Aerial Robot Wire Avoidance Through Pixel-Based Distance Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Aerial robots struggle to detect thin objects such as electrical wires due to their size and shape, leading to potential collisions and damage.
Innovation Solution
Equipping aerial robots with image sensors and machine learning models, specifically convolutional neural networks, to identify thin objects by tracking pixel changes across images and estimating distances based on movement data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If aerial robots use conventional sensors to detect objects, then they can detect most objects in the environment, but they fail to detect thin objects such as electrical wires due to the objects' sizes and shapes
Solution Approach 1:
The patent transitions from relying solely on depth measurements (3D point cloud data) to incorporating 2D image data with pixel-level analysis. By examining objects in the 2D image plane and tracking pixel positions across multiple frames, the system can detect thin objects that lack sufficient depth information, effectively adding a dimensional approach to detection.
Solution Approach 2:
The patent replaces conventional sensor-based detection (relying on physical sensor limitations) with a computational approach using neural networks and image processing algorithms. This substitution of mechanical/sensor-based detection with information-processing methods enables the system to overcome physical sensor constraints and detect thin objects through pattern recognition in image data.
2Extent of automation
If aerial robots navigate autonomously without manual control, then they can operate independently, but they collide with thin objects such as wires because the robots cannot detect them
Solution Approach 1:
The patent implements a feedback mechanism by continuously capturing images across multiple frames and tracking pixel positions of detected objects. The system compares object positions between frames, uses distance data to estimate movement, and adjusts navigation decisions based on this feedback loop, enabling autonomous detection and avoidance of thin objects that would otherwise be undetectable.
Solution Approach 2:
The patent performs preliminary detection and classification of objects as thin or non-thin before final navigation decisions are made. By identifying potential thin objects early in the processing pipeline and tracking them across frames, the system prepares collision avoidance actions in advance rather than reacting too late, improving the reliability of autonomous navigation.
3Loss of information
If aerial robots rely on point cloud data with depth measurements, then they can obtain 3D information about the environment, but they cannot generate sufficient depth measurements for thin objects such as wires
Solution Approach 1:
The patent creates a complementary 2D representation (image data with pixel coordinates) of the environment that serves as a copy or alternative view to the 3D point cloud data. This 2D copy contains information about thin objects that is lost in the 3D representation, allowing the system to recover depth and position information for objects that appear insufficient in the point cloud data.
Data Source
AI summary
An aerial robot includes an image sensor for capturing images of an environment. The robot receives a first image captured at a first location. The robot identifies one or more first pixels in the first image. The first pixels correspond to one or more targeted features of an object identified in the first image. The robot receives a second image captured at the second location. The robot receives its distance data that estimates a movement of the robot from the first location to the second location. The robot identifies second pixels in the second image. The second pixels corresponding to the targeted features of the object as appeared in the second image. The robot determines an estimated distance between the robot and the object based on the changes of locations of the second pixels from the first pixels relative to the movement of the robot provided by the distance data.