Wind Turbine Object Capture for Autonomous Drone Flight Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating calibration data for autonomous drone flights over wind turbines are hindered by insufficient or inaccurate manufacturer data and lack of reliable models, such as CAD models, which complicates the determination of flight paths and waypoints.
Innovation Solution
A method involving a drone capturing multiple images of a wind turbine from different positions, with each image assigned position and location information, allowing for feature recognition and position determination using AI or manual evaluation, to create a model of the turbine for autonomous flight planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If manufacturer data and pre-calculated waypoints are used for autonomous drone flights, then flight path determination is simplified, but the precision and reliability of position data become insufficient
Solution Approach 1:
The patent performs preliminary actions by capturing images and determining feature positions before the actual autonomous flight. A calibration flight is executed first to capture images of the wind turbine from multiple positions, and feature positions are determined in advance. This preliminary calibration data is then used to generate accurate waypoints for the subsequent autonomous inspection flight, resolving the contradiction by preparing precise position information beforehand without requiring complex real-time processing during the flight.
2Measurement precision
If detailed calibration data is captured to improve position determination accuracy, then more images and feature recognition are required, but the time and computational resources increase
Solution Approach 1:
The patent applies partial action by selecting only the most relevant features for position determination rather than processing all possible image data. Specifically, it identifies and uses key features such as the wind turbine tower center, rotor blade tips, and rotor blade flanges. This selective approach captures sufficient calibration data to achieve accurate position determination without the excessive time and computational resources that would be required to process every detail in the captured images.
3Measurement precision
If manual feature recognition is used to ensure accurate position determination, then quality is maintained, but the inspection process becomes slower and requires trained personnel
Solution Approach 1:
The patent implements self-service by enabling the system to automatically recognize features and determine their positions without requiring manual intervention. The evaluation unit autonomously processes the captured images, identifies features such as the wind turbine tower and rotor components, and calculates their positions using the assigned position and orientation information. This automation maintains accuracy while significantly increasing inspection speed and eliminating the need for trained personnel to manually analyze images.
4Adaptability or versatility
If autonomous flight paths are calculated without reliable CAD models, then flexibility is improved, but the safety and precision of collision avoidance deteriorate
Solution Approach 1:
The patent resolves this contradiction by performing preliminary actions to create an accurate spatial model of the wind turbine before the autonomous flight. During the calibration flight, images are captured from multiple positions and angles, and feature positions are determined to generate a reliable geometric representation of the turbine structure. This pre-acquired spatial information enables the autonomous flight system to navigate flexibly while maintaining high collision avoidance safety, as the drone can reference the pre-established model to avoid obstacles without requiring a pre-existing CAD model.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for capturing an object that can be described by a plurality of predetermined features, which method comprises take-off of the object and capture of a plurality of portions of the object by at least one capturing unit. Each of the portions is captured multiple times from different positions of the capturing unit in order to generate a set of images. Position and location information of the capturing unit are assigned to each image. The method further comprises recognising features in the sets of images and determining the positions and/or locations of the features using the position and location information of the images which contain the features.