquired from each system. This thesis aims to bridge this gap by evaluating the point cloud quality obtained from both single-camera and multi-camera systems, considering various factors such as lighting conditions, camera settings, and the stability of multi-camera setup in the 3D reconstruction process. Our research also aims to provide insights into how these factors influence the quality and performance of the reconstructed point clouds. By understanding the strengths and limitations of each system, researchers and professionals can make informed decisions when selecting the most suitable 3D imaging approach for their specific applications. To achieve these objectives, we designed and utilized a custom rig with three vertically stacked cameras, each equipped with a fixed camera lens, and maintained uniform lighting conditions. Additionally, we employed a single-camera system with a zoom lens and non uniform lighting conditions. Through noise analysis, our results revealed several crucial findings. The single-camera system exhibited relatively higher noise levels, likely due to non-uniform lighting and the use of a zoom lens. In contrast, the multi-camera system demonstrated lower noise levels, which can be attributed to well-lit conditions and the use of fixed lenses. However, within the multi-camera system, instances of significant instability led to a substantial increase in noise levels in the reconstructed point cloud compared to more stable conditions. Our noise analysis showed the multi-camera system preformed better compared to the single-camera system in terms of noise quality. However, it is crucial to recognize that noise detection also revealed the influence of factors like lighting conditions, camera calibration and camera stability of multi-camera systems on the reconstruction process.</p>

近來三維數值城市及數碼城市（Cyber City）為各界極欲發展及研究的課題，為了要增加三維數值城市的擬真性及美觀程度，通常是將建物模型敷貼真實拍攝之牆面影像，增加三維模型的細緻化程度。而欲精確的敷貼牆面紋理影像，必須嚴密地將所拍攝之近景影像定位定向，一般採用光束法區域平差解算，此時需加上適當的控制點控制資訊才能完成，因此控制點控制資訊若來自地面測量將相當耗費成本。多年來，各地方政府製作大比例尺地形圖時已拍攝相當多的航照影像，可用來做為上述的控制資訊，亦即航空影像控制實體，若能使用這些航空影像控制實體作為控制資訊，不但可有效利用資源，亦能減少控制點取得所需花費的成本。因此，本研究將使用航空影像控制實體所提供的控制資訊做為控制來源。 本研究探討以航空影像控制實體作為控制資訊時，使用非量測型相機以類似傳統航測拍攝方式及旋轉多基線交向拍攝方式拍攝涵蓋建物牆面的目標區影像後，於最少控制且不同控制分布時，對光束法區域平差精度之影響。因使用非量測型相機，故本研究先以iWitnessPRO近景攝影測量軟體率定相機參數，接著以PHIDIAS近景攝影測量軟體解算光束法區域平差。過程中探討使用航空影像控制實體作為控制資訊時，於最少控制且不同控制分布時，加入附加參數解算的自率光束法區域平差與與一般光束法區域平差之精度。根據實驗結果，低樓層取像的光束法區域平差之檢核點RMSE精度，其結果大多可應用於LOD 3精度等級的牆面敷貼。另，因都市地區高樓林立，狹小巷弄多，有鑒於此，本研究使用旋轉多基線交向攝影，結果顯示其將有機會運用於近景攝影測量LOD 3精度等級的牆面紋理敷貼。 / Recently, the studies about the cyber city have become a popular topic. For improving the level of detail of cyber city, photo-realistic textures from images are mapped onto the surfaces of 3D building models. Before the accurate texture mapping, bundle block adjustment can be performed to recover the parameters of exterior orientation for each close-range images more accurate and more precise, where the control information is necessary. For the past years, many aerial photogrammetry projects were done by local governments for the mapping of 1/1000 topographic maps. Those historic aerial images can be used as control information to reduce the cost and increase the efficiency. Therefore, this study investigates the accuracy of bundle block adjustment about non-metric close-range images, taken from the ways similar to the traditional aerial photogrammetry and the rotating multi-baseline photogrammetry, by using control entities from historic aerial images as the minimal controls under various control distributions. Since the non-metric camera is used for collecting the close-range images, the iWitnessPRO software is utilized for camera calibration. After that, the PHIDIAS software, a close-range photogrammetry software, is employed to performed the bundle block adjustment. During performing the bundle block adjustment, the camera parameters are regarded as unknowns and determined, called as self-calibration bundle adjustment. The results of self-calibration bundle adjustment will be compared with conventional bundle adjustment. The test results show that the accuracy of most self-calibration bundle adjustment about close-range images covered with low buildings can be used for the application of LOD 3 texture mapping. Moreover, the test results of using close-range images from rotating multi-baseline photogrammetry in urban areas show the potential possibility for LOD 3 texture mapping in urban areas with high buildings and narrow alleys.