Object-Based Target Detection Ranking in Spectral Imagery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing target detection processes in spectral digital imagery are hindered by massive data volumes and imperfect algorithms, requiring extensive manual verification, making real-time detection impractical due to time constraints.
Innovation Solution
A method and apparatus for object-based sorting and ranking of target detections, utilizing a statistical target detection filter to determine scores for pixels, identifying regions with higher scores, and providing object-based rankings, which can be automated and integrated with geospatial systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection of detection planes is performed for each target, then detection accuracy is improved, but analysis time increases significantly
Solution Approach 1:
The patent segments the analysis process into automated detection phase and selective verification phase. The automated system processes all detection planes using spectral matching algorithms, then segments results into high-confidence detections (requiring no manual review) and low-confidence detections (requiring manual verification). This segmentation reduces manual inspection time while maintaining detection accuracy through targeted verification of only uncertain cases.
Solution Approach 2:
The patent introduces an intermediary automated detection system that acts as a mediator between raw spectral data and manual analysis. This intermediary system performs initial filtering, scoring, and ranking of detection planes using spectral libraries and statistical algorithms, presenting only the most promising candidates to human analysts. This intermediary layer significantly reduces the volume of data requiring manual inspection while preserving detection accuracy through automated pre-screening.
2Reliability
If comprehensive manual verification of all detection planes is performed, then detection reliability is improved, but productivity decreases
Solution Approach 1:
The patent applies partial verification action by performing comprehensive automated detection on all pixels followed by selective manual verification only for detections below a confidence threshold. Rather than verifying all detections equally, the system performs excessive automated processing on high-confidence cases (where automation alone suffices) and partial manual processing on low-confidence cases, optimizing the balance between reliability and productivity through differentiated verification intensity.
Solution Approach 2:
The patent replaces the mechanical manual verification process with an automated computational system for the majority of detection cases. Statistical detection algorithms, spectral matching computations, and automated scoring systems substitute human analysts for routine verification tasks. This mechanical-to-computational substitution maintains detection reliability through rigorous algorithmic validation while dramatically increasing productivity by processing volumes of data beyond human capacity.
3Measurement precision
If spectral detection algorithms process every pixel in every image, then detection completeness is improved, but data processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-computing spectral signatures from reference libraries, pre-calculating background spectral characteristics from each scene, and pre-ranking pixels by their spectral distinctiveness before full detection processing. These preliminary computations prepare optimized data structures and statistical parameters in advance, enabling faster complete processing of all pixels while maintaining detection completeness through pre-established spectral comparison frameworks.
Solution Approach 2:
The patent dynamically changes processing parameters based on scene characteristics and detection confidence levels. Statistical thresholds, spectral matching sensitivity, and verification requirements are adjusted as parameters based on initial pass results. High-confidence detections use streamlined processing with fewer computational steps, while low-confidence detections trigger more intensive parameter-based verification, optimizing processing time across the complete dataset while maintaining comprehensive detection coverage.
Data Source
AI summary
A method, non-transitory computer readable medium, and apparatus that provides object-based identification, sorting and ranking of target detections includes determining a target detection score for each pixel in each of one or more images for each of one or more targets. A region around one or more of the pixels with the determined detection scores which are higher than the determined detection scores for the remaining pixels in each of the one or more of images is identified. An object based score for each of the identified regions in each of the one or more images is determined. The one or more identified regions with the determined object based score for each region is provided.