Object Recognition via Regional Characteristic Quantities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition technologies require significant computational resources and memory to process image data, especially when recognizing the position of objects, and often struggle with noise removal and accurate positioning.
Innovation Solution
An object recognition apparatus that divides images into regions and derives characteristic quantities based on pixel values within these regions, comparing these quantities between images to identify the object's location with reduced computational load, allowing for efficient and accurate position recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image data themselves are used for object recognition, then accurate position recognition is achieved, but large volume of memory is required and amount of calculations increases
Solution Approach 1:
The image is divided into multiple regions, and characteristic quantities are extracted for each region separately. This segmentation allows the system to process smaller data units with reduced memory requirements while maintaining the ability to accurately locate objects through regional comparison.
Solution Approach 2:
Instead of using the entire image data, the invention extracts only the essential characteristic quantities from each region. This extraction process removes unnecessary information while preserving the key features needed for object recognition, thereby reducing memory and calculation requirements.
2Measurement precision
If image data themselves are used for object recognition, then accurate position recognition is achieved, but heavy processing such as noise removal is required
Solution Approach 1:
The invention extracts characteristic quantities that inherently represent the essential features of each region while filtering out noise and irrelevant information. This extraction approach simplifies the processing requirements compared to handling complete image data.
Solution Approach 2:
The invention transforms image data from pixel-level information to region-level characteristic quantities, changing the parameter representation. This transformation reduces the data dimensionality and simplifies subsequent processing operations while maintaining recognition accuracy.
3Quantity of substance
If feature vector is derived from the whole image, then calculation amount is reduced, but position of object cannot be recognized
Solution Approach 1:
The image is segmented into multiple regions, each with its own characteristic quantity. This segmentation preserves spatial information that enables position recognition while keeping the data volume manageable for efficient calculation.
Solution Approach 2:
The invention introduces a regional dimension by dividing the image into multiple areas and assigning characteristic quantities to each region. This dimensional approach maintains position information while reducing the overall computational burden compared to processing the entire image as a single feature vector.
Data Source
AI summary
In an object recognition apparatus, a first division unit and a second division unit each partitions an image into a plurality of regions. A first calculation and a second calculation unit each derives, for each of the plurality of regions partitioned. A first comparison unit and a second comparison unit each compares the derived characteristic quantities in between at least two images, for each of the plurality of regions. A recognition unit recognizes a region where the object is located, based on the comparison result.