Autonomous Dictionary Update for Object Angle Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image-based object recognition techniques are limited in extracting additional information from images and do not autonomously update data for estimation, especially when the object's angle changes, which is crucial for diverse applications.
Innovation Solution
A system comprising a terminal, a database, and a robot that captures images of an object using a camera platform device, generates dictionary data including angle information, and updates it based on recognition results to estimate and adjust the object's angle, enabling autonomous data updating and improved recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dictionary data is prepared per angle of object to enhance recognition availability, then recognition accuracy improves, but data quantity and storage requirements increase
Solution Approach 1:
The system performs preliminary action by automatically generating dictionary data for multiple angles through rotation before actual recognition tasks. The offline generation process creates pre-computed feature amounts for various angles, so that during online recognition, the system can directly match against these pre-prepared data without real-time computation overhead.
Solution Approach 2:
The system creates copies of the object's feature amounts at different angles by rotating the object model. Instead of storing only one original object representation, multiple angular variations are generated and stored in the dictionary, enabling the system to handle objects at any orientation by matching against the appropriate angular copy.
2Reliability
If manual preparation of dictionary data per angle is performed, then recognition robustness improves, but labor time and operational complexity increase
Solution Approach 1:
The system performs self-service by automatically generating its own dictionary data without requiring manual intervention. The automatic generation unit rotates the object model, captures images at different angles, extracts feature amounts, and stores them in the dictionary all autonomously. This eliminates the need for operators to manually prepare data for each angle, significantly reducing operational complexity while maintaining recognition robustness.
Solution Approach 2:
The system replaces manual mechanical operations with automated computational processes. Instead of physically rotating objects and manually capturing images for dictionary creation, the system uses computational geometry to rotate object models and automatically processes images through computer vision algorithms, substituting manual labor with automated mechanical and computational systems.
3Productivity
If offline dictionary generation is performed to reduce online processing time, then recognition speed improves, but initial data preparation time increases
Solution Approach 1:
The system performs preliminary action by generating the dictionary offline before actual recognition operations begin. The offline generation unit creates the complete angular dictionary in advance, so that during online recognition, only simple matching operations are needed. This shifts the time investment from the recognition phase to the preparation phase, dramatically improving real-time recognition speed.
Solution Approach 2:
The system uses periodic action by generating dictionary data at fixed angular intervals (e.g., every 5 degrees or 10 degrees). Instead of continuously capturing images at all possible angles, the system samples at periodic intervals, creating a discrete angular dictionary that balances completeness with efficiency. This periodic sampling approach reduces the total number of images needed while maintaining adequate recognition coverage.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
According to a certain aspect of the present invention, there is provided an information processing device including a processor that realizes: a dictionary data acquisition function that acquires dictionary data related to an object; an image acquisition function that acquires a first image of the object; a first angle estimation function that estimates an angle of the object in the first image on the basis of the first image and the dictionary data; a second angle estimation function that re-estimates an angle of the object in the first image after a physical operation related to the object; and a dictionary data update function that updates the dictionary data in response to a result of estimation by the first angle estimation function and a result of re-estimation by the second angle estimation function.