Methods for automatic registration of 3D image data

1. A system comprising:

a red-green-blue (RGB) camera for capturing two-dimensional (2D) image data associated with a physical environment;

a depth camera for capturing depth data associated with the physical environment;

an inertial measurement unit (IMU) for capturing motion data associated with the system;

one or more processors;

non-transitory computer-readable media storing computer executable instructions which when executed by the one or more processors cause the one or more processors to perform operations including:

receiving first 2D image data captured by the red-green-blue (RGB) camera at a first time;

receiving first depth data captured by the depth camera at the first time;

determining a first pose associated with the system at the first time based at least in part on the motion data at the first time;

performing registration of the RGB camera to the depth camera based at least in part on the first pose, the first 2D data, and the first depth data;

receiving second 2D image data captured by the RGB camera at a second time;

receiving second depth data captured by the depth camera at the second time;

determining a second pose associated with the system at the second time based at least in part on the motion data at the first time;

identifying a set of points common to the first depth data of the first pose and the second depth data of the second pose;

determining a color error function from the set of points; and

updating the registration of the RGB camera to the depth camera to reduce misalignment between the RGB camera and the depth camera based at least in part on the color error function.

2. The system of claim 1 , wherein the second depth data and the second 2D image data is captured in response to a period of time elapsing.

3. The system of claim 1 , wherein the second depth data and the second 2D image data is captured in response to the motion data indicating a change in position of the system.

4. The system of claim 1 , wherein:

the first 2D image data includes foreground of a first color against a background of a second color;

the second 2D image data includes foreground of the first color against a background of the second color;

the first depth data includes depth data associated with the foreground and the background of the first 2D image data;

the second depth data includes depth data associated with the foreground and the background of the second 2D image data; and

performing the registration includes:

generating first combined image data by combining the first 2D image data with the first depth data;

detecting a profile of an object based at least in part on the combined image data;

generating second combined image data by combining the second 2D image data with the second depth data;

overlaying the second combined image data over the first combined image data;

detecting differences between the second combined image data and the first combined image data;

identify pixels of the first depth data corresponding to the differences;

identify pixels of the second depth data corresponding to the pixels of the first depth data; and

correcting the registration based at least in part on the pixels of the first depth data and the corresponding pixels of the second depth data.

5. The system of claim 1 , further comprising:

wherein the non-transitory computer-readable media stores additional computer-executable instructions, which when executed by the one or more processors cause the one or more processors to perform operations including:

receiving user input corresponding to one or more registration parameters; and correcting the registration of the RGB camera to the depth camera based on at least in part on the user input.