Surround View Image Processing for Automated Floor Stain Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current floor stain detection systems using computer vision and image processing are prone to errors due to incorrect sensing distance and coverage area, inability to capture defects in proper dimensions, and require mechanical modifications to existing machines, making them inefficient and unreliable.
Innovation Solution
A system and method utilizing surround view images captured by image capturing devices mounted on a floor cleaning device, generating undistorted virtual top view images, and employing a pre-trained machine learning model to detect floor stains, extract attributes such as dimensions, type, and location, and automate the cleaning process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional camera systems are used for floor stain detection, then the system is simple to implement, but the sensing distance and coverage area are incorrect and defects cannot be captured in proper dimensions
Solution Approach 1:
The patent transitions from conventional 2D camera views to a 3D point cloud representation of the floor surface. By capturing images from multiple cameras and converting them into a point cloud with three-dimensional coordinates, the system achieves accurate depth and distance information, enabling precise measurement of defect dimensions that cannot be obtained with standard 2D imaging systems.
Solution Approach 2:
The patent introduces a point cloud as an intermediary representation between the captured images and the final defect analysis. The point cloud serves as a mediator that preserves spatial relationships and dimensional information, allowing for accurate measurement of defects while maintaining compatibility with standard image processing workflows.
2Extent of automation
If computer vision techniques are used to detect floor stains, then automation is improved, but the system cannot distinguish between defects and floor texture
Solution Approach 1:
The patent applies local quality analysis by examining the three-dimensional characteristics of specific regions in the point cloud. Instead of treating the entire floor surface uniformly, the system analyzes local geometric properties such as height variations, curvature, and point density to distinguish actual defects from normal floor texture patterns, thereby improving detection reliability.
Solution Approach 2:
The patent changes the parameter space from 2D image intensity values to 3D point cloud coordinates and geometric features. By transforming the data representation and analyzing different parameters such as z-height, point spacing, and surface normal vectors, the system can reliably differentiate between defects and texture variations that appear similar in conventional 2D images.
3Extent of automation
If existing machines are retrofitted with stain detection systems, then automation is improved, but mechanical modifications are required
Solution Approach 1:
The patent designs the image capturing system with universal applicability, using standard cameras and processing algorithms that can be adapted to different floor cleaning machine types without requiring custom mechanical modifications. The system processes visual data and point cloud information in a way that is independent of the specific cleaning mechanism, allowing easy integration across various platforms.
4Reliability
If human operators manually detect floor stains, then no additional equipment is needed, but detection is error-prone and stains are left over
Solution Approach 1:
The patent replaces the human visual inspection mechanism with an automated computer vision system that processes images and point cloud data. This substitution eliminates human error and fatigue-related detection failures, providing consistent and reliable stain identification through algorithmic analysis of three-dimensional floor surface characteristics.
Data Source
AI summary
A method for detecting floor stains is disclosed. The method includes capturing images of a floor surface using one or more image capturing devices mounted on exterior top sides of the floor cleaning device body aimed in a forward drive direction. The images correspond to wide-angle view images. The method includes generating undistorted virtual top view image of the floor surface. The undistorted virtual top view image corresponds to a surround view image of the floor surface. The method includes detecting floor stain from undistorted virtual top view image using a first pre-trained machine learning model. The method includes processing floor stain to extract floor stain attribute. The floor stain attribute comprises at least one of: dimensions, type of floor stain, a distance of floor stain from image capturing devices, and a location of floor stain. The method includes cleaning floor stain based on the processing of the floor stain.