AI Image Matching Using Two-Stage Features for Speed and Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing image retrieval systems face a contradiction between efficiency and precision due to simplified machine learning models and reduced matching times, which affect the accuracy of image retrieval.
Innovation Solution
An AI-based image processing method involving two-stage matching: first, using low-dimensional image features to screen a preset number of reference images, followed by high-dimensional feature matching to determine the image processing result, balancing efficiency and precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a simplified machine learning model is used to reduce calculation amount, then image retrieval efficiency is improved, but image retrieval precision deteriorates
Solution Approach 1:
The patent segments the image retrieval process into two distinct stages: a first stage using low-dimensional image features for initial screening and matching, and a second stage using high-dimensional image features for precise matching. This segmentation allows the system to use simpler, faster computations for initial filtering while reserving complex, accurate computations for final verification, thereby resolving the contradiction between efficiency and precision.
Solution Approach 2:
The patent employs image features of different dimensions (low-dimensional and high-dimensional) in sequence. The low-dimensional features provide a coarse-grained representation for efficient initial matching, while the high-dimensional features provide a fine-grained representation for accurate final matching. This dimensional transition enables the system to achieve both speed and accuracy.
2Loss of time
If the number of matching times of the base library is reduced to reduce matching time consumption, then image retrieval efficiency is improved, but image retrieval precision deteriorates
Solution Approach 1:
The patent performs preliminary matching using low-dimensional image features before the final high-dimensional matching. This preliminary action filters out obviously unrelated images early in the process, reducing the number of full matching operations needed while ensuring that only the most promising candidates undergo detailed verification, thus balancing time consumption with precision.
Solution Approach 2:
The matching process is divided into two phases: a first phase using low-dimensional features for rapid initial screening, and a second phase using high-dimensional features for precise final matching. This segmentation reduces the total number of high-cost matching operations while maintaining high precision through the two-stage verification process.
Data Source
AI summary
An artificial intelligence (AI)-based image processing method includes: acquiring a first image feature and a second image feature of a preset image, and first reference image features and second reference image features of reference images in a first reference image set; determining first matching degrees between the preset image and the reference images in the first reference image set based on the first image feature and the first reference image features, and selecting, based on the first matching degrees, a preset number of reference images from the first reference image set to generate a second reference image set; and determining second matching degrees between the preset image and reference images in the second reference image set based on the second image feature and second reference image features, and determining an image processing result of the preset image based on the second matching degrees.