Satellite Image Segmentation for Efficient Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The large data size of satellite images makes it difficult for users to handle and process them effectively.
Innovation Solution
An information processing apparatus and method that acquires and processes satellite images by dividing them into smaller, manageable cell images, which are then classified using machine learning classifiers to provide easy-to-handle observation data with added metadata.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If satellite images are provided as complete observation data, then comprehensive information is obtained, but data handling difficulty increases due to huge data size
Solution Approach 1:
The patent divides the satellite image into multiple cell images (e.g., 5km x 5km grid cells) to segment the large observation data into smaller, manageable units. This allows users to access only the necessary portions of the image rather than handling the entire large dataset, thereby reducing data handling difficulty while preserving comprehensive information about the target area.
Solution Approach 2:
The patent extracts and provides only the necessary cell images that contain the target object or area of interest, rather than providing the complete satellite image. This extraction approach reduces the data volume users must handle while maintaining the essential information needed for their purposes.
2Loss of information
If complete satellite images are provided, then full observation coverage is achieved, but processing time increases
Solution Approach 1:
By segmenting the satellite image into smaller cell images, the processing time is reduced because users only need to process the relevant cells containing the target object rather than the entire image. The segmentation maintains observation coverage for the target area while significantly decreasing processing requirements and time.
Solution Approach 2:
The patent extracts only the necessary cell images for processing, eliminating the need to process the entire satellite image. This extraction approach maintains full observation coverage for the area of interest while reducing processing time and computational resources required.
3Measurement precision
If machine learning classification is applied to cell images, then classification accuracy is improved, but computational resources increase
Solution Approach 1:
The patent applies machine learning classification to individual cell images rather than the entire satellite image. This segmentation approach maintains high classification accuracy for the target area while significantly reducing computational resources required, as only the relevant cells are processed through the classification algorithm.
Solution Approach 2:
The patent extracts and classifies only the necessary cell images containing the target object, rather than processing the entire satellite image. This extraction approach maintains classification accuracy for the area of interest while reducing computational resource consumption and energy usage.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing apparatus (1) comprises: an acquisition unit that acquires observation information obtained through observation of a target region from a flying object flying in outer space; a classification unit that inputs the observation information acquired by the acquisition unit to a classifier so trained as to output a classification result obtained by classifying a target object present in the target region if the observation information is input, and classifies the target object; an acceptance unit that accepts designation input for designating the target object; and an output unit that outputs the observation information including a classification result of the target object designated.