THREE-DIMENSIONAL SCANNING SYSTEM

Abstract

A three-dimensional scanning system, characterized in that it comprises a base (1) supporting a series of detachable posts (2) forming a frame, wherein the posts (2) comprise cameras (3), and a processing base (5) using Gaussian splatting.

Description

Three-dimensional scanning system TECHNICAL SECTOR The present invention relates to a three-dimensional scanning system, of the type used to generate a three-dimensional, 3D, model of an object, including living beings, by means of a series of cameras. This model can then be used as an avatar in games, for marketing, to be scanned with real clothes, or for other uses. STATE OF THE ART When creating a three-dimensional model of an object from images, several methods are known. The most traditional involves generating a mesh with a large number of vertices and polygons from those images of the object. This method is limited in quality because it cannot scan semi-transparent, highly reflective, or very thin materials. Furthermore, polygon-based techniques lose all angle-dependent effects such as glare, etc. On the other hand, since late 2023, the Gaussian Splatting technique has been known, which makes it possible to generate a three-dimensional model without requiring the definition of polygons. This results in a much higher quality material and allows working with any material and taking into account angle effects, achieving a much more realistic result. This technique generates projected Gaussian primitives, whose position, orientation, and opacity characteristics are calculated to generate a three-dimensional model that is not based on a mesh. BRIEF EXPLANATION OF THE INVENTION The invention consists of a system for scanning an object using Gaussian Splatting, which can thus be technically and economically viable for use outside of large studios. This system thus offers an advantage when performing marketing functions, augmented reality, etc. The three-dimensional scanning system comprises a base that supports a series of detachable posts forming a perimeter around the object or person to be scanned. The posts include cameras, preferably with their own processing units, and a processing unit using Gaussian splatting. For better modeling, it is preferred that the cameras be approximately the same distance from the object, for which the posts are curved, with the center of curvature over the center of the base. The system may include lights, including the possibility of structured lights (with light and shadow patterns), so that the lighting conditions are controlled. DESCRIPTION OF THE DRAWINGS A section of drawings is presented to facilitate understanding of the invention. These drawings show examples of embodiments of the invention, which are intended only to illustrate it. Figure 1 shows a diagram of the system. Figure 2 shows a base with a post, in exploded view. Figure 3 shows a detail of an exploded pole, showing the back of the face facing the object to be scanned. MODES OF REALIZING THE INVENTION Figure 1 shows a schematic of the preferred embodiment. It comprises a base (1) on which a series of posts (2) are arranged. The base (1) may be made of several detachable parts. Likewise, the posts (2) may be detachable from the base (1), so that the assembly can be easily stored and transported, resulting in a highly portable device. The base (1) may be a circular, polygonal, or any other shape, supported on the ground. The interior of the base (1) may be solid, so that objects rest on the ground at the height desired by the user. The ground may be a special color, which the system can learn to discard, similar to a chroma key. The preferred posts (2) are curved, with the center of curvature over the center of the base (1), so that any object (including people or animals) placed over the center of the base (1) is equidistant from the posts (2), and from most points of these. Cameras (3) are mounted on the posts (2) and pointed towards the center of the base (1). These cameras (3) can have their own processing unit (4), a minicomputer, so that the captured images are preprocessed before being sent to a local or remote processing base (5). This processing base (5) is configured to perform three-dimensional modeling from the images of the cameras (3) using the Gaussian Splatting technique. The number of cameras (3) can be high, for example, more than 100, since this increases the quality of the result with little increase in processing requirements. The number of posts (2) and cameras (3) can be designed for each application, so that the definition of the generated model is configurable. To manage this seamless communication, several network protocols are used: two custom protocols and one generic protocol that enable bidirectional communication between the processing units (4) and the processing base (5). One of these protocols manages the connection status, updating it periodically to ensure all modules are operational. Another critical system launches capture actions to all minicomputers simultaneously. Finally, a third management protocol controls device shutdown and handles file transfers across the network using the SSH (Secure Shell) standard, ensuring an efficient and automated workflow. The images need to be coordinated so they are taken in parallel, for example via a wireless or wired connection. To confirm that the shots were simultaneous, timestamps can be added to the images. The preferred example of a method that ensures simultaneous firing works as follows: The system sends an initial control signal to all cameras. Upon receiving it, each camera records a timestamp and takes an initial control image. After a precise, brief, and pre-programmed time interval (for example, 1 second), the final image is captured. This step is essential because, while the initial synchronization is good, each system may have slight variations in the time it takes to prepare for capture. Since camera preparation only needs to be performed once, for the first image, the second image is captured in perfect sync. This two-step process, by compensating for internal variations in each camera, ensures that the final images are truly captured simultaneously. Furthermore, the system can add timestamps to the images to confirm that the shots were indeed simultaneous. The system may have lights (6), preferably structured lights, that is, lights that project a known pattern of points, lines or geometric figures, so that they help to recognize the curvature of the object by following the lines of light and shadow. In the invention, the preferred calibration method comprises using AI to recognize the presence of human silhouettes, or any object, for automatic extraction, facilitating the extraction of relevant features without visual interference from repetitive or overly homogeneous areas. Since the position of the posts (2) is known, it will suffice to calibrate two of them and apply the Umeyama algorithm to accurately estimate the overall orientation and scale, automatically generating the extrinsic parameters for all cameras. Ideally, methods will be available to compress and package the captured information to reduce its size. The generated three-dimensional models can be sent to an application, a server, or any other type of user interface, in the most suitable format for manipulation with the corresponding program. Ideally, the delivery to an individual should include a viewer to appreciate the result. The invention introduces complete automation of the entire data compression and packaging process in the field of 3D visualization, specifically optimized for Gaussian Splatting. The system introduces complete automation of the entire data compression and packaging process in the field of 3D visualization, specifically optimized for Gaussian splatting. It is a comprehensive system that manages and optimizes the workflow from image capture to the visualization of the three-dimensional model. This system, without manual intervention and within minutes of capture, performs the following functions: - AI Channel Generation: Using artificial intelligence, the system generates extra channels that help increase the training speed of the model and reduce the weight of the final result by ignoring parts of the images and deleting all splats that have an excessive contribution outside the mask area. - Error Detection and Elimination: The system automatically detects and eliminates errors such as floaters (poorly positioned floating points that ruin the reconstruction). - Compression and packaging: The model is compressed into various formats (online vs. offline) and packaged along with additional files. From this package, the system can also generate a website and a thumbnail, offering a complete solution for viewing and using the model. This "all-in-one" approach (capture, data cleaning, and compression) is key for production environments where speed and workflow predictability are the goal. From capture to display, the system: - Generates extra channels through AI that help with training speed and the lightness of the final result. - Detects and automatically eliminates errors, such as the well-known floaters (poorly positioned floating points that ruin the reconstruction). - Performs model compression in various formats (online vs offline). All of this happens without manual intervention and within minutes of capture. This is especially relevant for production environments where speed and workflow predictability are key.

Claims

1. A three-dimensional scanning system, characterized in that it comprises a base (1) supporting a series of detachable posts (2) forming a frame, wherein the posts (2) comprise cameras (3), and a processing base (5) using Gaussian splatting.

2. A three-dimensional scanning system according to claim 1, characterized in that the posts (2) are curved, with the center of curvature at the center of the base (1).

3. A three-dimensional scanning system according to claim 1, characterized in that the base (1) is polygonal.

4. A three-dimensional scanning system according to claim 1, characterized in that the cameras (3) each have processing units (4).

5. A three-dimensional scanning system according to claim 1, characterized in that it comprises lights (6).

6. A three-dimensional scanning system according to claim 5, characterized in that the lights (6) are structured.