Mosaic Seamline Generation Using Semantic Segmentation to Avoid Structures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating geospatial mosaics fail to account for the semantic meaning of objects within images, leading to seamlines that often pass through important structures like buildings, roads, and bridges, resulting in unnatural-looking mosaics.
Innovation Solution
A machine learning-based approach that combines pixel-level similarity with semantic segmentation to generate seamlines that avoid structures by using a Convolutional Neural Network (CNN) to predict structure probabilities and integrate these scores with a distance-based cost function to find optimal seamlines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional seamline generation methods are used that rely only on pixel-level similarity, then the computational process is simple and fast, but the seamlines pass through important structures like buildings resulting in unnatural-looking mosaics
Solution Approach 1:
The patent applies segmentation by dividing the image processing into two distinct components: pixel-level similarity analysis and semantic structure detection. The semantic segmentation model separately identifies structures (buildings, roads, bridges) from the image data, allowing the seamline generation to respect these segmented regions while maintaining computational efficiency through specialized processing pathways for each component.
Solution Approach 2:
The patent introduces an intermediary semantic segmentation model that acts as a mediator between the raw image data and the seamline generation process. This intermediary component provides structure probability maps that guide the seamline optimization, enabling the system to avoid structures without requiring direct complex analysis between all image features.
2Reliability
If semantic segmentation is integrated into seamline generation to avoid structures, then the visual quality of mosaics is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by performing semantic segmentation and generating structure probability maps before the actual seamline optimization process. This preprocessing step identifies all structures in advance, allowing the subsequent seamline generation to efficiently navigate around them using pre-computed probability guidance rather than performing complex analysis during the optimization phase.
Solution Approach 2:
The patent changes parameters by transforming the semantic segmentation output into structure probability maps with specific numerical values representing the likelihood of each pixel belonging to a structure. This parameter transformation enables the seamline optimization algorithm to work with quantitative probability data rather than qualitative semantic labels, improving computational efficiency through numerical optimization.
3Reliability
If semantic segmentation is used to identify structures, then seamlines can avoid buildings and other structures, but the system requires machine learning models increasing device complexity
Solution Approach 1:
The patent applies universality by designing a semantic segmentation model that can identify multiple types of structures (buildings, roads, bridges, water bodies) using a single unified system. This multi-functional approach allows the same model architecture and processing pipeline to handle diverse structure types, reducing overall system complexity compared to having separate detection systems for each structure type.
Solution Approach 2:
The patent uses parameter changes by transforming diverse structure types into a unified probability representation. Instead of requiring separate processing pathways for different structure types, the system converts all structures into structure probability maps where the numerical probability values guide seamline avoidance, simplifying the system architecture through parameter standardization.
Data Source
Figure 1A~1C
Figure 2
Figure 3A~3D
AI summary
Methods and systems for avoiding buildings and other structures using semantic segmentation, the method including obtaining a plurality of image strips of a geographic location, identifying one or more structures that are present in one or more image strips, and generating one or more seamlines between adjacent image strips using a machine learning model. When two or more adjacent image strips include one or more structures, the method includes circumventing the one or more structures with the generated seamlines therebetween, wherein the generated one or more seamlines form a visual connection between the adjacent image strips.