Real-Time Merchandise Recognition via Dimensional Feature Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques fail to enable real-time merchandise management by recognizing merchandise in video images, as they do not efficiently process and identify items in a captured video.
Innovation Solution
A system comprising a first local feature storing unit, an imaging unit, and a recognizing unit that generates and compares 1-dimensional to j-dimensional feature vectors for local areas in images, allowing for real-time recognition of merchandise by determining the correspondence between stored and captured features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional image recognition techniques are used, then recognition accuracy can be maintained, but real-time processing capability is lost
Solution Approach 1:
The patent divides the image recognition process into multiple stages: first extracting local features from the input image, then comparing these features against a pre-built model dictionary, and finally making recognition decisions. This segmentation allows computationally intensive operations to be distributed and optimized, enabling real-time processing while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary actions by pre-building a model dictionary from training images before actual recognition occurs. This dictionary contains pre-extracted features and classification information, allowing the system to skip time-consuming training phases during real-time operation and only perform fast feature matching and classification.
2Measurement precision
If detailed feature analysis is performed, then recognition accuracy improves, but processing speed decreases
Solution Approach 1:
The patent applies partial action by selectively processing only the most discriminative local features rather than analyzing every pixel or feature in the image. The system identifies and processes key feature points that provide sufficient recognition information, avoiding unnecessary computational overhead while maintaining high accuracy.
Solution Approach 2:
The patent implements local quality by focusing computational resources on extracting and analyzing local features from specific regions of interest in the image, rather than uniformly processing the entire image. This allows detailed analysis where needed while skipping less important areas, balancing accuracy and speed.
Data Source
AI summary
Merchandise management is implemented by recognizing a piece of merchandise in an image on a video in real time. A piece of merchandise and m-number of first local feature which are respectively 1-dimensional to i-dimensional feature vectors are stored after being associated with each other, n-number of feature points are extracted from an image on a video captured by an imaging unit, n-number of second local feature which are respectively 1-dimensional to j-dimensional feature vectors are generated, a smaller number of dimensions of the number of dimensions i and the number of dimensions j is selected, and a recognition that the merchandise exists in the image on the video is made when it is determined that a prescribed proportion or more of the m-number of first local feature of the selected number of dimensions correspond to the n-number of second local feature of the selected number of dimensions.