Camera Occlusion Assessment for Reliable Robot Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In automation environments, robots face challenges in accurately interacting with objects due to occlusions within the camera field of view, which can lead to errors in object recognition caused by noise and inaccuracies in camera data.
Innovation Solution
A computing system and method that determine occlusion by receiving camera data, identifying a target feature, determining a 2D region co-planar with the target feature, and calculating the size of an occluding region within a 3D region defined by the camera's location and the 2D region's boundary, thereby assessing the confidence in object recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If robot interaction is controlled based on camera data, then automation capability is improved, but object recognition accuracy deteriorates due to occlusions and noise
Solution Approach 1:
The patent introduces an occlusion detection mechanism as an intermediary between camera data acquisition and robot interaction control. This intermediary layer analyzes camera images to detect occluding objects and determines their impact on target object recognition, thereby mediating the trade-off between automation and accuracy by conditionally processing camera data based on occlusion assessment
Solution Approach 2:
The system implements feedback by using camera data to detect occlusions and feed this information back into the robot control decision-making process. The occlusion detection results are used to adjust robot interaction operations, creating a closed-loop control system that continuously refines automation based on real-time visual feedback about occlusion conditions
2Measurement precision
If multiple cameras are used to reduce occlusion impact, then object recognition accuracy is improved, but system complexity increases
Solution Approach 1:
The patent transitions from two-dimensional image analysis to three-dimensional spatial reasoning by determining a 3D region based on camera position and 2D region boundaries. This dimensional transformation enables more accurate occlusion detection without requiring multiple cameras, as the 3D spatial model can infer occlusion relationships from single-camera data combined with object structure information
Data Source
AI summary
A system and method for determining occlusion are presented. The system receives camera data generated by at least one camera, which includes a first camera having a first camera field of view. The camera data is generated when a stack having a plurality of objects is in the first camera field of view, and describes a stack structure formed from at least an object structure for a first object of the plurality of objects. The system identifies a target feature of or disposed on the object structure, and determines a 2D region that is co-planar with and surrounds the target feature. The system determines a 3D region defined by connecting a location of the first camera and the 2D region. The system determines, based on the camera data and the 3D region, a size of an occluding region, and determines a value of an object recognition confidence parameter.