RGB-D Floor Obstacle Detection for Mobile Machine Blind Spots
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional navigation sensors like lidar, sonar, and infrared sensors struggle to distinguish small on-floor obstacles from the ground, leading to incomplete obstacle detection, as they are typically mounted forwardly and cannot cover obstacles below their height.
Innovation Solution
The use of an RGB-D camera mounted on a mobile machine to capture images with both depth and RGB channel data, allowing for ground plane estimation and foreground detection, enabling the differentiation between obstacles and background through distribution modeling and shape-based methods, thereby covering the entire area where the machine is moving.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional navigation sensors (lidar, sonar, infrared) are mounted forwardly on the mobile machine, then they can detect obstacles in the forward direction, but they cannot detect small on-floor obstacles below their mount height due to the blind spot problem
Solution Approach 1:
The patent transitions from forward-mounted sensors to downward-facing sensors, changing the detection dimension from horizontal to vertical. This allows the sensors to scan the floor surface directly, eliminating the blind spot problem caused by forward mounting height and enabling detection of small obstacles that were previously invisible below the sensor mount level.
2Reliability
If conventional sensors are used to detect obstacles, then they can sense forward obstacles, but they cannot distinguish small obstacles from the ground, reporting all objects including both obstacles and ground
Solution Approach 1:
The patent applies different processing methods to different regions of the sensor data. By analyzing the depth information and comparing it with the estimated ground plane, the system selectively identifies deviations that represent obstacles while filtering out the ground surface. This local differentiation approach enables precise distinction between obstacles and ground, improving measurement precision.
3Adaptability or versatility
If RGB-D sensor is mounted to detect forward ground, then it can cover every portion where the mobile machine is marching to, but it requires an algorithm to distinguish obstacle versus background
Solution Approach 1:
The patent performs preliminary ground plane estimation before obstacle detection. By first establishing what constitutes the ground surface through depth data analysis, the system creates a reference model that simplifies subsequent obstacle identification. This preliminary action reduces the complexity of the overall algorithm by breaking down the obstacle detection task into manageable steps: ground estimation followed by deviation detection.
Data Source
AI summary
On-floor obstacle detection using an RGB-D camera is disclosed. An obstacle on a floor is detected by receiving an image including depth channel data and RGB channel data through the RGB-D camera, estimating a ground plane corresponding to the floor based on the depth channel data, obtaining a foreground of the image corresponding to the ground plane based on the depth channel data, performing a distribution modeling on the foreground of the image based on the RGB channel data to obtain a 2D location of the obstacle, and transforming the 2D location of the obstacle into a 3D location of the obstacle based on the depth channel data.