Adaptive Partitioning for High-Resolution Remote Sensing Target Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for high-resolution remote sensing image processing on edge computing platforms face challenges with resource occupation and low calculation efficiency, particularly when expanding input resolution, which leads to slower model training and increased power consumption.
Innovation Solution
A target detection method that adaptively divides high-resolution remote sensing images into different cluster areas based on target information, selecting appropriate models for each area, and dynamically allocating computing resources to prioritize detection accuracy and efficiency, thereby optimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the input resolution is expanded to improve detection accuracy, then the detection precision is improved, but the model training speed decreases and power consumption increases
Solution Approach 1:
The high-resolution remote sensing image is divided into multiple sub-images, which are then processed in parallel by multiple detection models. This segmentation approach maintains high detection accuracy while reducing the computational burden on each individual model, thereby improving overall processing speed and reducing power consumption.
Solution Approach 2:
Multiple detection models with different input resolution requirements are deployed to handle different sub-images. Some models process full-resolution sub-images for high-precision detection, while others process down-sampled sub-images for faster processing. This multi-functional approach allows the system to achieve both high detection accuracy and efficient processing speed.
2Measurement precision
If the input resolution is expanded to improve detection accuracy, then the detection precision is improved, but the power consumption increases
Solution Approach 1:
The image is segmented into multiple sub-images that are processed in parallel. This distribution of computational workload across multiple smaller tasks reduces the peak power consumption compared to processing one large high-resolution image at a time, while maintaining overall detection accuracy.
Solution Approach 2:
Different sub-images are processed with different model resolutions based on their specific characteristics. Sub-images containing important targets are processed with higher resolution models, while other areas use lower resolution models. This localized quality adjustment optimizes power consumption by applying high computational resources only where necessary.
3Productivity
If image segmentation is used to process high-resolution images, then the processing efficiency is improved, but the detection accuracy is affected
Solution Approach 1:
The image is divided into overlapping sub-images with appropriate overlap regions. This segmentation enables parallel processing efficiency while the overlap ensures that targets spanning multiple sub-images are not missed, thereby maintaining detection accuracy despite the division of the image.
Solution Approach 2:
The detection results from multiple sub-images are merged and integrated to form the final detection result. This merging process combines the advantages of parallel processing efficiency with comprehensive target detection, ensuring that no targets are lost during the segmentation process.
Data Source
AI summary
The present application discloses a target detection method and device for a high-resolution remote sensing image, which comprises: acquiring an original high-resolution remote sensing image from a sensor; acquiring target information of an area and an expanded area of the area expanding around by a predetermined distance; adaptively partitioning the original high-resolution remote sensing image to obtain different cluster areas to be detected, and obtaining an area required to be detected and an area not required to be detected; selecting a model of the area required to be detected and generating a target detection scheme; executing the target detection scheme to obtain a detection result; determining whether a computing platform has extra computing resources to detect the area not required to be detected, if so, performing dynamic partition detection for the area to obtain a detection result, and merging the detection results into a target detection result.