Photogrammetric methods and devices related thereto

What is this patent about?

'774 is related to the field of photogrammetry, specifically improving the accuracy and ease of use of 3D digital representation generation. Traditional photogrammetry often requires specialized equipment like structured light emitters or multiple cameras, or cumbersome calibration processes. Existing passive methods using single cameras can suffer from insufficient parallax or require user guidance that is prone to error.

The underlying idea behind '774 is to generate accurate 3D digital representations of objects using only a single, passive image-capture device, such as a smartphone camera. This is achieved by processing a series of overlapping 2D images using a structure from motion (SfM) algorithm to extract 3D geometry. The key insight is that sufficient detail can be obtained from these overlapping images to resolve object parameters and derive accurate measurements without active sensing or complex calibration.

The claims of '774 focus on a computerized method for obtaining measurements of an object. This involves receiving a plurality of 2D images (digital images or video frames) of a scene from a single passive image-capture device. The images must include the object of interest and have overlapping portions. The method then generates a 3D point cloud of the object and extracts spatial distances between point pairs to derive primitive geometry information (edges, lines, surfaces). Finally, this information is converted into 3D coordinates.

In practice, the invention leverages the sequential nature of video to improve the quality of 3D reconstruction. By tracking points within and even outside the image boundaries, the system can better correlate features across frames. The method also parameterizes lines with two endpoints, creating a duality between points and lines that facilitates modeling of lens distortion, even with uncalibrated cameras. This allows the system to filter out noise, reduce data processing needs, and provide object boundary detection.

The invention differentiates itself from prior approaches by using a single passive camera and avoiding cumbersome calibration steps. Unlike methods that rely solely on point clouds, this approach can utilize points, edges, and lines simultaneously, creating point clouds, line clouds, and edge clouds. Furthermore, the system can extract 3D measurements directly from the cloud data, potentially eliminating the need for a separate scaling step. This enables accurate measurements of objects, even at a distance, with accuracies down to a fraction of an inch, depending on the camera and distance.