3D Data Flattening for Deformed Shipping Label OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Optical character recognition (OCR) systems struggle to accurately recognize address information on deformed shipping labels, particularly when the deformation is significant, leading to false recognition or failure in identifying the label.
Innovation Solution
An image processing apparatus that generates three-dimensional data from images of packages using a stereo camera or light-field camera, followed by two-dimensional data flattening and recognition image generation, to correct deformations and enhance recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional OCR image processing is used to correct deformation, then processing simplicity is maintained, but recognition accuracy deteriorates when deformation is significant
Solution Approach 1:
The patent transitions from two-dimensional image processing to three-dimensional data processing. By generating 3D data from multiple images and performing flattening processing in three-dimensional space, the system can accurately represent and correct severely deformed labels that cannot be properly corrected using conventional 2D image processing techniques alone.
2Measurement precision
If multiple images are acquired from different positions to improve recognition accuracy, then recognition rate improves, but processing time and system complexity increase
Solution Approach 1:
The patent performs preliminary three-dimensional data generation and flattening processing before OCR recognition. By pre-processing the images to create accurate 3D models and flatten them appropriately, the system optimizes the input for subsequent recognition processes, improving overall efficiency and accuracy while managing processing time effectively.
Solution Approach 2:
The patent combines multiple images acquired from different positions into a unified three-dimensional data structure. By merging the information from multiple viewpoints into a single 3D model, the system achieves comprehensive label representation that improves recognition accuracy without requiring separate processing of each image individually.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus has a three-dimensional data generation unit, a two-dimensional data generation unit, and a recognition image generation unit. The three-dimensional data generation unit generates three-dimensional data of an object on the basis of an image of the object on which object information represented by a character or a symbol is attached, and a distance image indicating a distance between a reference position and the object for each pixel. The two-dimensional data generation unit flattens the three-dimensional data generated by the three-dimensional data generation unit to generate two-dimensional data. The recognition image generation unit generates a recognition image which is used to recognize the object information, on the basis of the two-dimensional data generated by two-dimensional data generation unit.