Vehicle Sensor Cleaning Control via Image Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In vehicle sensor systems, dirt on incident surfaces affects sensing accuracy, and existing cleaning techniques may inaccurately distinguish between dirt and adjacent objects, leading to unnecessary cleaning during automated driving modes.
Innovation Solution
A cleaning control device that compares outside light images with camera images to extract unmatched pixel groups or target objects with abnormal reflected light intensity changes, instructing the cleaning system to remove dirt from incident surfaces, ensuring accurate cleaning control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing cleaning techniques are used to detect dirt on incident surfaces, then cleaning can be performed, but inaccurate distinction between dirt and adjacent objects leads to unnecessary cleaning operations
Solution Approach 1:
The patent merges the functions of multiple sensors (optical sensor and sensing camera) with overlapping sensing areas to detect dirt. By combining data from both sensors and comparing images, the system achieves more accurate dirt detection than individual sensors alone, resolving the contradiction between cleaning accuracy and avoiding unnecessary cleaning operations.
Solution Approach 2:
The patent introduces an image comparison process as an intermediary between sensor detection and cleaning execution. The extraction unit compares images from multiple sensors to identify unmatched pixel groups, serving as a mediator that filters out false positives and ensures cleaning is only performed when actual dirt is detected.
2Measurement precision
If cleaning is performed frequently to maintain sensing accuracy, then sensing accuracy is improved, but automated driving continuity may be disrupted by erroneous cleaning operations
Solution Approach 1:
The patent implements a feedback mechanism where the cleaning control device continuously monitors sensing data from multiple sensors, compares images to detect dirt, and only triggers cleaning when dirt is confirmed. This feedback loop prevents erroneous cleaning operations while maintaining sensing accuracy, thus preserving automated driving continuity.
Solution Approach 2:
The patent performs preliminary image comparison and dirt verification before initiating cleaning operations. By预先 comparing images from overlapping sensing areas and identifying unmatched pixel groups, the system confirms actual dirt presence before cleaning, preventing disruptions to automated driving continuity.
3Measurement precision
If multiple sensors with overlapping sensing areas are used to improve dirt detection accuracy, then cleaning precision is improved, but device complexity increases
Solution Approach 1:
The patent divides the sensing area into overlapping regions covered by different sensors (optical sensor and sensing camera). Each sensor independently captures images of the same incident surface, and the extraction unit segments the analysis by comparing images to identify unmatched pixel groups corresponding to dirt, achieving high detection accuracy without excessive complexity.
Data Source
AI summary
A cleaning control device is configured to control a cleaning system of a vehicle equipped with (i) an optical sensor configured to acquire an outside light image according to intensity of outside light while light irradiation for sensing reflected light is stopped, (ii) a sensing camera configured to acquire a camera image according to the intensity of the outside light, and (iii) the cleaning system configured to clean an incident surface on which light is incident from sensing areas of the optical sensor and the sensing camera overlapping with each other. The cleaning control device includes an extraction unit configured to extract an unmatched pixel group by comparing the outside light image with the camera image, and a control unit configured to instruct the cleaning system to perform cleaning control to remove dirt from the incident surface, the dirt being estimated to correspond to the unmatched pixel group.