Object Detection Using Reduced Reference Feature Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection methods are inefficient due to the need to perform time-consuming feature extraction on multiple reference images, and redundant operations occur when instances of the target object have similar appearances, reducing overall efficiency.
Innovation Solution
A method involving feature extraction on a plurality of reference images to select a reduced set of identification features, forming a first feature dataset, which is stored for similarity determination with a to-be-detected image, allowing efficient detection of target objects by reducing the number of images processed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature extraction is performed on all reference images, then detection accuracy is improved, but processing time increases
Solution Approach 1:
The patent extracts and stores only the identification features from reference images rather than storing the complete images. This allows the system to maintain detection accuracy by preserving essential feature information while significantly reducing the data volume that needs to be processed during detection, thus resolving the contradiction between accuracy and processing time.
Solution Approach 2:
The patent segments the reference images into their essential identification features, separating the critical information needed for detection from the redundant visual data. This segmentation enables the system to work with a reduced feature dataset that maintains detection accuracy while reducing processing requirements.
2Loss of information
If all reference images are processed, then feature completeness is improved, but redundant operations increase
Solution Approach 1:
The patent extracts only the identification features from reference images and discards redundant visual information. This extraction process maintains feature completeness by preserving all necessary identification data while eliminating redundant operations that would arise from processing complete images, thereby improving processing efficiency.
Solution Approach 2:
The patent changes the data representation parameter from complete images to extracted identification features. This parameter transformation reduces the amount of data that needs to be processed while maintaining the essential information needed for accurate detection, thus resolving the contradiction between feature completeness and processing efficiency.
3Reliability
If multiple reference images with similar instances are processed, then detection robustness is improved, but computational resources are wasted
Solution Approach 1:
The patent extracts identification features from reference images and stores only these features. When processing multiple reference images with similar instances, the system can efficiently compare extracted features rather than processing complete images, maintaining detection robustness while significantly reducing computational resource consumption.
Solution Approach 2:
The patent applies partial action by processing only the necessary identification features from reference images rather than all possible visual data. This partial processing approach maintains detection robustness by preserving essential feature information while avoiding the waste of computational resources on redundant data from similar instances.
Data Source
AI summary
A method for object detection includes: extracting a plurality of identification features from a plurality of reference images that are related to a target object; selecting a plurality of selected identification features respectively from the identification features so as to obtain a first feature dataset, and storing the first feature dataset, a quantity of selected identification features being smaller than a quantity of the identification features; in response to receipt of a to-be-detected image, performing a feature extraction operation on the to-be-detected image to obtain a second feature dataset; performing a similarity determination operation with respect to the to-be-detected image based on the first feature dataset and the second feature dataset, and calculating a quantity of instances of the target object in the to-be-detected image.