In recent years, using consumer-grade 360° cameras, such as Insta360 and DJI Osmo, for 3D reconstruction has become far more than a niche laboratory experiment.

Across numerous technical practices worldwide, the workflow has become remarkably similar: an operator carries a 360° camera around the scene to capture panoramic imagery, exports the images in Equirectangular format, and then converts them into multiple perspective views. The processed images are fed into COLMAP, where the camera poses are estimated using Structure-from-Motion (SfM). The resulting data is then processed using open-source frameworks such as nerfstudio and gsplat to train a 3D Gaussian Splatting (3DGS) model. In small environments with rich texture features and consistent lighting, the resulting models can deliver excellent visual quality.

There are also more advanced technical approaches. For example, some workflows use raw dual-fisheye frames directly as input, while others rely on specialized panoramic rasterization frameworks such as gsplat360 and PFGS360, eliminating the need for cube-map conversion. These technical approaches continue to evolve, supported by highly active open-source communities.

The conclusion is clear: consumer-grade 360° cameras are capable of producing 3DGS models.

So where is the limitation?

When the above 3D reconstruction workflow is extended from small indoor rooms to larger environments such as entire floors, outdoor streets, and large areas of woodland, the core challenge consistently comes down to one critical stage: camera pose estimation.

Purely vision-based SfM has several unavoidable limitations:

• Texture dependency: In weak-texture environments such as forests, snow-covered terrain, large white walls, and glass curtain walls, feature points can be difficult to match. Camera poses may drift or fail to be calculated altogether.
• Lighting dependency: In underground parking garages, poorly lit indoor areas, or nighttime environments, reduced image signal-to-noise ratios can directly degrade reconstruction quality.
• Drift over long trajectories: The longer the acquisition route and the more loop closures involved, the more significant accumulated errors become. Eventually, the model may fail to align or stitch correctly.
• No true physical scale: SfM reconstructs a scene only up to a similarity transformation. The resulting model does not inherently know how long one meter is. Additional scale references, calibration objects, or control points are therefore required for dimensional measurements.

There is another practical challenge: when data collected over a large area fails, repeating the field survey can be costly. In many workflows, operators cannot immediately determine whether the captured imagery is suitable for model training. They often have to return to a workstation and run the complete COLMAP workflow before discovering that certain sections of the dataset cannot be used.

Consumer-grade 360° cameras already offer excellent imaging performance, and mainstream panoramic cameras can outperform many self-developed surveying cameras in terms of cost and image quality.

The weakness of the overall solution is not the imaging hardware itself, but the lack of a stable pose-estimation system capable of supporting real-world spatial measurement.

To address the fundamental limitations of purely visual SfM reconstruction—including failure in weak-texture environments, degradation under low-light conditions, drift over long trajectories, and the lack of a true measurement scale—TourPop provides a new solution.

As a handheld LiDAR SLAM scanning device, TourPop is equipped with a LiDAR sensor featuring a point rate of 48,000 points per second, together with an integrated power handle, control unit, and data storage system. The device does not include a built-in panoramic imaging module. Instead, it features a standardized universal mounting interface that supports a wide range of third-party 360° cameras, allowing users to continue leveraging their existing imaging equipment.

The two sensing systems have clearly defined roles and work together:

By using LiDAR point clouds to calculate the device's trajectory, TourPop does not rely on image-feature matching for pose estimation. It can maintain stable positioning in weak-texture environments such as forests and snow-covered areas, as well as in low-light environments such as parking garages and nighttime scenes. Continuous long-distance acquisition can be performed without significant cumulative drift.

LiDAR observations also provide the model with a true physical scale, ensuring that spatial dimensions correspond directly to the real-world environment. Users can perform on-site measurements of distances, widths, heights, and other spatial dimensions directly within the captured scene.

The rendering quality of 3DGS is highly dependent on the quality of the original imagery. When paired with a mature commercial 360° camera, TourPop offers significant advantages over all-in-one spatial cameras in terms of acquisition cost, imaging performance, and hardware flexibility.

TourPop's software integrates a hybrid 3DGS training architecture. Rather than relying solely on SfM-based pose estimation, it uses LiDAR point clouds as the primary spatial reference, with image-based SfM serving as an auxiliary mechanism for image selection and filtering.

Its key advantages include:

• More uniform and structured Gaussian point distribution, significantly reducing the scattered floating-point noise commonly found in purely vision-based solutions.
• Stable 3D reconstruction in complex environments with weak textures or low illumination, reducing the impact of environmental conditions on the modeling process.

The concept of the "spatial camera" has emerged within the industry, representing a clearly defined technological value proposition. Its core concept is that the device does not simply capture 360° panoramic imagery. Instead, it records complete spatial data together with accurate spatial poses and a measurable real-world scale.

TourPop follows the same technological direction while adopting a differentiated implementation:

An open 360° camera architecture combined with LiDAR SLAM as the core pose-estimation technology.

• Users can mount their existing 360° cameras directly onto TourPop and use them as the visual imaging unit.
• LiDAR builds a complete spatial framework while accurately determining the device's position, orientation, movement distance, and the real-world dimensions of the environment.
• After acquisition, the output is not simply a collection of panoramic images, but a sequence of panoramas with accurate pose information, accompanied by measurable LiDAR point clouds and 3D Gaussian Splatting models.

In simple terms, TourPop upgrades a consumer-grade camera that can only capture panoramic imagery into a professional spatial data acquisition terminal capable of comprehensively recording three-dimensional spatial information.

The open, modular camera architecture is designed around two practical considerations.

First, within the surveying and mapping hardware sector, it is difficult to independently develop a panoramic imaging sensor that combines both low cost and excellent image quality.

Second, 360° camera products have relatively short product cycles and evolve rapidly. A modular compatibility architecture allows the TourPop main unit to support a broad range of camera models. When users upgrade their cameras or new generations of 360° cameras enter the market, the TourPop system can continue to be used without replacing the entire acquisition system.

TourPop can generate a range of spatial data products:

• Colorized Point Clouds: With a true physical scale, enabling measurements of distance, height, area, and other spatial dimensions.
• 3DGS Models: Delivering photo-realistic visual quality, with support for free-viewpoint exploration and trajectory-based navigation.
• Pose-Tagged Panoramic Imagery: Panoramas can be viewed and navigated spatially, with each image containing its precise position within the environment.
• Mesh Models: Generated from imagery and/or point clouds using third-party software for visualization, editing, and rendering.
• BIM Models: Further developed using third-party software for engineering-oriented information delivery.

Note: Mesh and BIM models require additional processing using professional third-party software.

1. Users who already own 360° imaging equipment and regularly perform 3DGS reconstruction or large-scale spatial digitalization.
2. Applications involving indoor and outdoor spatial documentation, cultural and tourism digital presentation, and rapid 3D reconstruction of real-world environments.
3. Scenarios that require an overall representation of spatial dimensions but do not have strict millimeter-level surveying accuracy requirements, including rapid 3D reconstruction of traffic accident scenes, forest and landscape resource surveys, and similar applications.

Precision surveying and mapping projects that require millimeter-level measurement accuracy.

TourPop is designed as a centimeter-level, rapid real-world spatial acquisition tool. It is not intended to replace high-precision surveying instruments such as total stations or professional-grade 3D laser scanners.