Multi-Temporal Raft Culture Extraction for Medium-Resolution Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting raft culture areas from remote sensing images face challenges in large-scale accuracy, omission, and type identification, particularly with medium-resolution images, due to data source constraints, robustness limitations, and complex aquatic environments.
Innovation Solution
A method involving a raft culture marker sample library, a UNet network model enhanced with Atrous Spatial Pyramid Pooling (ASPP), shape constraints, and an attention mechanism, combined with NDWI index calculation and multi-temporal image fusion, to accurately extract and identify raft culture types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If single-temporal optical image extraction is used, then the extraction process is simple, but a large number of aquaculture areas are omitted and boundaries are blurred
Solution Approach 1:
The patent applies periodic action by using multi-temporal remote sensing images to extract aquaculture areas at different time points. This allows the system to capture aquaculture areas that may be missed in single images, reducing omission rates while maintaining manageable processing complexity through systematic temporal sampling.
Solution Approach 2:
The patent merges extraction results from multiple temporal images using a fusion algorithm. This combining approach integrates information from different time points, clarifies blurred boundaries through consistent features across time, and improves overall extraction accuracy while distributing the computational workload.
2Measurement precision
If deep learning methods are used, then extraction accuracy is improved, but comprehensive and authoritative sample library construction is lacking
Solution Approach 1:
The patent performs preliminary action by constructing a comprehensive sample library before the actual extraction process. This pre-prepared library contains diverse aquaculture area samples with standardized annotations, enabling the deep learning model to be trained on authoritative data, thereby improving extraction accuracy while establishing a reusable foundation that reduces future construction complexity.
Solution Approach 2:
The patent uses copying by creating standardized sample templates from actual aquaculture areas. These templates serve as reusable training data that can be replicated and adapted for different extraction scenarios, maintaining high accuracy while reducing the need to manually construct samples for each new extraction task.
3Productivity
If local area extraction experiments are performed, then model development is rapid, but large-scale and large-area raft culture extraction cannot be achieved
Solution Approach 1:
The patent applies segmentation by dividing large-scale remote sensing images into smaller sub-images or blocks that can be processed individually by the deep learning model. This allows the model to maintain its development speed on manageable segments while the systematic processing and fusion of multiple segments enable extraction across large-scale areas that would be impossible to process as a single image.
Data Source
AI summary
The present disclosure discloses an extraction method of raft culture area based on multi-temporal optical remote sensing images, including: constructing a raft culture marker sample library including culture types such as fish, shellfish, and algae; optimizing the deep learning model of the UNet network by using ASPP (Atrous Spatial Pyramid Pooling) and the shape constraint module; using the deep learning model to extract a corresponding multi-temporal raft culture area by using the multi-temporal optical remote sensing images with medium resolution in the target area; combining prior knowledge, fusing the extraction results of the raft culture area to obtain a final extraction results of the raft culture area.