In the upcoming pages, you will discover best practices and recommendations for gathering visual data for 3D model reconstruction, aligned with MYX's standards.
One3D is compatible with a variety of camera types, including drones, mobile phones, compact digital cameras, DSLRs, fisheye lenses, photogrammetric cameras, and multi-camera systems. It can process both still photographs and individual frames extracted from digital video cameras. However, it does not support linear push-broom cameras, or rolling shutter cameras when capturing fast-motion scenes.
Recommended DJI drones:
Mavic 2 Pro
Mavic 3/Pro/Enterprise
Inspire 2 (with ZENMUSE X7, ZENMUSE X5S, or ZENMUSE X4S gimbal)
Inspire 3
Matrice 300/RTK (Zenmuse P1 gimbal)
Matrice 350
Matrice 30
Suggested DSLRs:
Sony APS-C cameras (a6000/6300/6500)
Sony A7/A7R III/IV
Canon Mark 3/4
Nikon D3
When evaluating camera options, prioritize key features such as higher resolution, larger pixel size, lower crop factor, and superior low-light performance.
While there is no specific minimum camera resolution requirement for One3D, using a higher resolution camera enables more precise subject capture with fewer photographs, resulting in a quicker acquisition process compared to using a lower resolution camera.
To ensure comprehensive coverage, it is recommended to photograph each part of the subject from a minimum of three distinct viewpoints. However, these viewpoints should not be drastically different. Ideally, there should be an overlap of at least two-thirds between consecutive photographs. Additionally, the viewpoints of the same part should not be more than 15 degrees apart. For straightforward subjects, you can accomplish this by capturing approximately 30-50 evenly spaced photographs, encircling the subject from all angles. In the case of aerial photography, it is advisable to maintain both longitudinal and lateral overlap of 80% or more. For optimal outcomes, it is recommended to capture both vertical and oblique photographs. This approach allows for the simultaneous reconstruction of building facades, narrow streets, and courtyards. For a more systematic acquisition, you may choose to prepare a flight plan.
The term 'projected pixel size' refers to the expansion of the traditional ground resolution to encompass a broader range of acquisition configurations, including non-aerial setups. The quality and accuracy of the resulting 3D model depend on the projected pixel size applied to the subject. To achieve a desired projected pixel size, it is necessary to find an appropriate combination of focal length and distance to the subject, as indicated by the following formula: projected pixel size × focal length × largest dimension of the photo = sensor width × distance to the subject[measurement unit/pixel] [millimeters] [pixels] [millimeters] [meters]Having a uniform projected pixel size throughout the entire image is not mandatory since One3D can automatically adjust for variations in projected pixel size, thereby affecting the resolution and accuracy of the resulting 3D model. However, One3D cannot seamlessly merge photographs with significantly different projected pixel sizes. If a wide range of projected pixel sizes is required, it is advisable to use intermediate values to create a smooth transition.
It is advisable to maintain a consistent focal length throughout the entire acquisition process. To achieve varying projected pixel sizes, it is recommended to adjust the distance to the subject. In cases where using multiple focal length settings is unavoidable, such as when the subject distance is restricted, it is advisable to capture several series of photographs, with each series using a fixed focal length. When using a zoom lens, it is important to keep its position fixed across a series of photographs. You can use adhesive tape to secure a manual zoom lens in place. If the camera model type is specified, One3D can automatically estimate extreme lens distortion when using wide-angle or fish-eye lenses. Avoid using digital zoom.
Choose exposure settings that minimize the effects of motion blur, defocus, and noise, as well as over or under-exposure, which can significantly impact the accuracy of 3D reconstruction. For optimal results, manual exposure is recommended as it reduces the chances of color inconsistencies in the texture map of the resulting 3D model. However, using manual exposure requires proficient photography skills and is best suited for stable and uniform lighting conditions. In situations where these conditions cannot be guaranteed, automatic exposure can be employed. It is advisable to disable both optical and digital image stabilization during the capturing process.
Consistent ambient lighting is preferable over direct and/or changing lighting conditions, as the latter can lead to overexposure or underexposure. When capturing images indoors, it is recommended to use fixed lights rather than flash. For outdoor capture, cloudy conditions (with high-altitude cirrus clouds and no rain) are preferable to direct sunlight. If capturing photos on a sunny day is necessary, it is advised to take them around noon to minimize shadow areas. It's important to note that correctly-exposed shadows do not affect the performance of One3D; however, they will be visible in the texture map of the resulting 3D model.
Before importing photographs into One3D, it's important not to alter them by resizing, cropping, rotating, denoising, sharpening, or adjusting parameters such as brightness, contrast, saturation, or hue. Additionally, ensure that the auto-rotate feature on your camera is deactivated.