Aircraft Pose Estimation Using Machine Learning Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated methods for estimating the pose of aircraft in images, such as position and orientation, are unreliable, especially when the aircraft is far from the camera and occupies a small portion of the image.
Innovation Solution
A computer-implemented image processing method using a machine-learning aircraft pose estimation model that includes segmentation, position determining, and orientation determining stages to identify aircraft parameters like range, position, and orientation, employing techniques like centre voting and semantic segmentation to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If shape recognition methods are used to detect specific parts of the aircraft, then the method can identify aircraft components, but the reliability of determining actual orientation deteriorates when the aircraft is further away and occupies a small portion of the image
Solution Approach 1:
The patent applies segmentation by dividing the image processing task into multiple components: detecting the aircraft body, identifying specific parts (nose, wingtips, tail), and determining pose parameters. This segmentation allows the system to handle different aspects of pose estimation separately, improving overall reliability even when the aircraft occupies a small image portion
Solution Approach 2:
The patent transitions from 2D image coordinates to 3D pose parameters by introducing depth estimation and spatial orientation calculations. This dimensional transformation enables the system to determine actual aircraft orientation and position in three-dimensional space, resolving the reliability issue when aircraft appear small in the image
2Extent of automation
If automated shape recognition is used for pose estimation, then the process can be performed automatically, but the method fails to reliably determine orientation when aircraft are at distances where they occupy only a small portion of the image
Solution Approach 1:
The patent performs preliminary detection of the aircraft body and specific parts before calculating pose parameters. This preliminary action establishes reference points and geometric relationships that are essential for accurate orientation measurement, enabling automated processing to maintain precision even at long distances
Solution Approach 2:
The patent introduces intermediate calculations and reference frames that mediate between the 2D image coordinates and the 3D pose parameters. These intermediaries include detected aircraft parts serving as reference points and calculated geometric relationships, which bridge the gap between automated detection and precise orientation measurement
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to pose estimation of aircraft, in particular to pose estimation of aircraft using machine learning. According to an aspect of the present invention, there is provided a computer-implemented method for performing pose estimation of aircraft. The method comprises: obtaining an input image; and using a machine-learning aircraft pose estimation model to obtain one or more aircraft parameters associated with one or more aircraft in the input image.