Imaging Sensor Degradation Detection Using Neural Network Sub-Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Imaging sensors in autonomous driving and driver assistance systems face degradation due to environmental factors like dirt, rain, and fog, leading to flawed system outputs in convolutional neural networks and traditional computer vision algorithms, which compromises sensor availability and safety.
Innovation Solution
A method using a fully convolutional neural network to determine imaging degradation by subdividing images into sub-images, training the network to recognize and localize degradation, and providing a control signal for system response, such as cleaning functions, to maintain sensor availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If imaging sensors are used in autonomous driving and driver assistance systems, then the system can capture surroundings and provide driving functionality, but the sensors are subject to degradation from environmental factors (dirt, rain, fog) which compromises sensor availability and safety
Solution Approach 1:
The system performs preliminary detection of imaging degradation by analyzing images captured by the sensor before the degradation critically impacts system functionality. The neural network continuously monitors image quality and detects degradation patterns early, enabling proactive responses such as activating cleaning mechanisms or alerting the driver before the sensor becomes completely unreliable.
2Reliability
If the system shuts down imaging sensors when degradation is detected, then safety is maintained, but sensor availability and system productivity are reduced
Solution Approach 1:
Instead of completely shutting down the imaging sensor when degradation is detected, the system applies partial action by selectively reducing functionality or adjusting operational parameters. The neural network can identify specific degraded regions or conditions and allow the sensor to continue operating at reduced capacity or with modified processing, maintaining partial functionality while ensuring safety. This approach keeps the sensor available rather than completely disabling it.
3Measurement precision
If a neural network is trained to detect degradation in entire images, then detection capability is provided, but the network cannot ensure functionality during degraded operation without special training
Solution Approach 1:
The system segments the image analysis task by training the neural network to detect degradation in sub-images or specific regions rather than requiring the entire image to be processed as a single unit. This segmentation approach allows the network to identify localized degradation patterns and continue functioning with partial image data, improving adaptability to degraded conditions while maintaining detection precision in the visible portions of the image.
Data Source
AI summary
A method for detecting imaging degradation of an imaging sensor includes (i) providing an image of a surrounding area, said image being generated by the imaging sensor; (ii) detecting imaging degradation for each sub-image of a plurality of sub-images of the image using a neural network trained for this purpose; and (iii) detecting the imaging degradation of the sensor, said imaging degradation exhibiting the ratio of the number of sub-images of the image with detected degradation to the plurality of sub-images.