UAV Obstacle Avoidance Using Multi-Focal-Length Vision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing obstacle detection systems for unmanned aerial vehicles face measurement interference and positioning errors due to onboard inertial devices, affecting the precision of depth information calculation during autonomous flight and return.
Innovation Solution
The system employs a camera apparatus with optical zoom capabilities to obtain detection images at multiple focal lengths, calculating obstacle position information, including distance and height difference, without relying on precise GPS and inertial data, and adjusts the vehicle's velocity direction to avoid obstacles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS and onboard inertial devices are used to obtain position and attitude information, then obstacle detection can be performed, but measurement interference and positioning errors occur, reducing calculation precision
Solution Approach 1:
The patent extracts and removes the dependency on GPS and onboard inertial devices from the depth calculation process. By using only visual information from monocular stereoscopic vision and binocular stereoscopic vision, the system eliminates the measurement interference and accumulated errors from inertial sensors, thereby improving calculation precision without relying on imprecise positioning data
Solution Approach 2:
The patent introduces an alternative intermediary system for obtaining position and attitude information - using visual odometry and feature point matching from camera images instead of inertial sensors. This intermediary approach provides more reliable positioning data by comparing sequential images and tracking feature points, avoiding the accumulated errors of inertial integration
2Length of stationary object
If monocular stereoscopic vision is used for remote obstacle detection, then detection range is extended, but calculation precision is affected by inertial device errors
Solution Approach 1:
The patent segments the obstacle detection system into two distinct functional parts: monocular stereoscopic vision for remote detection and binocular stereoscopic vision for proximity detection. This segmentation allows each subsystem to operate in its optimal range - monocular for extending detection range and binocular for providing precise depth information when the obstacle is closer, thereby maintaining high precision across the entire detection range
Solution Approach 2:
The patent uses visual feature point matching and image processing as an intermediary method to calculate depth information without relying on inertial device data. By tracking feature points across sequential images and using geometric relationships, the system obtains precise depth measurements that are independent of GPS and inertial sensor errors
3Measurement precision
If binocular stereoscopic vision is used for proximity detection, then depth information precision is improved, but detection range is limited by baseline length
Solution Approach 1:
The patent segments the detection system into monocular and binocular components, with binocular stereoscopic vision dedicated to proximity detection where high precision is needed. The limited detection range of binocular vision is acceptable in this segment since it only needs to cover the immediate vicinity of the UAV, while the monocular system handles remote detection
Solution Approach 2:
The patent merges monocular and binocular stereoscopic vision systems into a unified obstacle detection framework. The systems work complementarily - monocular vision provides remote detection capability while binocular vision provides precise proximity measurement. The integration allows the UAV to switch between or combine both methods based on distance, achieving both extended range and high precision
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances calculation precision by reducing measurement interference and positioning errors, enabling effective obstacle avoidance without the need for precise GPS and inertial data, improving the reliability of obstacle detection and navigation.
Implementation Method 1
The camera apparatus supports optical zoom
Implementation Method 2
obtaining depth information of a point by using a pixel position difference that is obtained by matching the same point in the images of the different viewing angles
Data Source
Figure 1~2
Figure 3
Figure 4a~5b
AI summary
Embodiments of the present invention disclose an obstacle avoidance method and apparatus and an unmanned aerial vehicle. The method includes: obtaining detection images of at least two different focal lengths from a camera apparatus; determining, according to the detection images of the at least two different focal lengths, that an obstacle exists in a detection area; and obtaining position information of the obstacle according to the detection images of the at least two different focal lengths. In the embodiments of the present invention, the at least two different focal lengths are set for the camera apparatus. The detection images of the different focal lengths are obtained. Then it is determined, according to the detection images of the different focal lengths, whether an obstacle exists in the detection area. Position information of the obstacle is obtained according to the detection images. Because the camera apparatus performs a zooming operation quickly, it's no need to precisely obtain position and attitude information by using a Global Positioning System (GPS) and an onboard inertial device, which avoids measurement interference and a positioning error caused by the inertial device, thereby improves calculation precision.