Object Identification Using Optical Code Filtering and Visual Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition systems face inefficiencies in large databases, where the search for nearest-neighbors becomes increasingly difficult and time-consuming as the number of recognized objects grows, often requiring manual intervention for no-code exceptions and lacking speed and accuracy in decoding obscured optical codes.
Innovation Solution
A method that reduces the search space by using a database filter unit to generate a subset of feature models based on decoded portions of optical codes, allowing for efficient comparison with extracted visual features to identify matches, thereby improving recognition speed and accuracy and handling no-code exceptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large database of feature models is used to improve object recognition accuracy, then recognition accuracy is improved, but search time and processing speed deteriorate
Solution Approach 1:
The patent segments the large database search process into two phases: first, optical code decoding provides a preliminary filter to identify candidate objects; second, visual feature recognition is performed only on this reduced subset. This segmentation transforms a single large search into a two-stage process with a small initial search followed by targeted detailed recognition, resolving the contradiction between database size and search speed.
Solution Approach 2:
The patent performs preliminary action by decoding the optical code before conducting visual feature recognition. The optical code contains identification information that can be decoded in advance to generate a filter parameter, which pre-identifies candidate objects from the database. This preliminary filtering action reduces the subsequent visual search space, allowing the system to maintain high accuracy while reducing search time.
2Measurement precision
If complete optical code decoding is required to ensure accurate object identification, then identification accuracy is improved, but system reliability deteriorates when codes are obscured or damaged
Solution Approach 1:
The patent applies partial action by using only the portion of the optical code that can be successfully decoded, rather than requiring complete code decoding. The system extracts available information from partially obscured codes to generate filter parameters, performs visual feature recognition on the resulting candidate subset, and identifies the object without needing the entire optical code intact. This allows the system to maintain reliability even when codes are damaged or obscured.
3Measurement precision
If manual intervention is implemented to handle no-code exceptions, then identification accuracy is improved for edge cases, but productivity and automation level deteriorate
Solution Approach 1:
The patent implements self-service by enabling the system to automatically handle cases that would traditionally require manual intervention. When optical code decoding fails or returns incomplete results, the system automatically performs visual feature recognition on candidate objects identified through alternative means, and can even prompt users to capture additional images if needed. This automated fallback mechanism eliminates the need for manual keyboard entry or operator intervention, maintaining both accuracy and productivity.
Data Source
AI summary
An object identification system comprises an optical code reader that scans an optical code of an object and decodes a portion of the optical code. Using the decoded portion of the optical code, a database filter unit generates a filtered subset of feature models from a set of feature models of known objects stored in a database. An image capture device captures an image of the object, and a feature detector unit detects visual features in the image. A comparison unit compares the detected visual features to the filtered subset of feature models to identify a match between the object and a known object.