Multi-Document Scanning with Corner Detection and AR Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document scanning methods struggle with accurately identifying and processing multiple documents in a scene, leading to inefficient use of processing power and reduced accuracy due to background interference and orientation issues.
Innovation Solution
A method using a multi-layer convolutional neural network to identify corner points of documents, frame them for processing, and map to a 3D augmented reality system, guiding users to optimal scanning positions for clear images, and automatically capturing pictures for further processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire scene is processed for document scanning, then coverage of all potential documents is maximized, but processing power is wasted on background areas and accuracy is reduced due to background interference
Solution Approach 1:
The patent segments the scene into multiple potential document regions by detecting corner points and forming frames around each detected document. This segmentation allows the system to process only the relevant document areas identified by corner point detection, excluding background interference while maintaining comprehensive document coverage.
Solution Approach 2:
The patent extracts only the relevant portions of the image that contain documents by using corner point detection to identify and isolate document frames. This extraction mechanism removes background elements from the processing pipeline, improving both accuracy by eliminating background interference and efficiency by reducing the processing area.
2Measurement precision
If documents are processed without orientation correction, then processing speed is maintained, but OCR accuracy decreases due to rotated text
Solution Approach 1:
The patent performs orientation detection and correction as a preliminary action before the main OCR processing. By detecting the orientation of documents through corner point analysis and applying rotation transformations in advance, the system ensures text is properly oriented for OCR while minimizing overall processing time through efficient algorithms.
3Measurement precision
If manual document selection is required, then processing accuracy can be controlled, but user time and operational complexity increase
Solution Approach 1:
The patent implements self-service functionality where the system automatically detects documents, determines their locations through corner point detection, and prepares them for processing without requiring manual user selection. This automation maintains high processing accuracy while significantly improving ease of operation by eliminating manual intervention steps.
4Measurement precision
If corner point detection is performed on all detected objects, then document identification accuracy is maximized, but system complexity and processing time increase
Solution Approach 1:
The patent applies local quality by detecting corner points specifically at locations where documents are likely to appear, rather than uniformly analyzing the entire image. This targeted approach uses document probability maps and spatial priors to focus computational resources on high-probability regions, maintaining high detection accuracy while reducing overall system complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of scanning paper document(s) for further processing includes filming/displaying a scene on a user device and recognizing the document(s), identifying corner point coordinates of the document(s) using a multi-layer convolutional neural network, building a frame around each of the recognized paper document(s) using the corner point coordinates and mapping these coordinates to an augmented reality engine's coordinate system and highlighting the document(s) on the user device's display by a highlighted object frame around each document. The user is guided by commands to move the user device to a scanning position to take a picture of the document(s) with a resolution, object coverage, sharpness suitable for further processing of the at least one paper document, and a picture of at least one of the paper document(s) is automatically taken when the user device has arrived at said scanning position, the picture being transmitted for further processing.