Orientation-Based Panoramic Feeds from Stationary Robot Cameras
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robots using rotatable camera heads face issues with wear and tear, cable stretching, restricted field of view, blind spots, and lack of real-time synchronization, which hinder effective spatial awareness and manipulation.
Innovation Solution
A system and method utilizing at least three image sensors on a robot, with a processing subsystem that includes an image acquisition module, feed rectification, estimation, panorama generation, and viewpoint rendering modules to provide a panoramic feed based on user device orientation, eliminating mechanical movements and enabling real-time synchronization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a rotatable camera head is used to acquire visual feed, then the robot can change its field of view, but the rotatable head is subjected to wear and tear and cables are prone to stretching and twisting
Solution Approach 1:
The patent divides the single rotatable camera into multiple stationary cameras positioned at different orientations. Instead of one camera rotating to capture different fields of view, multiple cameras are fixed at predefined angles (e.g., 0°, 90°, 180°, 270°) to simultaneously capture different directions, eliminating mechanical rotation and its associated wear.
Solution Approach 2:
The patent creates virtual copies of the camera system through software processing. Multiple images from stationary cameras are stitched together to create a panoramic view, and virtual camera positions are rendered based on user device orientation. This virtual replication eliminates the need for physical camera movement while maintaining the adaptability of changing fields of view.
2Adaptability or versatility
If a rotatable camera head is used, then the robot can acquire visual feed from different directions, but the limited freedom of movement leads to restricted field of view and blind spots
Solution Approach 1:
The patent positions cameras at multiple discrete orientations (0°, 90°, 180°, 270°) around the robot to capture scenes from different directions simultaneously. This segmentation of the viewing task among multiple stationary cameras provides comprehensive 360° coverage without the blind spots caused by a single rotatable camera's limited movement range.
Solution Approach 2:
The patent merges images from multiple stationary cameras into a single panoramic view through image stitching algorithms. By combining the fields of view of multiple cameras positioned at different orientations, the system achieves a complete 360° spherical view, eliminating blind spots and providing operators with full spatial awareness of the environment.
3Adaptability or versatility
If a rotatable camera head is used, then the robot can change its viewing angle, but lack of real-time synchronization hinders spatial awareness
Solution Approach 1:
The patent pre-positions multiple cameras at different orientations to capture images simultaneously before any user interaction occurs. The system continuously captures images from all cameras in real-time, ensuring that when a user views the panoramic feed, the images are already synchronized with the current robot state, eliminating synchronization delays.
Solution Approach 2:
The patent implements real-time feedback by continuously updating the panoramic view based on live images from all cameras and the current orientation of the user device. The system processes images from multiple cameras simultaneously and dynamically adjusts the displayed panoramic view to match the user's viewing angle, ensuring real-time synchronization and eliminating delays in spatial awareness.
4Loss of information
If a rotatable camera head is used, then the robot can acquire visual feed, but the robot lacks the ability to determine the distance between the robot and a target object
Solution Approach 1:
The patent introduces depth information as an intermediary layer between the visual feed and the manipulation system. By processing multiple images from cameras at different positions and orientations, the system extracts depth data that serves as a mediator, enabling the robot to determine distances to target objects and facilitating accurate manipulation without requiring complex additional sensors.
Data Source
AI summary
A system to provide a panoramic feed corresponding to an orientation of a user device is disclosed. The system includes a processing subsystem including an image acquisition module to receive a corresponding visual feed from image sensors. The processing subsystem includes a feed rectification module is configured to rectify one or more anomalies in the visual feed to obtain a rectified feed. The processing subsystem includes an estimation module to estimate a homography matrix for each pair of the rectified feed. The processing subsystem includes a panorama generation module to blend each pair of the corresponding rectified feed based on the homography matrix to obtain the panoramic feed. The processing subsystem includes a viewpoint rendering to evaluate at least one viewpoint based on orientation parameters received from the user device. The viewpoint rendering module is to render a predefined field of view of the panoramic feed in the user device.