3D Orthoimage Mapping with Automated Object Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The manual process of generating orthoimages for three-dimensional surfaces, particularly façades, is tedious and error-prone, requiring significant user input to define mapping areas and volumes, which is inefficient for complex shapes.
Innovation Solution
A computer-implemented method that automatically selects mapping areas and content using machine learning and neural networks, such as PointNet and PointCNN, to generate orthoimages by specifying a 3D mapping volume and performing segmentation to identify wanted and unwanted objects, with user validation and refinement options.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual processes are used to define mapping areas and volumes, then user control and precision are improved, but the process becomes tedious and error-prone
Solution Approach 1:
The system performs automatic segmentation and object identification without requiring manual intervention. The computer automatically defines mapping areas and volumes by processing structure data and identifying wanted versus unwanted objects, eliminating the need for users to manually create bounding boxes or perform masking operations.
Solution Approach 2:
The patent replaces manual mechanical operations (dragging, rotating, positioning bounding boxes) with automated computational processes. Machine learning algorithms and neural networks substitute for human operators in defining mapping parameters, thereby reducing both time consumption and human error while maintaining precision.
2Manufacturing precision
If manual masking processes are used for complex shapes, then mapping accuracy is improved, but the process becomes more complex and error-prone
Solution Approach 1:
The patent applies segmentation by dividing the 3D structure data into distinct objects and regions. The system automatically segments wanted objects (façade features) from unwanted objects (nearby structures) based on spatial relationships and object characteristics, eliminating the need for manual masking while maintaining accuracy for complex shapes.
Solution Approach 2:
The system changes parameters automatically by computing mapping volume parameters (bounds, orientation, extent) based on the segmented objects. Instead of requiring manual parameter input, the system derives mapping parameters from the 3D structure data and object segmentation results, simplifying the process while maintaining precision.
3Productivity
If automated methods are used to select mapping content, then productivity is improved, but automation extent and error risk are worsened
Solution Approach 1:
The system incorporates feedback mechanisms where segmentation results and object identification are validated against the original structure data. The automated selection process includes verification steps that compare selected mapping content with the input data, ensuring accuracy while maintaining automation. Errors can be detected and corrected through this feedback loop.
Solution Approach 2:
The patent introduces an intermediary processing stage between raw structure data and final orthoimage generation. The segmentation and object identification modules act as intermediaries that prepare and validate the mapping content selection, ensuring that automated processes produce reliable results by filtering and verifying data before final image generation.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The invention pertains to a computer-implemented method (100) for generating an orthoimage of a 3D structure, the method comprising receiving (110) structure data comprising object data related to wanted objects and unwanted objects, selecting (120), based on the structure data, a mapping area including at least a part of the 3D structure, selecting (140) mapping content comprising one or more wanted objects, and generating (150) an orthoimage showing the mapping content, wherein selecting (140) the mapping content comprises, at least partially automatically, specifying (130) a 3D mapping volume enclosing the mapping area, wherein the selected mapping content comprises objects that are located in the mapping volume, and/or performing a segmentation (135) of the surface data to identify wanted and/or unwanted objects.