GeoAI Data Cube Inference With Task-Model Matching at Scale
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing spatiotemporal big data computing platforms face challenges in achieving high inference accuracy and speed in multi-spatiotemporal or large-scale scenarios due to high computational complexity and dependency on specific temporal and spatial conditions, which limits the performance of single models.
Innovation Solution
An AI Cube-based online inference method that integrates spatiotemporal data and GeoAI models using a unified cube organizational model, employing explicit and implicit matching mechanisms, and a distributed inference framework for parallel processing of tile data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single GeoAI model is used for inference in multi-spatiotemporal scenarios, then the system complexity is low, but the inference accuracy deteriorates due to dependency on specific temporal and spatial conditions
Solution Approach 1:
The patent implements a multi-model inference system where multiple GeoAI models with different spatial and temporal adaptabilities work together. The system includes a first GeoAI model for spatial inference and a second GeoAI model for temporal inference, allowing the system to handle diverse spatiotemporal scenarios universally while maintaining high accuracy through model selection based on data characteristics
2Productivity
If distributed parallel processing is implemented for tile data inference, then the inference speed improves, but the computational resource requirements increase
Solution Approach 1:
The patent divides the inference task into multiple segments by splitting tile data into different regions and processing them in parallel. The processor divides a second target tile into multiple third target tiles and performs parallel inference on each segment, significantly improving inference speed while managing computational resources through distributed processing
Solution Approach 2:
The system implements incremental inference by processing tile data in parts rather than all at once. It performs preliminary inference on entire tiles first, then focuses computational resources on boundary regions where accuracy needs improvement, avoiding excessive computation on already-processed areas while maintaining overall accuracy
3Measurement precision
If comprehensive model matching with multiple criteria is performed, then the task-to-model matching precision improves, but the matching time increases
Solution Approach 1:
The patent pre-calculates and stores spatial and temporal coverage information for each GeoAI model in advance. When a inference task arrives, the system quickly matches models by comparing task requirements with pre-stored model characteristics, avoiding time-consuming real-time analysis while ensuring accurate matching through comprehensive pre-evaluation of model capabilities
4Measurement precision
If boundary region refinement is performed to improve accuracy, then the inference precision in boundary areas improves, but the processing time increases
Solution Approach 1:
The patent applies different processing quality levels to different regions of tile data. Boundary regions receive refined processing with higher computational effort to improve accuracy, while internal regions use standard processing. This local differentiation ensures high precision at critical boundaries without unnecessarily increasing processing time for entire tiles
Data Source
AI summary
The disclosure pertains to the field of data analysis and services, specifically to an online inference method, service system, and device based on an artificial intelligence geospatial data cube. The method involves constructing a cube organizational model based on a spatiotemporal grid, where the cube organizes and manages geospatial data and GeoAI models in a unified manner. It performs task-oriented model matching through a combination of explicit and implicit matching, where explicit matching converts user inference requests into multidimensional query conditions to retrieve candidate models, and implicit matching determines the optimal model by computing the feature similarity between inference data tiles and candidate models. Based on a distributed inference framework, an efficient inference workflow is executed in parallel across multiple computing nodes. The disclosure aims to address the limitations of existing technologies in multi-scenarios or large-scale spatial scenarios, such as low inference accuracy and slow inference speed.