Method for establishing semantic distance map and related moving device
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sweeping robots rely on single lens cameras for distance measurement, leading to ineffective obstacle avoidance, as they cannot determine the height of obstacles and may collide with or get stuck under them during sweeping.
Innovation Solution
A method and device that create a semantic distance map using a monocular image capturing device and a non-contact range finding module, which captures images, performs obstacle recognition, and determines the distance, type, and recognition probability of obstacles, enabling precise obstacle avoidance by calculating the semantic distance map.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single lens camera is used for distance measurement, then the device complexity is reduced, but the measurement precision of obstacle distance and height is insufficient
Solution Approach 1:
The patent combines a monocular camera and a non-contact range finding module (such as a time-of-flight sensor) into an integrated obstacle detection system. The camera captures visual information while the range finding module provides accurate depth data, and their results are fused to create a comprehensive semantic distance map that includes both visual recognition and precise distance measurement.
Solution Approach 2:
The patent introduces a processing unit as an intermediary that receives data from both the camera and the non-contact range finding module. This processing unit fuses the visual information with depth information to generate a semantic distance map, acting as a mediator that combines multiple data sources to achieve superior measurement precision.
2Device complexity
If obstacle avoidance is performed based on single-point distance measurement, then the device complexity is reduced, but the reliability of obstacle avoidance is insufficient
Solution Approach 1:
The patent transitions from single-point distance measurement to a two-dimensional semantic distance map that covers the entire visual field. Each pixel in the map contains distance information, transforming the measurement from a single dimension (one point) to two dimensions (spatial distribution across the image), thereby improving the reliability of obstacle avoidance.
Solution Approach 2:
The patent performs preliminary obstacle recognition and distance mapping before the sweeping robot approaches the obstacle. By pre-establishing the semantic distance map and identifying potential obstacles with their distances and types, the system can plan avoidance paths in advance, improving the reliability of obstacle avoidance.
3Speed
If the robot directly avoids obstacles without determining height, then the speed of obstacle avoidance is improved, but the productivity of sweeping is reduced
Solution Approach 1:
The patent implements dynamic obstacle avoidance behavior based on the semantic distance map. When an obstacle is detected, the system determines its height and distance, then dynamically adjusts the avoidance strategy: for low obstacles that allow passage, the robot continues sweeping; for high obstacles, the robot performs avoidance. This dynamic response maintains sweeping productivity while ensuring safe avoidance.
Data Source
AI summary
An establishing method of semantic distance map for a moving device, includes capturing an image; obtaining a single-point distance measurement result of the image; performing recognition for the image to obtain a recognition result of each obstacle in the image; and determining a semantic distance map corresponding to the image according to the image, the single-point distance measurement result and the recognition result of each obstacle of in the image; wherein each pixel of the semantic distance map includes an obstacle information, which includes a distance between the moving device and an obstacle, a type of the obstacle, and a recognition probability of the obstacle.