Back to results
US 12,711,776 B2Granted
Systems and methods for constructing high resolution panoramic imagery for feature identification on robotic devices
AssigneeBrain Corporation
InventorsMicah Richert, Guru-Koushik Senthil-Kumar
Life of this patent
Estimating the term…
Abstract
Systems and methods for constructing high resolution panoramic imagery for feature identification on robotic devices are disclosed herein. According to at least one non-limiting exemplary embodiment, a robot collects a plurality of images of an environment, these images include large overlap in their visual scenes. Using additional image data from the overlapping images, resolution of labels, price tags, and other inventory tags may be enhanced when constructing panoramic imagery to improve feature identification.
Claims20
A robotic system, comprising:
- a memory comprising computer readable instructions stored thereon;
- and
- a processor configured to execute the computer readable instructions to:
- receive, via a sensor coupled to the robotic system, a first image of an object and a second image of the object as the robotic system moves along a route, wherein the route comprises one or more scanning segments that extend parallel to surfaces of corresponding objects to be imaged;
- determine, via a computer readable map, distance to the object within the first and second images and horizontal translation of the robotic system between the first and second image, wherein the horizontal translation of the robotic system on the computer readable map is determined based on odometry data of the robotic system, wherein the distance to the object is based on a pre-determined distance of the one or more scanning segments to the surfaces of the corresponding objects to be imaged;
- determine a bounding box for each price label depicted in the first and second images;
- determine an image quality matrix based on a level of contrast detected within the bounding boxes corresponding to the price labels within the first and second images, wherein the contrast is detected based on color transitions between adjacent pixels within the bounding boxes;
- align the first and second images to form a panoramic image at least in part based on locations of the bounding boxes and the horizontal translation of the robotic system, such that price labels of the object are neither duplicated nor omitted in the panoramic image, and the panoramic image begins and ends proximate to edges of the object on the computer readable map;
- and
- communicate the panoramic image to a server.
7 more elements
A method for forming an image by a robotic system, comprising:
- receiving, via a sensor coupled to the robotic system, a first image of an object and a second image of the object as the robotic system moves along a route, wherein the route comprises one or more scanning segments that extend parallel to surfaces of corresponding objects to be imaged;
- determining, via a computer readable map, distance to the object within the first and second images and horizontal translation of the robotic system between the first and second image, wherein the horizontal translation of the robotic system on the computer readable map is determined based on odometry data of the robotic system, wherein the distance to the object is based on a pre-determined distance of the one or more scanning segments to the surfaces of the corresponding objects to be imaged;
- determining a bounding box for each price label depicted in the first and second images;
- determining an image quality matrix based on a level of contrast detected within the bounding boxes corresponding to the price labels within the first and second images, wherein the contrast is detected based on color transitions between adjacent pixels within the bounding boxes;
- aligning the first and second images to form a panoramic image at least in part based on locations of the bounding boxes and the horizontal translation of the robotic system, such that price labels of the object are neither duplicated nor omitted in the panoramic image, and the panoramic image begins and ends proximate to edges of the object on the computer readable map;
- and
- communicating the panoramic image to a server.
4 more elements
A non-transitory computer readable medium comprising computer readable instructions stored there that when executed by at least one processor configure the at least one processor to,
- receive, via a sensor coupled to a robotic system, a first image of an object and a second image of the object as the robotic system moves along a route, wherein the route comprises one or more scanning segments that extend parallel to surfaces of corresponding objects to be imaged;
- determine, via a computer readable map, a distance to the object within the first and second images and horizontal translation of the robotic system between the first and second image, wherein the horizontal translation of the robotic system on the computer readable map is determined based on odometry data of the robotic system, wherein the distance to the object is based on a pre-determined distance of the one or more scanning segments to the surfaces of the corresponding objects to be imaged;
- determine a bounding box for each price label depicted in the first and second images;
- determine an image quality matrix based on a level of contrast detected within the bounding boxes corresponding to the price labels within the first and second images, wherein the contrast is detected based on color transitions between adjacent pixels within the bounding boxes:
- align the first and second images to form a panoramic image at least in part based on locations of the bounding boxes and the horizontal translation of the robotic system, such that price labels of the object are neither duplicated nor omitted in the panoramic image, and the panoramic image begins and ends proximate to edges of the object on the computer readable map;
- and
- communicate the panoramic image to a server.
4 more elements
Dependencies are inferred from the claim text.