Coin Recognition System Using Segmented Processing and Bright Field Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current coin recognition systems face challenges in efficiently processing large numbers of different coin types due to high computational costs, energy consumption, and the need for multiple passes, which increases wear and decreases productivity, while also failing to effectively identify complex coin attributes like year, mint-mark, and condition.
Innovation Solution
A coin recognition system utilizing a Circuital Coin Path with bright field imaging and machine learning algorithms that processes coins in two passes, using a combination of Imager-Sorter and Recognizer-Controller components to identify types and attributes efficiently, reducing computational resources and maintenance needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If digital image recognition is performed on bulk coins using existing systems, then coin types can be recognized, but the processing speed is slow (approximately 10 coins per second) and computational cost is high ($1200 to $5000)
Solution Approach 1:
The patent segments the coin recognition task into multiple specialized processing units, each dedicated to recognizing specific coin types or attributes. This allows parallel processing of different coin subsets, dramatically increasing throughput while maintaining recognition accuracy through specialized optimization for each segment.
Solution Approach 2:
The patent introduces a temporal dimension by implementing multi-pass processing where coins are recognized across multiple cycles. Each pass focuses on different aspects or subsets of coin types, allowing the system to accumulate recognition data over time and achieve both high speed and high accuracy through distributed temporal processing.
2Speed
If template data for all coin types is stored in memory for rapid recognition, then recognition speed improves, but memory requirements become prohibitively large (200 GB or more)
Solution Approach 1:
The patent divides the coin type database into multiple smaller segments or categories. Each processing unit maintains only the template data for its assigned segment in fast memory, while other segments are stored in slower external storage. This segmentation allows the system to achieve rapid recognition for common coin types while using economical storage for less frequent types.
Solution Approach 2:
The patent implements local quality by providing high-speed memory access for frequently encountered coin types while using slower, cheaper storage for rare coin types. The system dynamically adjusts which coin templates reside in fast memory based on observed frequency, ensuring that the most commonly processed coins benefit from rapid access while minimizing overall memory requirements.
3Measurement precision
If multiple passes are used to process bulk coins, then recognition accuracy can be improved, but coin wear increases and productivity decreases
Solution Approach 1:
The patent extracts and separates the imaging function from the mechanical handling function. Coins are imaged once during a single pass through the system, and multiple copies of the images are created and processed computationally. This eliminates the need for physical re-handling of coins for additional imaging passes, thereby preventing wear while enabling multi-analyis of coin attributes through digital image processing.
Solution Approach 2:
The patent creates digital copies of coin images that can be analyzed repeatedly without affecting the physical coin. Once a coin is imaged during a single pass, the digital image is stored and can be processed multiple times to extract different attributes or verify recognition accuracy, replacing the need for multiple physical passes that would cause wear.
4Adaptability or versatility
If existing recognition systems are used, then basic coin types can be identified, but complex attributes like year, mint-mark, and condition cannot be effectively identified
Solution Approach 1:
The patent implements a universal recognition platform that can identify both basic coin types and complex attributes through a single integrated system. Multiple processing units are configured to detect different features (type, year, mint-mark, condition) from the same set of coin images, allowing the system to perform multiple functions simultaneously without requiring separate specialized systems for each attribute.
Data Source
AI summary
A coin handling system performs a one or two-part coin handling and recognition process. In the first part of the two-part process, the system images bulk coins, determines the coin types or other attributes, and returns at least some of the coins to the bulk coin receptacle; in the second part, the system re-images the coins and uses the coin types or attributes determined in the first part to efficiently and economically perform machine recognition of attributes of the coins. The output is used to handle the coins and to determine a price to pay for the coins. The images are bright field images suitable for use by people.