Cleaning Device Dirt Detection via Image Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sweeping robots rely heavily on big data training to identify dirt, which is costly and inefficient due to the difficulty in collecting comprehensive data, leading to potential misidentification and inaccurate cleaning.
Innovation Solution
A control method for a cleaning device that compares pre-cleaning images with post-cleaning images to identify dirt, allowing the device to clean accurately without the need for extensive big data training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If big data training is used to identify dirt, then the cleaning device can learn to recognize dirt patterns, but the cost and complexity of data collection and training increase significantly
Solution Approach 1:
The patent uses image copying and comparison techniques by capturing pre-cleaning images, storing them as reference templates, and comparing real-time images against these templates to identify dirt. This eliminates the need for complex big data training while maintaining high identification accuracy through direct image template matching.
Solution Approach 2:
The patent performs preliminary action by capturing and storing clean surface images before the cleaning process begins. These pre-captured images serve as reference templates that are stored in advance, eliminating the need for subsequent complex training processes and enabling immediate dirt identification through comparison.
2Measurement precision
If comprehensive big data training is performed to improve dirt identification, then accuracy improves, but the time and resources required for data collection and processing increase
Solution Approach 1:
The patent captures and stores reference images of clean surfaces in advance before the cleaning task begins. This preliminary action creates ready-to-use templates that enable immediate dirt identification through comparison, eliminating the time-consuming big data training process while maintaining high accuracy.
Solution Approach 2:
The patent uses simple image copying and storage of pre-cleaning reference images instead of complex big data training. The copied reference images are stored and directly used for comparison, providing rapid dirt identification without the extensive time investment required for comprehensive data collection and training.
3Productivity
If image comparison with stored templates is used to identify dirt, then the cleaning device can accurately detect dirt in real-time, but the need for storing and processing reference images increases memory requirements
Solution Approach 1:
The patent applies local quality by storing and comparing only the specific regions of interest (pre-cleaning images of clean surfaces) rather than entire images or comprehensive datasets. This approach reduces memory requirements while maintaining real-time detection capability by focusing computational resources on relevant local areas.
Solution Approach 2:
The patent uses simple copying of pre-cleaning reference images and stores them for comparison. This copying approach requires minimal memory storage compared to big data training models, enabling real-time dirt detection with reduced memory requirements by only storing essential reference templates.
Data Source
AI summary
Disclosed are a control method for a cleaning device and the cleaning device. The cleaning device includes a cleaning component, a processor, a memory and at least one camera. The control method includes: obtaining, via a processor, a contrast image of a position to be cleaned, and storing the contrast image in the memory; capturing, via a camera, an image of the position to be cleaned as a first image; comparing, via the processor, the first image with the contrast image to obtain a dirt mark; and controlling, via the processor, the cleaning component to clean dirt at the dirt mark.