Image Recognition Device Using Attribute-Based Area Specification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image recognition devices face challenges in achieving high recognition accuracy, as the accuracy depends on the object being recognized, leading to inefficiencies and potential misrecognition.
Innovation Solution
An image recognition device comprising a first recognition unit, an obtaining unit, an object specifying unit, an area specifying unit, and a second recognition unit, which identifies attributes of objects, references object correspondence information, and specifies areas for targeted recognition to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image recognition is performed without attribute-based filtering, then processing speed is maintained, but recognition accuracy deteriorates due to misrecognition of similar objects
Solution Approach 1:
The image recognition process is segmented into multiple stages: first recognition unit identifies candidate objects, obtaining unit extracts attributes, object specifying unit filters candidates based on attributes, area specifying unit defines search regions, and second recognition unit performs final recognition. This segmentation allows the system to achieve high accuracy by processing objects in hierarchical stages rather than attempting single-pass recognition of all objects.
Solution Approach 2:
The system performs preliminary actions by first identifying candidate objects and extracting their attributes before final recognition. The object specifying unit uses attribute information to pre-filter candidate objects, and the area specifying unit pre-defines search regions based on spatial relationships. This preliminary processing reduces the search space and improves accuracy of the final recognition stage.
2Measurement precision
If attribute-based object specification is performed, then recognition accuracy is improved, but processing time increases due to additional recognition steps
Solution Approach 1:
The system applies partial action by performing complete attribute-based filtering only for candidate objects that require disambiguation. The object specifying unit selectively applies attribute filtering based on recognition needs, and the area specifying unit defines search regions only where necessary. This approach achieves high accuracy for critical recognitions while avoiding unnecessary processing time for straightforward cases.
3Reliability
If comprehensive object recognition is performed across the entire image, then all objects are detected, but processing efficiency deteriorates due to large search space
Solution Approach 1:
The image is segmented into multiple regions of interest based on spatial relationships. The area specifying unit divides the image into first, second, and third areas based on positions of reference objects and candidate objects. The second recognition unit then performs focused recognition only within the relevant second area, rather than scanning the entire image, thus maintaining detection completeness while improving processing efficiency.
Solution Approach 2:
The system performs preliminary spatial analysis to identify relevant search regions before conducting detailed object recognition. The area specifying unit pre-defines search regions based on positions and attributes of reference and candidate objects, allowing the second recognition unit to concentrate computational resources on areas where target objects are likely to be found, thereby improving processing efficiency without sacrificing detection completeness.
Data Source
AI summary
An image recognition device including: a first recognition unit that performs image recognition within an image to find a first object; an obtaining unit that obtains an attribute of the first object found by the first recognition unit; an object specifying unit that refers to object correspondence information showing identifiers of second objects and associating each identifier with an attribute, and specifies an identifier of one of the second objects that is associated with the attribute of the first object; an area specifying unit that refers to area value information showing values that are associated with the identifiers of the second objects and are related to a first area occupied by the first object, and specifies a second area within the image by using a value associated with the identifier of the one of the second objects; and a second recognition unit that performs image recognition within the second area to find the one of the second objects.