Offline 2D Image Filter Selection for DPM Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing code reader systems face inefficient and lengthy setup procedures due to the limited availability of pipelined complex image processing filters, leading to burdensome configurations and slow decoding rates, especially for poorly printed machine-readable indicia.
Innovation Solution
Implement an offline learning process to select a limited number of pipelined complex image processing filters based on decodability metrics, using a large training set to improve decoding rates and reduce setup time, allowing non-technical personnel to perform automatic setups.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pipelined complex image processing filters are used, then decoding rates improve, but setup time increases exponentially
Solution Approach 1:
The patent performs filter selection in advance during an offline training phase using a large dataset of machine-readable indicia. The system pre-evaluates multiple pipelined complex image processing filters and stores the optimal filter set, so that during online operation the automatic setup procedure can quickly retrieve pre-determined filters rather than exploring the full filter space in real-time.
2Adaptability or versatility
If more image processing filters are offered, then decoding capability improves, but device complexity increases
Solution Approach 1:
The patent extracts and stores only the most effective pipelined complex image processing filters that were identified during offline training. Instead of providing access to all possible filters, the system selectively retains a curated subset of filters that have been proven to work optimally for specific types of machine-readable indicia, thereby reducing configuration complexity while maintaining high decoding capability.
3Measurement precision
If automatic setup procedure explores more filter possibilities, then decoding accuracy improves, but convergence speed decreases
Solution Approach 1:
The system pre-performs the exhaustive filter evaluation and selection during an offline training phase using a comprehensive dataset. The results of this preliminary exploration are stored as optimal filter configurations, allowing the online automatic setup procedure to achieve high decoding accuracy without repeating the exhaustive search, thus maintaining fast convergence speed.
Data Source
Figure 1
Figure 2
Figure 3A~4B
AI summary
A code reader may include a non-transitory memory configured to store multiple image processing filters including at least one pipelined complex image processing filter. An image sensor may be configured to image machine-readable indicia. Processor(s) of the code reader may be configured to perform an automatic setup procedure that captures one or more sample machine-readable indicia. The sample machine-readable indicia may be decoded by utilizing the image processing filters including at least one of the pipelined complex image processing filter(s) determined offline from the code reader. A subset of the image processing filters may be selected based on performance of the image processing filters in decoding the multiple sample machine-readable indicia for use during normal code reading operations of the code reader may be performed. Normal code reading operations may be performed in reading machine-readable indicia using the selected subset of the image processing filters.