В рамках проекта требовалось быстро и с высокой точностью воссоздать городскую территорию площадью 1,132 км2.
Съемка осуществлялась специализированной камерой Vexcel UtraCam на высоте 1200 м. Всего было обработано 142 000 фотографий (14 000 плановых и 128 000 перспективных) — это 6500 гигапикселей и 23 терабайта данных.
Размер плановых фотографий — 210 МПикс, перспективных — 80 МПикс.
Использовалось около 100 опорных точек, применяемых чтобы привязать съемку к географической системе координат и повысить точность модели.
Обработка осуществлялась в сетевом режиме, одновременно на шести компьютерах. Процесс занял около трех месяцев и происходил без участия человека.
The ContextCapture Master is the main user interface of ContextCapture. This graphical user interface, it allows you to:
Import the data sets,
Define the processing settings,
Monitor the progress of submitted tasks,
Visualize results, etc.
The Master does not perform the processing tasks. Instead, it separates tasks into elementary jobs which it submits to a job queue.
ContextCapture Master’s main interface manages the different steps of the ContextCapture workflow through a project.
A project is organized along a tree structure. It contains items of different types, corresponding to each step of the workflow:
Project: A project manages all data relative to a scene processed by ContextCapture. It contains one or several blocks as sub-items.
Block: A block manages a set of input photos and their properties (photogroup properties: sensor size, focal length, principal point, lens distortion / pose: position, rotation), based on which one or several reconstructions can be created. These reconstructions are represented as sub-items of the block in the tree structure.
Reconstruction: A reconstruction manages a 3D reconstruction framework (spatial reference system, region-of-interest, tiling, retouching, processing settings), based on which one or several productions can be launched. These productions are represented as sub-items of the reconstruction in the tree structure.
Production: A production manages the generation of a 3D model, with error feedback, progress monitoring and notifications about updates of the underlying reconstruction (e.g. retouching).
A project can contain multiple items corresponding to a same step of the workflow, which allows complex versioning and/or variant management. This is very useful to experiment on a same scene with different input data and different processing settings.
The main interface is in the form of a project explorer from which you can browse all the items of a project.
ContextCapture Master Main Interface
ContextCapture Engine is the worker module of ContextCapture.It runs on a computer in the background, without user interaction. When it is not busy, the Engine takes a pending job in the queue, depending on its priority and date of submission, and executes it. A job usually consists of an Aerotriangulation or 3D reconstruction process, using various computationally intensive algorithms (keypoint extraction, automatic tie point matching, bundle adjustment, dense image matching, robust 3D reconstruction, seamless texture mapping, texture atlas packing, level-of-detail generation …).
ContextCapture Engine makes an extensive use of general-purpose computation on graphics processing units (GPGPU). Each Engine can exploit a single GPU.
Starting/Ending the engine
Click on the ContextCapture Engine desktop shortcut to start the engine.
ContextCapture Engine window
Once running, ContextCapture Engine listens to the Job queue directory which is configured in the ContextCapture Settings.
To close ContextCapture Engine, simply close the engine console. Any running Job will be moved back into the job queue, with a pending status and its original priority. Pending jobs will remain in the Job queue, waiting to be processed by a next execution of ContextCapture Engine.
Title:?Bentley Acute3D ContextCapture Center 4.1
Acute3D, a Bentley Systems company, has released 4.1 version of ContextCapture Center, its a new solution developed on the basis of the former Smart3DCapture technology, a software solution allowing to produce high resolution 3D models from simple photographs, without any human intervention.
The principle of ContextCapture is to analyze several photographs of a static subject, taken from different viewpoints, and to automatically detect pixels corresponding to a same physical point. From many such correspondences, relative orientations of photos and accurate 3D shape of the scene can be inferred.
ContextCapture solves this problem with cutting edge photogrammetry, computer vision and computational geometry algorithms fulfilling industrial-quality requirements in terms of precision, scalability, efficiency, usage, robustness and interoperability.
ContextCapture 4.1 release notes:
– Fisheye camera support.
– New camera database (camera model, user database, user interface).
– New user interface to import positions/rotations from text files.
– Video frame import.
– Aerotriangulation algorithm improvement (frustum, new input data min/max viewing distance, block type nadir).
– Texturing improvement (texturing per sub-triangle).
– New command line tool to export a unique LOD mesh from a 3D reconstruction composed of several tiles.Migration to ContextCapture for Smart3DCapture users
All Smart3DCapture Advanced/Expert and Ultimate users, with a valid subscription service, are entitled to migrate to an equivalent ContextCapture Edition. Please contact your Acute3D representative or your local Bentley office to learn more about the procedure.
Acute3D is a technological software company created in 2011 by Jean-Philippe Pons and Renaud Keriven. Acute3D?s cutting edge technology originates from 25 man-years of research done at two major French research institutes, ?cole des Ponts ParisTech and Centre Scientifique et Technique du B?timent.
We are proud to announce that the Acute3D team has joined Bentley Systems, Incorporated, the leading company dedicated to providing comprehensive software solutions for sustaining infrastructure.