Aerial Image Stitching Using Metadata and Top-Level Reference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for stitching aerial images from small unmanned aerial vehicles (UAVs) are time-consuming due to the need to account for large pitch and roll angles, often requiring heavy and costly camera gimbal mounts, and lack efficient use of image metadata such as latitude, longitude, altitude, pitch, roll, and yaw angles.
Innovation Solution
A method and system that utilize image metadata to generate transformed images, calculate quality of fit, normalize captured images, and assemble a new aerial image by fitting them to a top-level image, reducing processing power and time by using variables for each parameter and incorporating pitch, roll, and yaw angles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional stitching methods are used to assemble aerial images from small UAVs, then accurate mosaic images can be generated, but the stitching time becomes extremely long (several hours)
Solution Approach 1:
The patent applies preliminary action by using a top-level image to pre-establish the geographic framework and boundaries of the mosaic area before stitching individual aerial images. This preliminary setup allows the stitching algorithm to quickly determine where each image should be positioned and how it should be transformed, eliminating the need to process all possible angle combinations during the actual stitching operation. The top-level image serves as a pre-computed reference that guides the rapid placement and orientation of subsequent images.
2Measurement precision
If camera gimbal mounts are used to maintain stable camera orientation during flight, then image quality improves, but the system becomes heavier and more expensive
Solution Approach 1:
The patent replaces the mechanical gimbal stabilization system with a computational approach. Instead of using physical mechanisms to maintain camera orientation during flight, the system captures images with varying pitch and roll angles and then applies geometric transformations and perspective corrections during the stitching process. This substitution eliminates the need for heavy mechanical gimbals while achieving the same goal of producing high-quality mosaic images from images captured at different angles.
3Adaptability or versatility
If stitching software is designed to tolerate large pitch and roll angles, then compatibility with small UAVs improves, but processing time and computational resources increase
Solution Approach 1:
The patent applies local quality by treating different regions of the image set differently based on their characteristics. Images captured with small pitch and roll angles are processed using standard stitching algorithms, while images with large pitch and roll angles undergo specialized perspective transformation and geometric correction. The top-level image provides a reference framework that allows the system to selectively apply different processing strategies to different images, optimizing both compatibility and processing efficiency.
Data Source
AI summary
A method, a computer program product, and a system are disclosed for stitching aerial data using information from at least one previous image. The method includes capturing a plurality of images of the landscape; obtaining, image metadata for each of the captured images; generating, for each of the captured images, a set of transformed images based on the image metadata, comprises: setting a variable for each of the parameters; preparing a plurality of sets of transformed image metadata by applying the variables to the parameters; and preparing the set of transformed images from the captured image based on the plurality of sets of transformed image metadata, respectively; identifying, for each set of transformed images, one of the transformed images by calculating quality of fit to the top level image for each of the transformed images; and assembling a new aerial image based on the plurality of the identified transformed images.