Satellite Building Image Synthesis from AI Surface Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a need to generate synthetic images that simulate the appearance of new buildings in satellite images, particularly in undeveloped areas or areas obscured for security reasons, using artificial intelligence.
Innovation Solution
A method and device that utilize a neural network to infer the boundaries of building surfaces from satellite images, generate new buildings based on these boundaries, and synthesize them into the satellite images, considering positional relationships and user-provided information such as coordinates, dimensions, and surface areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a neural network is used to infer building boundaries from satellite images, then the accuracy of building boundary detection is improved, but the computational complexity and processing time increase
Solution Approach 1:
The building detection process is segmented into multiple stages: first detecting building boundaries using neural network, then segmenting the upper surface and side surface boundaries separately. This segmentation allows the complex neural network processing to be broken down into manageable steps, improving accuracy while controlling computational complexity through staged processing.
Solution Approach 2:
The system performs preliminary action by first inferring the building boundary from the satellite image before proceeding to generate the new building. This preliminary boundary detection using neural network establishes the foundation for subsequent building generation, ensuring accuracy is achieved early in the process while the complexity is managed through sequential operations.
2Manufacturing precision
If user information such as coordinates and dimensions is collected to generate new buildings, then the precision of building positioning is improved, but the operation complexity increases
Solution Approach 1:
The system implements self-service by automatically utilizing the collected user information (coordinates, dimensions, area) to generate and position the new building without requiring complex manual adjustment. The neural network processes this information automatically to create the building model, reducing operational complexity while maintaining high positioning precision through automated computation.
Solution Approach 2:
The system changes parameters such as building coordinates, width, height, and area to generate the new building. By transforming the user-provided information into specific building parameters through automated processing, the system achieves precise positioning while simplifying the user interaction to basic parameter input rather than complex operational procedures.
3Reliability
If the new building is synthesized into the satellite image considering positional relationships, then the naturalness of the synthetic image is improved, but the processing time increases
Solution Approach 1:
The system performs preliminary action by first determining the positional relationship between the new building and existing buildings before synthesizing the final image. This preliminary positioning calculation ensures that the new building is correctly integrated into the satellite image context, achieving naturalness through pre-computed spatial relationships rather than time-consuming real-time adjustments.
Solution Approach 2:
The system uses copying by generating a new building model that replicates the characteristics and positional relationships of existing buildings in the satellite image. The neural network copies the spatial patterns and relationships from the original image to create a realistic synthetic image, achieving naturalness through pattern replication rather than complex real-time rendering.
Data Source
AI summary
Disclosed is a method of generating a synthetic image based on artificial intelligence using a satellite image, which is performed by a processor. The method of generating a synthetic image based on artificial intelligence using a satellite image includes applying a satellite image captured by a satellite to a neural network to infer a boundary of an upper surface and a boundary of a side surface of a building included in the satellite image, and generating a new building based on the inferred boundary of the upper surface and the inferred boundary of the side surface of the building and synthesizing the new building into the satellite image.