Hand-Drawn Chart Recognition via Connectivity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems fail to effectively recognize and represent the meaning of hand-drawn charts in ink input, as they are sensitive to stroke order and number, and struggle with grouping strokes that form complex diagrams or charts.
Innovation Solution
A system and method that includes a chart recognizer with connectivity-based, connected container, and curve recognition capabilities, allowing for the detection and recognition of hand-drawn charts with connected areas, containers, and curves, regardless of stroke order or number, and generates a drawing from recognized charts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If incremental recognition algorithms are used to recognize simple geometric shapes, then recognition of basic shapes from specific stroke sequences is achieved, but the system fails to handle multi-stroke shapes and diagrams due to sensitivity to stroke order and inability to group strokes
Solution Approach 1:
The patent segments the hand-drawn diagram into multiple strokes and groups them based on spatial proximity and connectivity relationships. Each stroke is analyzed independently initially, then clustered into coherent shapes using distance thresholds and connection detection, allowing the system to handle diagrams drawn with any number of strokes in any order.
Solution Approach 2:
The patent transitions from sequential stroke-based recognition to spatial relationship-based recognition by analyzing the geometric positions, distances, and connectivity between stroke endpoints. This dimensional shift from temporal (stroke order) to spatial (position and connection) analysis enables handling of multi-stroke diagrams regardless of drawing sequence.
2Ease of manufacture
If stroke order and number are fixed requirements for recognition, then simple shape recognition is possible, but complex hand-drawn charts with arbitrary stroke sequences cannot be recognized
Solution Approach 1:
The patent creates a universal recognition system that handles multiple types of charts (pie charts, bar charts, flowcharts, etc.) and multiple stroke configurations through a single unified approach. The system uses general spatial analysis and connectivity detection that works regardless of the specific diagram type or number of strokes, making the recognition process adaptable to arbitrary user input.
Solution Approach 2:
The patent changes the recognition parameters from fixed stroke sequences to flexible spatial relationships. By using distance thresholds, connection detection, and geometric property analysis instead of requiring specific stroke orders, the system maintains simplicity while achieving flexibility to handle any stroke input configuration.
3Reliability
If the system requires specific number of strokes and orders for recognition, then basic shape detection works, but robust recognition of hand-drawn objects with variable stroke patterns fails
Solution Approach 1:
The patent enables the recognition system to automatically adapt to the user's drawing style without requiring explicit guidance. The system self-adjusts by analyzing the actual stroke patterns present in the input, detecting connectivity relationships, and grouping strokes into coherent shapes based on spatial criteria rather than predefined stroke requirements, thus maintaining reliability while preserving user freedom.
Data Source
AI summary
A system and method for recognition of hand-drawn charts in ink input is provided. A chart recognizer may be provided that may recognize a hand-drawn diagram or chart in ink input. The chart recognizer may include a connectivity-based recognizer for recognizing a hand-drawn chart having connected areas such as a pie chart, a connected container recognizer for recognizing a hand-drawn chart having connected containers such as a cycle diagram, and a curve recognizer for recognizing a hand-drawn chart having a curve. The connected container recognizer may also recognize a hand-drawn chart having intersecting containers such as a Venn diagram or a hand-drawn chart having a container that may include another container such as a target diagram.