Image Object Region Selection for Accurate Circular Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image identification techniques struggle with accurately specifying regions for objects with non-quadrangular shapes, particularly circular objects, leading to low identification accuracy.
Innovation Solution
An information processing apparatus that specifies multiple regions with different reference positions and directions, generates identification regions based on these specifications, and identifies objects using learned models to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single quadrangular region is specified for object detection, then the detection process is simple, but the identification accuracy is low for non-quadrangular objects
Solution Approach 1:
The patent divides the object detection process into multiple stages: first specifying a coarse quadrangular region, then generating multiple candidate regions with different reference positions and directions, and finally selecting the optimal identification region. This segmentation allows the system to maintain simplicity in the initial detection phase while achieving high accuracy through subsequent refinement steps.
Solution Approach 2:
The patent introduces additional dimensional parameters beyond the basic quadrangular region by specifying multiple candidate regions with varying reference positions and reference directions. This multi-dimensional approach enables accurate detection of objects with different shapes and orientations, particularly improving performance for circular and non-quadrangular objects.
2Measurement precision
If multiple regions with different reference positions and directions are specified, then object identification accuracy improves, but computational cost increases
Solution Approach 1:
The patent performs preliminary action by first specifying a coarse quadrangular region that narrows down the search space. This preliminary region specification enables subsequent generation of candidate regions to focus only on relevant areas, reducing the overall computational burden while maintaining high identification accuracy.
Solution Approach 2:
The patent applies local quality by generating candidate regions with different reference positions and directions only within the initially specified quadrangular region. This localized approach ensures that computational resources are concentrated on the most relevant areas of the image, improving accuracy without proportionally increasing overall computational cost.
3Measurement precision
If multiple candidate regions are generated with different reference directions, then circular objects can be uniquely identified, but the processing time increases
Solution Approach 1:
The patent employs dynamics by adaptively adjusting the number and configuration of candidate regions based on the detected object's characteristics. For circular objects, multiple candidate regions with different reference directions are generated to determine the optimal angular position. This dynamic approach ensures accurate angular measurement while minimizing unnecessary processing for objects that don't require such detailed analysis.
Data Source
AI summary
In order to enhance the accuracy of identifying an object included in an image, an information processing apparatus (10) includes: a first specifying section (11) that specifies, as a region in which the object in the image exists, a plurality of regions having respective different reference positions or respective different reference directions; a second specifying section (12) that specifies an identification region for the object on the basis of the plurality of regions specified by the first specifying section (11); and an identifying section (13) that identifies the object on the basis of the identification region specified by the second specifying section (12).