How To Clean Up Agisoft Point Cloud

Agisoft PhotoScan (ordinarily known as PhotoScan) is a professional tool for a photogrammetry pipeline. It performs photogrammetric processing of digital images and generates 3D spatial data. Agisoft Photoscan is the dominant photogrammetric software on the market, being used by a range of professions, from archaeology, cartographers to creators of virtual worlds and game developers. Unlike Visual SfM, it contains a complete program to numerous specific tasks and dissimilar types of information, and, as a novice to photogrammetric software, was easily managed, with an efficient workflow system.

Digital Artefact: Agisoft PhotoScan – Construction from Movement/Photogrammetry Software

Digital Tools Used: Digital Camera, Agisoft

Skills Needed: Searching for software online. Bones skills in downloading software and installing them onto computer. Workflow organisation. Understanding of dissimilar file formats.

Level: Intermediate with support. I followed several excellent online tutorials in written and video format.

Licence: Professional and Standard (commercial) and Educational (non-commercial)

Possible Use for Community Archaeology Projection: Create complete 3D models of artefacts, georeferenced orthophotos, consign of texture.

Using Agisoft PhotoScan

Photographs

As with Visual SfM, a number of photographs must be taken completing a 360 degrees of the object to be scanned. In this instance, another Ogham Stone from Burnham House. Forty four images in full. This is not a high number for 3D information map, merely at this point, I was more interested in learning the process. For more authentic results, more photographs would have to exist taken.

Agisoft PhotoScan Interface

Beneath is a screen shot of the Agisoft PhotoScan interface. At that place is a toolbar on the top, side window on the left, and the primary model area in grey.

Preferences

Earlier starting, the preferences can be changed. This is not necessary, just could amend the quality of the terminal data fix. You can utilize your graphics card to compute all the algorithms.

Go to Tools > Preferences> OpenCL > Text at the bottom of the dialogue box suggests you lot deactivate 1 CPU core for each GPU in use for optimal performance. Select the box abreast the graphics carte du jour. Lower the Agile CPU cores one bar. In this instance, from 4/4 to three/iv.

You tin besides change the background colour to a darker greyness. This may exist helpful if your points are a light colour and hard to see. Become to Tools > Preferences > Appearance > Change the groundwork color.

Workflow

In the Agisoft Photoscan software I worked through the workflow arrangement. Workflow is the series of activities that are necessary to complete a task; each step in a workflow has a specific footstep before it and a specific pace afterward it, with the exception of the first step [Rouse 2022]. The workflow can be different, depending on the last model you want, and how yous are going to use the model. For enquiry purposes, you may merely need a mesh, merely for visuals and animation, you may need texture. The speed at which the software will process the data set, and the resultant 3D model, will depend on how y'all want to apply the data later you have created the model. This is important as you can set diverse levels such every bit accuracy and quality. This volition exist discussed in more than item as I get though the steps to creating a 3D model of an Ogham Stone.

In Agisoft Photoscan, it is a linear workflow. I will be working with five bones steps:

• Calculation Cameras: The start step is initiated by an exterior result, in this case the calculation of photographs, or cameras. This step uploads all of your images into the Agisoft interface

• Align Photograph: The 2d step is aligning the photos or cameras, to create a thin signal cloud. The calculator program compares the pixels in your photos to observe matches and estimates the camera locations; these are a series of multidimensional points along the surface which give all the necessary information, to render the data properly, that is, to recreate a physical object in 3-dimensional digital infinite. A bespeak cloud is a set of data points in some coordinate organization; in a iii-dimensional coordinate system, these points are usually defined by 10, Y, and Z coordinates, and often are intended to represent the external surface of an object viii. For case, in GIS, geographic information systems, the 10 and Y tin can represent footing coordinates such every bit latitude and longitude, and the Z can represent an elevation. The sparse cloud will contain a sure corporeality of points.

• Dense Deject: In step three, we will create a dumbo point cloud. This will have much higher number of points, or spatial data; therefore having much greater detail. Based on the estimated camera positions the program calculates depth information for each photographic camera to be combined into a single dense point cloud [Agisoft 2022]. With an accurate clean sparse dense deject, this procedure tin be carried out much faster and with better results.

• Build Mesh: The fourth footstep is building a mesh. Afterward dumbo point cloud has been reconstructed information technology is possible to generate a polygonal mesh model based on the dense cloud data [Agisoft 2022]. The 3 coordinates 10,Y,Z, are connected into a triangular face up, which combine seamlessly to produce a continuous mesh over the surface of your model. A typical final photogrammetric product is the polygonal mesh, which can be then used for a range of research purposes and/or used in other software programs such as Blender, Maya, SketchFab. However it is possible to become a step farther and add together texture and colour.

• Build Texture: every bit stated previously, for gathering data sets it is not necessary to build texture, just it will create a photo-realistic 3D model. If you are using the model for animation purposes, creating texture will add to the effects.

Step i: Add together Photos (cameras)

There are 2 means yous can upload photos to the software interface.

1. Workspace > Add Photos

2. Workflow > Add Photos

Select All > Open or Drag and driblet images into workspace.

You will see the following in the Workspace area:

• Workspace refers to your workflow in progress.
• Clamper refers to 1 lot of photos. It is possible to work with two or more chunks. This can be used when you desire to scan both ends of an artefact. In my case its not possible considering the rock is in the ground.

Cameras (o/61 aligned) is the number of photos that are added, that points will be processed from.

Stride 2: Marshal photos

Once all the photos have been added, I clicked on Workflow and selected align photos

Workflow > Select Align Photos. This will open up the Align Photo dialogue box.

I had several choices to brand, depending on the quality of your photos, the quality of output I want, and the speed at which the process takes. The higher the accuracy is set to, creates a higher bespeak limit, and the longer the process will take.Afterwards selecting high, I then clicked on Advanced. In order to shorten the process, I started at key point limit of 40,000, and selected OK.

Matching the points and estimating the camera locations as the photos are aligned is the slowest part of the process, and tin take up to 20 minutes, depending on the quantity and quality of the photos. When the process is finished you will see the sparse dense cloud of the model. The blue planes are the camera positions, the dots are the tie points – unique tracking points identified by the software. You tin can enable/disable the cameras to get a amend view.

Cameras On                             /                                     Cameras Off

Suit Region

Before moving onto the next step, information technology is worth cleaning up the model and re-running the process once more. There are a few different steps I can take to remove actress points. Firstly, I tin can re-size and adjust the box around the model. This box is called a region. The software will only configure the points inside the region box, then by readjusting it, we are cutting out multiple tie-points that we exercise not need.

Cleaning the Model

It is important to clean around the model, to go rid of any unnecessary extra background points. This allows the software to create a better image and rendering, and also speeds up the process. You can do it 2 ways:

1. Select the points y'all desire to delete and press delete on keyboard.
2. Select the points you want to go on and select crop on the toolbar.

Become back to navigation tool, zoom in and delete points y'all don't want.

For a higher resolution, I repeated this stride past re-aligning the photos over again. I repeated the steps mentioned above, that is, Workflow > Align Photos, but this time, prepare the key points to 60,000.

After this process was finished, I repeated the step for a 3rd fourth dimension. I cleaned up the model by deleting points, and readjusted the region box. I increased the primal points to 80,000 and ran the process again.

Beneath are three images of the model. It seems like a long-winded way of working the procedure, just information technology is actually shorter than using the eighty,000 central points straight away. In a scan with this few photos, there might not be much of a difference, but if using a larger quantity of photos, there will be.

Model aligned at 40,000, lx,000 and eighty,000 key points

Pace 2: Build Dense Cloud

Building a dumbo cloud will further improve an build more points to create a better polygon mesh. Before starting this step, it is worth checking again to see if the model needs more tidying upwardly. I zoomed in and out, effectually the model, checking for more stray points and deleting them. Then click on Workflow and select Build Dense Deject

          Workflow > Build Dense cloud

As with the procedure for the alignment of the photos, I had multiple choices, when building the dense cloud. Ultra-high takes the longest to procedure, and isnt necessary unless you lot have a lot of data, and really high quality sharp images. Low takes the shortest amount of time. I selected medium. I left the depth field at aggressive. This is useful if your images are not of the highest quality. If they are, y'all tin can select medium. Select OK, so permit the procedure run. Information technology can take several minutes.

After the dense cloud has been created you can see it past clicking on the dense cloud buttons.

Below are the three types of information currently available to view. Thin Cloud or Tie-points, Dense Cloud and Dense Cloud Classes.

Pace 3: Build Mesh

Build mesh, is the creation of a polygonal mesh model, based on the dense cloud data. Before moving onto step 3, it is worth tidying up the model again. If in that location are points floating around, the mesh will connect these points. The cleaner the model is, the more than accurate and precise the mesh will exist. It is better to accept the time and create a good model than have to repeat the whole process again, because of stray points in the mesh.

After clicking on the Workflow menu, I selected Build Mesh. This opened the Build Mesh dialogue box. In that location is non much y'all demand to modify here except the polygon count. This is number of faces in the resulting mesh, based on the number of points in the dense cloud. 500,000 works well, especially if you are going to utilize your data prepare in 3D software such equally Maya. Interpolation is the default mode, this makes an endeavor to close any gaps in the mesh, if at that place is an area of little or scant point. I then selected OK.

When the process is complete, the 3D model icon, in the shape of a pyramid, tin can exist seen in the Workspace pane. On the main toolbar, in that location is now a option of new icons, again, the pyramid. I clicked on these to view the 3D Model that has at present been created.

Shaded                              /                        Solid                         /                            Wireframe

If there is a hole or gap at whatever bespeak on the model, it could be because of two factors:

1. There were photos or tie-points missing.
2. There was enough photos, but the expanse was likewise like, and the software couldn't find enough unique points.

At this signal you can export the model, if that is all the data you demand, or you can go onto Step 4 and Build Texture.

Step five: Build Texture

The Texture data collected from Agisoft can be exported separately from the 3D model for use in other software such as Maya. There are dissimilar mapping modes, depending on the time to come use of your outputted data. It is important to remember that the lighting in a texture 3D model is not existent lighting, but purely the colour of the texture, at the time the photographs were taken.

To complete this step, I went dorsum to my Workflow Card and selected Build Texture.

As stated above, there are several options depending on future use of texture. I was non looking to save the texture equally a dissever file, just I wanted to understand the process.

Below is an epitome of the model with a texture layer. As mentioned previously, the texture makes for a better looking model, but for the purpose of the Ogham in 3D project, it is not necessary.

I exported the texture data file for possible utilize in the future. I clicked on Tools, in the meridian main menu, and and so export, and export texture. A dialogue box will appear asking where to salve the file. Consign the texture data as a tiff or png, not as a jpeg equally its a destructive format. Select OK.

Step 6: Export Model

To export the texture get to the model in workflow right click export model – select where to relieve – prompted with a dialogue box – consign as a tiff or png, not a jpeg as its a destructive format – click OK.

I saved the information of the 3D model equally a .ply file. A PLY is a computer file format known as the Polygon File Format or the Stanford Triangle Format. It was principally designed to store three-dimensional data from 3D scanners.

For the Ogham in 3D projection, the customs group are emailing the files to Gary Devlin (The Discovery Programme), and Nora White (DIAS). The files are uploaded to MeshLab, an open source arrangement for the processing and editing of unstructured 3D triangular meshes, and SketchFab, a web platform to publish, share and embed interactive 3D files, and finally onto two websites, the Ogham in 3D website and the CorcaDhuibhne3D website.

Screenshot from Ogham in 3D website showing search options

Screenshot of The Discovery Programme on SketchFab showing Ogham stone from Reask, Dingle Pensinula, Co. Kerry

 Screenshot from Corca Dhuibhne 3D new website showcasing all the Ogham Stones in Dingle Pensinula. The above Ogham is from Reask.

Save

Save

Save

Save

Relieve

Relieve

Save

Salve

Save

Save

Save

Salvage

Source: http://fitefuaite.com/digital-artefact/agisoft-photoscan/

Posted by: wrightdeshe1946.blogspot.com

Share this post