Remotely Sensed Image Target Detection via Adaptive Pixel Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for retrieving targets from remotely sensed data are inefficient, particularly in large-scale applications, due to the need for scanning entire images and the challenges of identifying targets with insufficient or noisy information, which is not scalable and time-consuming.
Innovation Solution
A method and system that uses computation-efficient, time-efficient, and resource-efficient sampling techniques to analyze a small percentage of pixels, employing active, evenly distributed, or hybrid sampling strategies to detect and localize targets based on spectral and spatial properties, using machine learning models to compute an evidence score for target presence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If entire images are scanned to identify targets, then detection accuracy is improved, but processing time and computational resources increase significantly
Solution Approach 1:
The patent divides the image into multiple blocks and further segments each block into super-pixels or regions. This segmentation allows the system to process smaller units independently, applying sampling techniques to select representative pixels from each region rather than scanning all pixels, thus reducing processing time while maintaining detection accuracy through strategic sampling of key areas.
Solution Approach 2:
The patent applies partial action by scanning only a subset of pixels (sampling) rather than the entire image. It uses strategies like uniform sampling, adaptive sampling, or sampling based on regions of interest to process a partial but representative portion of the image data, achieving acceptable detection accuracy with significantly reduced computational effort and time.
2Measurement precision
If manual tagging is performed by experts, then target identification accuracy is improved, but the process is not scalable to large volumes of data
Solution Approach 1:
The patent implements self-service by enabling the system to automatically perform target detection and image suitability assessment without requiring manual expert tagging. The algorithm autonomously analyzes image blocks, applies sampling techniques, computes target presence indicators, and generates results independently, making the process scalable to large volumes of remotely sensed data while maintaining consistent accuracy.
Solution Approach 2:
The patent replaces the mechanical manual process of expert tagging with an automated computational system. Instead of relying on human experts to manually examine and tag images, the system uses algorithmic processing, sampling strategies, and automated target detection mechanisms to substitute human labor, thereby achieving both accuracy and scalability.
3Productivity
If automated target detection is applied to wide coverage images, then productivity is improved, but detection accuracy decreases for non-conspicuous targets
Solution Approach 1:
The patent applies local quality by treating different regions of the image differently based on their characteristics. It divides the image into blocks and identifies regions of interest or areas with higher probability of containing targets. Sampling strategies are adapted locally to focus computational resources on areas more likely to contain non-conspicuous targets, improving detection accuracy in critical regions while maintaining overall productivity.
Solution Approach 2:
The patent performs preliminary action by pre-processing images to identify potential regions of interest before conducting detailed target detection. It uses initial analysis to flag areas that may contain targets, even non-conspicuous ones, and then applies more rigorous sampling and detection methods to these pre-identified regions, improving the chances of detecting subtle targets while maintaining scalability.
4Speed
If sampling techniques are used to reduce processing load, then processing speed is improved, but the risk of missing targets increases
Solution Approach 1:
The patent incorporates feedback mechanisms where the results from sampled pixels are used to guide further processing. If targets are detected in sampled regions, the system can adjust sampling density in surrounding areas or trigger more detailed analysis. This feedback loop allows the system to maintain high processing speed while dynamically adjusting the level of scrutiny based on initial findings, thereby reducing the risk of missing targets.
Solution Approach 2:
The patent uses parameter changes by adjusting sampling density, block size, and analysis thresholds based on image characteristics and target properties. It can dynamically modify these parameters to balance processing speed and detection reliability - using coarser sampling for large-scale surveys and finer sampling when targets are suspected or in critical regions, thus optimizing the trade-off between speed and reliability.
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
State of art techniques performing image labeling of remotely sensed data are computation intensive, consume time and resources. A method and system for efficient retrieval of a target in an image in a collection of remotely sensed data is disclosed. Image scanning is performed efficiently, wherein only a small percentage of pixels from the entire image are scanned to identify the target. One or more samples are intelligently identified based on sample selection criteria and are scanned for detecting presence of the target based on cumulative evidence score Plurality of sampling approaches comprising active sampling, distributed sampling and hybrid sampling are disclosed that either detect and localize the target or perform image labeling indicating only presence of the target.