Multi-Sensor Calibration Using Connectivity Graphs and Target Variance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for calculating external parameters between multiple sensors are inadequate when their sensing ranges do not overlap or are limited, leading to insufficient conditions for estimation and reduced accuracy due to uneven target distribution.
Innovation Solution
A method that generates a sensor connectivity graph based on the number of shared observations of targets by multiple sensors, calculates variance in target positions, and determines an optimal order for estimating external parameters to ensure accurate relative positional relationships among sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If external calibration is performed using a spherical target observed by multiple sensors simultaneously, then the external parameter can be calculated, but the requirement for shared sensing ranges and simultaneous observations limits the applicability and causes biased estimation accuracy
Solution Approach 1:
The patent segments the calibration process into two distinct phases: (1) constructing a sensor connectivity graph that identifies which sensors can observe the target together, and (2) performing external parameter calculation only for sensor pairs with connectivity. This segmentation removes the requirement for all sensors to have shared sensing ranges, allowing each sensor pair to be calibrated independently based on their own connectivity.
Solution Approach 2:
The patent introduces dynamic target movement during the calibration process. The target is moved to multiple positions, and the system dynamically determines sensor connectivity at each position. This dynamic approach allows sensors that cannot simultaneously observe the target at a single position to still be calibrated by utilizing their connectivity at different target positions, thereby improving adaptability.
2Ease of manufacture
If a predetermined order is used for sensor selection in external calibration, then the process is simple to implement, but the estimation accuracy becomes biased due to uneven target distribution
Solution Approach 1:
The patent performs preliminary construction of a sensor connectivity graph before the actual external parameter calculation. This graph pre-identifies which sensor pairs have connectivity based on target observation capability. By preparing this connectivity information in advance, the system can then perform calculations in an optimized order that ensures uniform target distribution, combining implementation simplicity with improved accuracy.
Solution Approach 2:
The patent uses the sensor connectivity graph as feedback to guide the sensor selection order. The connectivity information feeds back into the calibration process, allowing the system to adjust which sensor pairs are selected for calibration and in what order, thereby avoiding biased estimation that would result from a simple predetermined order.
Data Source
AI summary
A computer is configured to: specify the number of times of a time point when the target is detected at the same time point among time points by each set of two sensors among the three or more sensors by using whether the target is detected by each of the three or more sensors at each time point; generate information indicating connectivity regarding estimation accuracy of a relative positional relationship between sensors by using the number of times specified for each set of sensors and a variance of arrangement positions of the targets at the time points; generate information indicating a first relative positional relationship between two sensors having connectivity by using the information indicating connectivity; generates information indicating a second relative positional relationship between two sensors having no connectivity by using the information indicating connectivity and the information indicating the first relative positional relationship; and outputs the generated information.