Stereo Camera Sensor Fault Detection via Reference Depth Map Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting image sensor errors in stereo cameras are inadequate, particularly in static scenarios and for ensuring safety standards, as they fail to reliably distinguish between noise and actual defects, leading to potential misclassification of distances and missed object detection.
Innovation Solution
A method that generates a reference depth map during a learning phase and compares it with subsequent depth maps to detect systematic deviations, using statistical measures like mean values to identify address decoder errors and other issues, thereby ensuring the functionality of image sensors and preventing safety-critical miscalculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current noise figure comparison methods are used to verify image sensor functionality, then the test can be performed, but the detection reliability is insufficient because defective pixels can exhibit noise corresponding to expectations
Solution Approach 1:
The patent applies preliminary action by capturing a reference depth map during a learning phase before actual operation. This reference map is stored and later compared with depth maps captured during operation to detect systematic deviations indicating image sensor errors, thereby improving detection reliability beyond simple noise figure comparison.
Solution Approach 2:
The patent implements feedback by continuously comparing captured depth maps with the stored reference depth map and using statistical measures to identify systematic deviations. This feedback mechanism enables reliable detection of image sensor errors by analyzing differences between current and reference measurements rather than relying on absolute noise figures.
2Reliability
If verification circuits are integrated into the image sensor as specified in US Patent 8,953,047 B2, then image sensor functionality can be verified, but the image sensor becomes more complex to manufacture and additional circuit components take up space
Solution Approach 1:
The patent uses an intermediary approach by introducing a reference depth map as a mediator between the image sensor and the verification process. Instead of integrating complex verification circuits into the sensor itself, the system captures reference depth maps during a learning phase and compares them with operational depth maps, thereby verifying sensor functionality through external comparison rather than internal circuitry.
Solution Approach 2:
The patent applies copying by creating a reference copy of the depth map during the learning phase when the image sensor is known to be functioning correctly. This reference copy is then used for comparison with subsequent depth maps to detect errors, eliminating the need for integrated verification circuits while maintaining reliable functionality checking.
3Reliability
If statistical measures like mean values are used to compare depth maps, then systematic deviations can be detected, but additional processing effort is required
Solution Approach 1:
The patent applies partial action by calculating statistical measures such as mean values only for specific pixel ranges or regions of interest in the depth maps, rather than processing all pixels. This selective approach enables detection of systematic deviations while reducing the overall processing effort compared to full-map statistical analysis.
4Loss of information
If the stereo camera is configured to detect objects and calculate distances, then depth information is provided, but image sensor errors can lead to incorrect distance calculations and missed object detection
Solution Approach 1:
The patent applies preliminary action by capturing and storing a reference depth map during a learning phase before actual operation begins. This reference map represents the correct depth information under normal conditions and is used later to detect systematic deviations in distance calculations, thereby maintaining accuracy even when image sensor errors occur during operation.
Solution Approach 2:
The patent implements feedback by continuously comparing the depth maps captured during operation with the stored reference depth map. Statistical analysis of the differences provides feedback on whether systematic deviations are present, enabling detection and correction of image sensor errors that would otherwise lead to incorrect distance calculations and missed object detection.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A safe stereo camera (10) for monitoring a monitoring area (12) is specified, comprising at least a first image sensor (16a) for capturing a first two-dimensional image and a second image sensor (16b) for capturing a second two-dimensional image of the monitoring area (12) from a perspective offset from each other, a stereoscopic unit (26) for generating a depth map from two-dimensional images according to the stereoscopic principle, an evaluation unit (24) for processing depth maps, and a test unit (28) for checking the functionality of the image sensors (16a-b), wherein the evaluation unit (24) is configured to generate a reference depth map of the monitoring area (12) in advance during a learning phase and to form a difference depth map from a captured depth map and the reference depth map during operation.The test unit (28) is designed to form an average of pixels of the difference depth map and to detect a fault in an image sensor (16a-b) based on a deviation of the average from a reference value.