UAV Distance Measurement Using Edge ROIs for Texture-Poor Obstacles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Unmanned aerial vehicles (UAVs) face challenges in remote-distance obstacle avoidance due to poor stereo matching precision and unavailability of stereo matching in extreme cases with no texture, low textures, or dense and repeated textures, which hinder effective obstacle detection and avoidance.
Innovation Solution
The method involves foreground and background segmentation, followed by edge feature extraction using enlarged Regions of Interest (ROIs) to enhance measurement changes in images, allowing for reliable remote-distance obstacle avoidance by determining the relative distance between the UAV and targets, even in extreme conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo matching algorithms (local or global) are used for remote-distance obstacle avoidance, then depth information can be obtained from image pixel points, but matching precision deteriorates or matching fails in extreme cases such as no texture, low textures, or dense and repeated textures
Solution Approach 1:
The patent introduces an intermediary mechanism by using foreground-background segmentation and edge feature extraction as intermediate processing steps between image acquisition and depth calculation. Instead of directly performing stereo matching on raw images, the system first segments the image into foreground and background, then extracts edge features from the foreground. This intermediary processing provides more robust features for distance measurement that are not dependent on texture patterns, thereby resolving the contradiction between matching precision and reliability in extreme texture conditions.
Solution Approach 2:
The patent replaces the traditional stereo matching mechanism (which relies on pixel intensity comparisons and epipolar geometry) with an alternative mechanism based on foreground-background segmentation and edge feature analysis. By substituting the core measurement mechanism, the system achieves distance measurement capability that is independent of texture characteristics, thus maintaining both precision and reliability in extreme cases where traditional stereo matching fails.
2Reliability
If foreground and background segmentation with edge feature extraction using enlarged ROIs is used, then obstacle avoidance reliability is improved in extreme cases, but device complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the image processing task into distinct stages: foreground-background segmentation, ROI identification, and edge feature extraction. Each stage processes a specific aspect of the image data, making the overall complex system more manageable and implementable. The segmentation approach allows the system to focus computational resources on relevant regions (foreground objects) rather than processing the entire image, thus balancing reliability improvement with acceptable device complexity.
3Speed
If traditional stereo matching is used, then processing speed is maintained, but measurement precision deteriorates in extreme texture conditions
Solution Approach 1:
The patent implements preliminary action by performing foreground-background segmentation and edge feature extraction before the actual distance measurement calculation. By pre-processing the image data to extract robust edge features from the foreground, the system prepares quality input data for the subsequent distance calculation step. This preliminary processing ensures that when the measurement is performed, high-precision features are already available, thus achieving both speed and precision without requiring complex real-time adjustments during the measurement process.
Data Source
Figure 1~3a
Figure 3b~4
Figure 5~7
AI summary
The present invention discloses a distance measurement method and apparatus, and an unmanned aerial vehicle using same. According to the distance measurement method and apparatus, and the unmanned aerial vehicle provided in embodiments of the present invention, foreground and background segmentation is performed on two neighboring frames of images, and edge feature extraction is performed by using enlarged regions of interest (ROIs) to obtain measurement changes of the images. By means of the measurement changes of the images, remote-distance obstacle avoidance in an extreme condition can be implemented, and problems such as poor stereo matching precision or unavailability of stereo matching in extreme cases such as no texture, low textures, and dense and repeated textures are resolved.