Binocular Vision Obstacle Recognition Using Depth-Height Map Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current binocular vision obstacle avoidance methods for robots suffer from low accuracy due to environmental interference, leading to incorrect obstacle recognition and ineffective avoidance processes.
Innovation Solution
An intelligent obstacle avoidance method and apparatus that combines depth maps and height maps from binocular cameras to identify target obstacles by calculating overlap ratios between initial obstacles in both maps, ensuring accurate recognition and effective avoidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If binocular vision obstacle avoidance method is used, then the robot can avoid obstacles in changing environments, but the accuracy of obstacle recognition is low due to environmental interference
Solution Approach 1:
The patent combines depth map information and height map information from binocular vision to comprehensively identify obstacles. By merging multiple information dimensions (depth and height), the system overcomes the limitations of single-map recognition and reduces environmental interference, thereby improving both recognition accuracy and avoidance reliability
Solution Approach 2:
The patent introduces an overlap ratio calculation as an intermediary mechanism to verify obstacle validity. By calculating the overlap ratio between depth map obstacles and height map obstacles, the system filters out false positives caused by environmental interference, improving recognition accuracy while maintaining avoidance effectiveness
2Device complexity
If depth map alone is used for obstacle identification, then the processing is simple, but wrong obstacles are recognized resulting in low accuracy
Solution Approach 1:
The patent merges depth map processing with height map processing, where the height map is derived from the depth map. This combination adds verification capability without requiring entirely separate sensing systems, balancing complexity improvement with accuracy enhancement
Solution Approach 2:
The patent transforms the problem from two-dimensional depth map analysis to three-dimensional spatial analysis by calculating height information from depth data. This dimensional transformation enables the system to filter obstacles based on height criteria, improving accuracy while using the same binocular vision hardware
Data Source
AI summary
Provided are an intelligent obstacle avoidance method and apparatus based on binocular vision, a robot and a medium. The method includes acquiring, through a binocular camera, a depth map of a work scene in a moving direction of a robot, and calculating a height map corresponding to the work scene according to the depth map; determining first obstacle information in the work scene according to the depth map and determining second obstacle information in the work scene according to the height map; determining a target obstacle existing in the work scene and position information of the target obstacle according to the first obstacle information and the second obstacle information; and controlling the robot to perform obstacle avoidance according to the position information and a preset path of the robot.