Image Processing Method for UAV Target Prioritization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Operators of remotely-controlled equipment, such as UAVs, face challenges in correctly identifying and prioritizing potential targets in images, as existing methods fail to effectively distinguish between relevant features of interest and false alarms, particularly in crowded scenes.
Innovation Solution
An image-processing method that convolves selected regions of interest with masks of different sizes, calculates the ratio of convolution responses, and assigns scores based on the size and contrast of features, using techniques like Laplacian of Gaussian convolution to prioritize targets by size, while filtering and thresholding to reduce noise and emphasize relevant features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple potential targets are presented to the operator, then the operator can identify more targets, but the operator cannot distinguish between relevant targets and false alarms
Solution Approach 1:
The patent applies parameter changes by computing multiple image quality metrics (sharpness, contrast, saturation, etc.) for each detected potential target and using these parameters to rank and filter targets. This transforms the raw target detection into a prioritized list where operators can focus on the most relevant targets first, resolving the contradiction between presenting multiple targets and maintaining identification accuracy
Solution Approach 2:
The patent replaces manual target prioritization with an automated computer-based system that objectively measures and ranks targets using image processing algorithms. This substitution of mechanical/manual assessment with computational analysis enables reliable differentiation between relevant targets and false alarms at scale
2Loss of information
If all features in the image are presented to the operator, then no information is lost, but the operator cannot efficiently identify the most important targets
Solution Approach 1:
The patent segments the set of all detected targets into prioritized groups based on image quality metrics. By dividing the comprehensive target list into ranked categories (highest priority, medium priority, lower priority), the system preserves all target information while enabling operators to efficiently process targets in order of importance, thus maintaining information completeness while improving productivity
3Quantity of substance
If small high-contrast features are detected, then more potential targets are identified, but small features like decoy flares cannot be distinguished from larger legitimate targets
Solution Approach 1:
The patent adds the dimension of target size measurement to the detection process by computing bounding box dimensions and area metrics for each potential target. This additional dimensional information (size) allows the system to differentiate between small decoy flares and larger legitimate targets, resolving the ambiguity that arises when only contrast information is available
Data Source
Figure 1~2
Figure 3~5
Figure 6~7
AI summary
An image-processing method comprising convolving a selected feature of interest (FOI) within the image with a mask of a first size, repeating the convolution with a mask of a second size, and calculating the ratio of the convolution responses, as an indication of the size of the FOI. Preferably the convolution masks are Laplacian of Gaussian. The method can be useful for prioritising potential targets in a field of view for presentation to an operator.