Contextual Object Matching for Robust Visual Task Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Visual inspection systems using autonomous moving bodies face challenges in accurately targeting inspection objects due to positional errors from GPS and environmental changes, which hinder effective image matching and damage detection.
Innovation Solution
A task execution system that includes a database management unit for recording contextual relationships, an imaging unit for capturing images with position information, an object detection unit, a segmentation unit, and a contextual relationship extraction unit to execute tasks based on spatial relationships between objects, allowing robust processing against changes in visual features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If GPS position information is used to navigate the moving body to the inspection target place, then the moving body can autonomously move to the target location, but the position accuracy deteriorates with errors of 30 cm to 5 m preventing accurate imaging
Solution Approach 1:
The patent introduces visual feature matching as an intermediary mechanism between GPS navigation and target imaging. The system uses template images stored in a database to match visual features of the inspection target, enabling the moving body to locate and image the target accurately even when GPS position information has large errors of 30 cm to 5 m.
2Adaptability or versatility
If visual SLAM technology is used to navigate in GPS-denied environments, then the moving body can operate indoors, but the calculation load increases and position accuracy remains limited to 20 to 30 cm error
Solution Approach 1:
The patent performs preliminary action by storing template images of inspection targets in a database before the actual inspection task. During operation, the system matches visual features against these pre-stored templates, eliminating the need for complex real-time SLAM calculations and achieving accurate positioning even in GPS-denied indoor environments.
3Measurement precision
If template image matching is used for navigation, then the moving body can reach the inspection target place, but the system fails when visual features change due to damage, loss, or environmental changes
Solution Approach 1:
The patent applies dynamics by making the template image adaptable to visual feature changes. When the inspection target or its environment changes, the system updates the template image stored in the database with the new visual features. This dynamic update mechanism maintains accurate matching capability even when the target is damaged, lost, or when seasonal environmental changes occur.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A technique of executing a task related to a target object robustly against a change in a visual feature is provided. A task execution system includes: a database management unit configured to record in advance a contextual relationship database indicating a spatial contextual relationship of a plurality of objects including a target object; an imaging unit configured to acquire image data that is data obtained by adding, to an image, position information indicating a position at which the image is captured; an object detection unit configured to detect objects from the image; a segmentation unit configured to extract the objects from the image; a contextual relationship extraction unit configured to extract a spatial contextual relationship of the objects extracted from the image; and a task execution unit configured to execute a task related to the target object based on the contextual relationship of the objects extracted from the image and the contextual relationship recorded in the contextual relationship database.