Occluded Symbol Identification in Engineering Drawings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying occluded objects in engineering drawings is challenging due to occlusions such as stamps, smudges, and ink blotches, which obscure the distinguishing features of symbols, leading to inaccuracies in image analysis.
Innovation Solution
A computer-implemented method that retrieves an image with occluded objects, removes occlusions to create a clear image, runs a detection model to predict symbol identifications, and applies geometric properties and contextual information like text and subcomponents to determine the probability of correct identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple camera angles with 3D images are used to identify occluded objects, then depth estimation and shape reconstruction improve, but the method cannot be applied to 2D shadowless engineering drawings
Solution Approach 1:
The patent transitions from 3D image analysis to 2D drawing analysis by leveraging geometric properties inherent in engineering drawings. Instead of using multiple camera angles to reconstruct 3D shapes, the system analyzes 2D geometric properties such as symmetry axes, line intersections, and shape characteristics to identify occluded symbols in flat engineering drawings.
2Measurement precision
If feature matching techniques with pyramid arc-length descriptors are used, then local curve and edge matching improve, but the method requires exact knowledge of shape features which is unavailable for occluded symbols
Solution Approach 1:
The system performs preliminary analysis by detecting geometric properties such as symmetry axes and line intersections from the visible portions of occluded symbols. These preliminary geometric constraints are then used to guide the identification process, allowing the system to predict what features should be present even when occluded, without requiring complete shape information.
Solution Approach 2:
The system uses iterative feedback by comparing detected geometric properties against known symbol databases. The geometric constraints derived from visible portions provide feedback that narrows down possible symbol identities, which is then refined by checking consistency with engineering drawing context and symbol relationships.
3Loss of information
If NLP and word-to-vector models are used to predict occluded objects from captions, then contextual prediction improves, but engineering drawings lack captions and contextual relationships between symbols
Solution Approach 1:
The patent introduces geometric properties as an intermediary between the visual data and symbol identification. Instead of using NLP to interpret captions, the system uses geometric feature extraction as a mediator that translates visual patterns in occluded symbols into identifiable characteristics, bridging the gap between partial visual information and complete symbol recognition.
Data Source
AI summary
In an approach to identifying occluded objects, a computer retrieves a first image that includes an object at least partially occluded by one or more occlusions. A computer removes the one or more occlusions from the first image to create a partial object in a second image. A computer runs a detection model with the second image to predict one or more identifications of a symbol represented by the partial object. A computer determines top predictions of the one or more identifications of the symbol by the detection model. A computer identifies at least one geometric property associated with the one or more identifications of the symbol included in the one or more top predictions. A computer applies the at least one geometric property to the partial object. A computer determines a probability of the one or more top predictions correctly identifying the symbol represented by the partial object.