Aerial Target Detection Using Segmented Confidence Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The detection accuracy of target objects in aerial images captured from flying objects deteriorates due to the reduction in object size, leading to potential erroneous results, as the height of the flying object increases.
Innovation Solution
A target object detection device that divides aerial images into multiple segments, calculates a confidence score for each segment, and displays a discrimination image highlighting areas with high confidence scores, allowing for manual confirmation and re-imaging of specific regions for detailed examination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If the height of the flying object is increased to capture aerial images from a larger region, then the coverage area is improved, but the size of the target object in the image decreases leading to deteriorated detection accuracy
Solution Approach 1:
The aerial image is divided into multiple smaller images (e.g., 4 or more images) that cover the same large region. This segmentation allows the system to maintain both wide coverage area and sufficient detail for accurate target detection, as each smaller image can be processed independently to identify targets while the collection of images provides comprehensive coverage.
Solution Approach 2:
The system transitions from using a single large aerial image to a multi-image approach, adding the dimension of image collection and processing. By capturing multiple images from slightly different positions or angles and then processing them together, the system achieves both broad coverage and detailed detection capability.
2Measurement precision
If the aerial image is divided into multiple smaller images to improve detection accuracy, then the detection precision is improved, but the complexity of image processing increases
Solution Approach 1:
The aerial image is divided into multiple smaller images (e.g., 4 or more images) that cover the same large region. This segmentation allows the system to maintain both wide coverage area and sufficient detail for accurate target detection, as each smaller image can be processed independently to identify targets while the collection of images provides comprehensive coverage.
Solution Approach 2:
The system processes multiple smaller images and uses the results to generate a comprehensive determination about target objects in the original large region. The confidence scores and detection results from individual smaller images provide feedback that is aggregated to make final detection decisions, improving overall accuracy while managing processing complexity through systematic evaluation.
Data Source
AI summary
A target object detection device capable of providing an image useful for discovering a target object from an aerial image is provided. A target object detection device for detecting at least one target object from an aerial image is provided with an imaging device, a computer, and a display device. The computer is configured to divide the first aerial image into a plurality of first images, output a confidence score that the target object is included, for each of the plurality of first images, and cause the display device to display a first discrimination image capable of discriminating a degree of the confidence score and a portion of the first aerial image corresponding to the confidence score.