LogInAlexey Gorinov talked to us about 3D scanning of a birch pole surface and shared a detailed tutorial on the entire workflow, discussing the choice of equipment, camera settings, the shooting process, and post-production.
Introduction
At the beginning of my professional career in 3D modeling, I thought about what direction I would like to take and what my dream profession should be, given the fact that I love playing computer games, traveling, and communicating with people.
I tried many different specializations in 3D, development pipelines, and a ton of software. As a result, I chose the perfect specialization that allows me to do everything I love and make good money – 3D Scan Artist.
A 3D Scan Artist can quickly and efficiently transfer almost any object from the real world to the virtual world – from the smallest (for example, pins) to facades of buildings and structures, which can then be used in advertising, computer games, and even motion films (including CGI effects)! Sounds like magic, doesn't it?
In addition, the profession of 3D Scan Artist is very much in demand today by a wide variety of companies around the world. This will allow you to not only always have a job but also to travel, combining enjoyment and utility!
3D Scan of the Firewood
In this article, I will tell you how to perform a 3D scan of a birch pole surface by photogrammetry in excellent quality at home. I will describe a step-by-step scanning process consisting of eight stages. As a result, we will get a 3D model of firewood in OBJ format with PNG texture maps.
Photogrammetry fundamentally differs from other areas of 3D scanning. To successfully master this method, you will need a number of practical skills rather than theoretical ones. You will not need to know the theory of color and composition, which is usually necessary for 3D Artists.
Photogrammetry requires you to know the basics of photography, the special 3D scanning process, and special, meticulous attention to detail. In the future, I may write a separate article that will help you expand on the knowledge you have gained about 3D scanning objects and help you get started scanning outdoors, as this is a big separate topic.
Selecting Object for Scanning
Step one: Choose a suitable object for scanning.
In general, it is not difficult to choose an object for scanning: it can be any object you see in front of you, in your apartment, or on the street. Do not limit yourself and scan anything that seems interesting and unique to you. However, there are some constraints. For example, you will not be able to scan by photogrammetry:
- transparent objects;
- objects that lack fine details (e.g., dust particles);
- objects painted in the same color.
The essence of these limitations lies at the very heart of the photogrammetry method, which is to find and combine the fine details of an object in a series of photographs. Different methods are used to circumvent these limitations. Still, they all involve applying additional matte surfaces to the object, or in the case of glass, for example, painting and glazing it so that it is no longer transparent.
Equipment Selection
Step two: Choose scanning equipment.
"I'm not rich enough to buy cheap things" is a well-known aphorism, the meaning of which very accurately describes the situation with the choice and purchase of scanning equipment. Naturally, you always want to save money and simplify something, but in our case, it will 100% lead to unnecessary spending of financial resources and loss of precious time. Therefore, I recommend purchasing a special initial set of 3D scanning equipment, which you can then improve and supplement, increasing the convenience and quality of scanning and saving your time, money, and nerves.
So, we will need the following initial set of equipment:
1. Camera
Any camera with a sensor resolution of at least 24 megapixels will do. We are not interested in other characteristics.
As a note. Personally, I use a mirrorless camera Sony Alpha 7 III. A great option to improve the Alpha 7 III would be the next model – Sony Alpha 7RM V with an effective resolution of 60 MP. The increase in resolution will allow you to get excellent quality even as the distance to the subject is increased.
2. Lens
The choice of lens for the camera often has a much greater impact on the final quality of 3D scanning than the choice of the camera itself. In this case, it is important to choose a lens with a fixed focal length because when shooting, it should not be changed in any case, as it will lead to a large number of inconsistencies between the photos and ruin all your work. As a result, for scanning at home, we can use a lens with a small focal length of 28-50 mm, which will also help save flash resources.
Lifehack: Tape the lens swivel mechanism for the duration of the shoot so you won't accidentally twist it.
Tip: To improve quality, I advise you to buy a macro lens with a fixed focal length, which will allow you to photograph small subjects. In the case of large subjects, a macro lens will help you shoot from a favorable close angle. In comparison, conventional lenses have a minimum focal length of about 25-35 cm, which can make it impossible to take quality pictures of small subjects.
3. Flash
It's simple: you buy a Godox Wistro AR400 400WS ring flash and don't think about replacing it. This flash is perfect for 3D scanning as it allows you to create shadowless photos, which is very important to us. It is also lightweight and easy to use.
Tip: Purchase extra batteries so you don't have to stop shooting to recharge them, but you can simply swap them out like a rifle clip.
Important! Your camera may not have a port for an external flash. If this is the case, you will have to buy a special adapter for the camera's «hot shoe» or a radio sync.
4. Polarizing Filters for the Flash
You will need two filters, one for the flash itself and one for the camera. The filters are needed to remove the glare from the flash in your pictures using the cross-polarization effect. When shooting subjects (especially glossy ones), when you change the angle of the shot, you are bound to get constantly changing and moving glare in your pictures, which will lead to texture errors and poor quality of the final model.
Important! You will need to set the correct angle of rotation of the filters before each shot. To do this, you can walk up to the mirror, turn on the illumination mode on the flash, and rotate the filter on the lens until the flashlight on the camera screen disappears or is very hard to see.
Lifehack: The Godox Wistro AR400 400WS doesn't have a filter mount (it doesn't exist in nature), so you'll have to 3D-print it yourself. You can find many ready-made mount files on the Internet. Choose whichever one you like.
Hardware Settings
Step three: Set up the correct equipment settings.
Once you have purchased your initial set of equipment, you need to set it up correctly before you start shooting.
Important! Never change the camera settings while shooting! This will create inconsistencies between the images, and the final scan will be poor.
1. ISO – 100. This setting will ensure high-definition photos without unnecessary noise.
2. Shutter speed – 1/200. This value of camera shutter speed will minimize the influence of external light in the frame, leaving only the light from the flash, as well as remove unwanted effects from accidental camera shake when shooting, which guarantees the absence of blurring in the photos.
3. Aperture – F16. This aperture setting is necessary to achieve a large depth of field in the frame. If some part of the scanned object in the photos is heavily blurred, then after assembling the 3D scan, we will get uneven, broken boundaries of the object, which is absolutely unacceptable.
4. Image format – RAW. Setting the format to RAW will allow us to get cleaner highlighting in the subject photos without color noise, as well as avoid quality loss in any subsequent image conversions (e.g., JPEG conversion).
5. Flash power – 1/4. This flash power setting will be sufficient if the subject to be scanned is close to the camera.
Important! The flash may overheat when shooting a large number of frames in succession. Don't worry, this is normal! Stop shooting, wait for the flash to cool down, and continue.
6. Focus method – continuous. This method ensures that the scanned object is always in focus and it is convenient for you to quickly correct it, if necessary, by selecting another position for focusing. Even a few unfocused, blurred images among high-quality images can spoil the result of your work. Therefore, if you find blurred images, you will have to delete or reshoot them.
7. White balance – constant. It is important to adjust the white balance so that the color in the photos is close to the real one. We will make further adjustments in Adobe Lightroom.
Preparing the Location and Object
Step Four: Prepare the location and object to be scanned before taking the picture.
Now that all the equipment is set up and ready, you can start to prepare the place and object to be scanned. Determine the space in which it will be convenient for you to put a round table with a diameter of about 70-80 cm with free space around it for comfortable movement. On the table, you should put any matte black material, for example, cloth or felt, which will remove the effect of light reflection from the flash on the scanning object.
In the room where you are scanning, you will also need to remove sources of direct harsh light, such as light from a window, making it diffuse if possible. Ideally, you will want a dark room with softboxes giving soft, diffused light.
It's time to place our gorgeous scanning subject on the table! In my case, it will be a birch pole. I want to emphasize that the scanning object can be anything you want, and all the steps we have already followed and will follow will be exactly the same for any of your objects.
For scanning, we will need to take two separate sets of photos of the object – the top and the bottom. Later, we will combine them into one scan.
Important! While taking photos, you need to keep the shape of the scanned object unchanged, so it is important to understand in advance how you will position and turn the object without damaging or deforming its surface. For example, on the surface of my log, there is bark, which easily bends when pressed. Also, if there are parts on the scanning object that move or fall off when you turn it over, it is better to get rid of them immediately or fix them in some way. In my case, it would be advantageous to put the field vertically rather than horizontally so the bark will not bend when you turn it over. Sometimes, you will have to approach this question more inventively if the object to put favorably does not work. In any case, remember that it is better to spend more time on preparation than to correct mistakes later.
Shooting
Step Five: Take a picture of the scanned object.
Now that we have correctly positioned the object of scanning, we can start taking pictures. You will need to scrupulously shoot the entire surface of the object from all possible angles without missing any protruding details and hollows of its relief. If you miss something, then in the best case, the missed places will simply be covered by a smooth mesh or will create a gap in the geometry of the final model.
I am sure that with a little practice, you will develop your own approach to shooting. Later, when you get the hang of it, you will notice that similar objects are scanned in a similar way, so you will soon be able to significantly increase the speed of your scanning.
I recommend using my own personal, already practiced shooting plan, which I think fits most cases.
1. Get as close to the subject as the focal length allows. Try to keep the same distance from the surface of the object when taking pictures and always keep it in focus.
2. Fix your angle and start photographing the subject by walking around it in a circle. Make one detour, moving smoothly in small steps, following the simple rule "take one picture – take a step –change position." Remember that each new shot must include some new detail of the object's relief; otherwise, the shot will be superfluous.
3. Change the shooting angle and repeat the previous point by circling the object, creating a hemisphere-like shape from the photos.
Important! For high-quality scanning, it is necessary to ensure approximately 50-60% overlap between the images for their subsequent combination into a model. On the one hand, you should not skimp on the number of images, as you will not always have the opportunity to take more pictures later, should the need arise. On the other hand, a very large number of images will only increase the time of shooting and their subsequent processing, while the quality of scanning will not improve. Unfortunately, no formula can be used to determine the necessary and sufficient number of shots for a quality scan of any object. But don't worry: with experience, you will get a feel for the right number.
4. If your object has a flat and smooth surface, for example, like a mug or a plate, you can finish taking pictures of the top half. However, in the case of a birch log or other objects with irregular topography, you will need to take additional pictures of small or complex surface details. Take several shots for each such detail, changing the angle of shooting.
5. Carefully turn the object over and perform all the actions described in steps 1-4 with the bottom part of the object.
Lifehack. Before you turn the object over, take a photo of your palm, for example, so that after you finish taking the photo, you don't get confused about where the first half of the photo ends and the second half begins.
Important! Depending on the size and number of objects you are scanning, taking a photo can be very physically demanding and energy-consuming. You may have to do it all day or even several days in a row. So, I advise you to keep yourself in good physical shape. Otherwise, you will quickly get tired and lose concentration, which will have a negative impact on the quality of your shooting.
Photo Processing
Step Six: Process the photos of the object being scanned.
If you manage to get a sufficient number of high-quality photos of the object's surface, taking into account all the details of its relief, then further processing of these photos will be easy. It's time to brew a cup of your favorite coffee or tea and start processing the photos of the object being scanned.
1. Copy the photos to your workstation.
Tip. Organize two separate drives: one for storing finished models and resulting textures and one for photos and the RealityCapture software cache. When building a 3D scan, you will need a disk with high performance and increased capacity. An NVME or SATA SSD drive is the best option.
2. Let's import your subject's photos into a special program to lighten, color correct, and then overwrite from RAW to JPEG format. I recommend using the Adobe Lightroom program.
3. Select the first photo. To lighten the shadows, we move the "Shadows" slider by +100 units; to darken the highlights, we move the “Highlights” slider by -100 units. To adjust the white balance, we move the “White Balance” slider until the photo matches the color of your subject as much as possible. Then, we select all photos and click the “Synchronize Settings” button.
4. Since photos from different angles may be taken at different distances from the subject, the final photo may be exposed differently (darkened or overexposed). To improve the quality of the scan, we will need to adjust the exposure on different groups of photos. Select several photos, then darken or lighten them to achieve the same brightness using the exposure slider.
5. Export all photos to a separate folder in JPEG format with 100% compression quality.
3D Model Assembly
Step Seven: Assemble the 3D model of the object.
The photos are processed and ready for assembling the 3D model of the object. I suggest using the free RealityCapture program for assembly. There are many opinions about which software is the best for this purpose, but most professionals still prefer RealityCapture. In fact, RealityCapture is an established industry standard backed by plenty of reference material, ease of use, and a wide community.
Tip: You can find a great series of tutorials on YouTube from the creators of RealityCapture that will help you learn all the features of this program.
In this article, I won't waste your time telling you about all the features of RealityCapture, but will briefly describe the necessary operations and give some tips.
1. To merge two parts of the object, we need to create masks on the photos so that we would not interfere with the environment that was in the frame. To do this, drag the photos of the first part into the program and start alignment in draft mode. Perhaps not all photos will be combined into one component, it all depends on how correctly you photographed everything. With experience you will understand how to do it.
2. Then, from the created point cloud it is necessary to create a mesh in preview mode. Now, on the resulting model, using lasso and other selection tools, remove everything unnecessary except for the model. In my case, it is the table. Now click the “Depth and Mask” button on the export panel and select the folder with our photos to export the masks. In the export settings window we choose to disable the depth map export, we don't need it. After that, the program will create masks for photos based on the created preview of the model.
3. Repeat steps 1 and 2 for the second part of the object photos.
4. Create a new project in RealityCapture and import photos of both parts. Select all the resulting masks and drag them into the program as well. This way, the program will apply these masks and calculate without including the environment, which we need. Start the final alignment with the "Align" button and wait for the result. Depending on the number of photos and the performance of your workstation, this and the next steps may take different amounts of time – from a few minutes to a few days, so be patient.
5. After creating the point cloud, edit the reconstruction area to make sure that no parts of the model go beyond it. Sometimes, the program thinks that small parts are not part of the model and puts the area wrong, cutting them off. Because of this, you would have to redo the miscalculation or correct it manually, which, of course, is an unnecessary waste of time. Move the reconstruction area sliders until your model is inside the cube.
6. Next, create a mesh using the High Detail button. The mesh can be very dense, containing more than 100 million polygons, so we need to simplify it. Click the Simplify Tool button and simplify it to 20 to 40 million polygons. Sometimes, due to poor quality or lack of necessary images, the mesh may turn out to be too messy and uneven, so it can't be used as a high poly for the normal map. In this case, the mesh will need to be further refined, for example, in ZBrush, using texture and smooth brushes, and then imported back. Create a sweep for the resulting mesh and texture it.
7. The resulting model is processed to a game-ready state. Depending on the project, you may need to retopologize it in a special program, for example, using the Decimation master tool in Zbrush or manually using the Topogun program. After retopology is completed, you need to import the model into the RizomUV program and sweep it.
8. Import the obtained model into RealityCapture and bake the Base Color and Normal textures with the model obtained at the end of step 6 at 8K resolution.
As a result, we have a scanned model ready to be used in games, movies, or for sale on the stock market. If you make a model only for rendering, then you don't need to do steps 7 and 8. That's what I'm going to do this time.
Texture Fixes
Step eight: Fix the texture and improve its quality.
The 3D model of the object obtained in the previous step is likely to have errors that need to be corrected. This is normal because it is very difficult to make a perfectly clean 3D scan. It requires a lot of experience and training.
To correct the textures and improve the quality of the 3D model of the object, use Adobe Substance 3D Painter to correct the errors with the Stamp tool and create a roughness map, and then use Topaz Gigapixel AI to improve the quality of the corrected Base Color texture.
Conclusion
Photogrammetry is a distinctive way of creating 3D models that movie and game makers can't do without if they need ultra-realistic quality. This technology has many advantages, which is what allows it to be in demand.
I hope that after reading this article, newcomers who were previously distrustful of 3D scanning will be excited to try this method. In turn, I hope that experienced 3D scanning artists will also learn something interesting for themselves, which will help them to develop further.
Thank you all for your attention, and happy scanning!