Region-Based Object Identification With Selective Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems face challenges in reducing computational load while maintaining or enhancing the accuracy of object identification in captured images.
Innovation Solution
An image processing apparatus that includes a stereo camera and a processor, utilizing a combination of deep neural networks and region-based feature extraction to identify objects in both full and region-specific images, allowing for selective use of feature quantities to optimize computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature quantities from the entire captured image are extracted for object identification, then identification accuracy is improved, but computational load increases
Solution Approach 1:
The captured image is divided into multiple regions of interest (ROIs) based on detection results from a first object identifier. Feature quantities are then extracted only from these specific ROIs rather than the entire image, reducing computational load while maintaining identification accuracy for objects within those regions.
Solution Approach 2:
Different processing strategies are applied to different regions of the image. Regions identified as containing objects of interest receive detailed feature extraction and second object identification, while other regions receive minimal or no processing, optimizing the balance between accuracy and computational efficiency.
2Measurement precision
If multiple feature quantities are extracted and processed for object identification, then identification accuracy is improved, but processing time increases
Solution Approach 1:
A first object identifier performs preliminary detection to identify potential regions of interest before detailed feature extraction. This preliminary action filters out irrelevant regions, so that subsequent detailed feature extraction and second object identification are performed only on necessary regions, reducing overall processing time.
Solution Approach 2:
Instead of performing complete feature extraction and identification on the entire image, the system performs partial action by limiting detailed processing to only those regions identified as containing objects of interest, achieving sufficient identification accuracy with reduced processing time.
Data Source
AI summary
An image processing apparatus includes a first extractor, a first object identifier, a region defining unit, a second extractor, a selector, and a second object identifier. The first extractor extracts a first feature quantity included in a captured image. The first object identifier identifies an object on the basis of the first feature quantity. The region defining unit defines an image region in the captured image. The second extractor extracts s a second feature quantity included in an image in the image region. The selector selects, on the basis of data related to the image region defined by the region defining unit, a part of the first feature quantity extracted by the first extractor. The second object identifier identifies the object on the basis of the second feature quantity and the part of the first feature quantity selected by the selector.