Wide-Angle Camera Calibration for Accurate Storage Grid Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems face challenges in accurately determining the location of load handling devices within a grid framework structure, particularly when communication is lost or devices become unresponsive, leading to potential collisions and operational inefficiencies.
Innovation Solution
A method and system for calibrating ultra wide-angle cameras above the grid framework using a neural network to map distorted images to the actual grid coordinates, involving obtaining initial camera parameters, processing images to detect tracks, and updating parameters to minimize discrepancies between mapped and known grid coordinates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a camera is used to capture images of storage media for identifying physical locations, then visual inspection capability is improved, but lens distortion causes inaccurate mapping between image pixels and grid points
Solution Approach 1:
The system performs preliminary calibration by capturing images of a known test pattern (grid of lines or circles) before actual storage media inspection. This pre-capture calibration establishes a distortion correction model that is stored and applied during subsequent operations, allowing the camera to compensate for lens distortion in advance rather than trying to correct it in real-time during inspection
Solution Approach 2:
The calibration process captures images at multiple different focal lengths and zoom levels. By varying these optical parameters during calibration, the system builds a comprehensive distortion correction model that accounts for different operating conditions. This allows accurate pixel-to-grid mapping even when the camera operates at various magnifications throughout the inspection process
2Adaptability or versatility
If calibration images are captured at multiple focal lengths and zoom levels, then mapping accuracy across different magnifications is improved, but the number of calibration images and processing time increase
Solution Approach 1:
The calibration approach captures images at multiple focal lengths and zoom levels during the calibration phase, creating a universal distortion correction model that works across all operating conditions. This single comprehensive calibration process replaces the need for separate calibrations at each magnification level, making the system universally accurate across all zoom levels while performing the multi-parameter calibration only once during setup
3Measurement precision
If distortion correction is applied to achieve accurate pixel-to-grid mapping, then location precision is improved, but computational complexity and processing requirements increase
Solution Approach 1:
The distortion correction parameters are calculated and stored in advance during the calibration phase, before actual storage media inspection begins. During operational use, the pre-calculated correction model is simply applied to each captured image rather than performing complex iterative optimization in real-time. This separates the computationally intensive calibration step from the faster inspection operations
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and system for calibrating a wide-angle or ultra wide-angle camera disposed above a grid of a storage system. An image of a grid section and initial values of a plurality of parameters corresponding to the camera are obtained. The image is processed, using a neural network trained to detect/predict sets of parallel tracks in images of grid sections captured by wide-angle or ultra wide-angle cameras, to generate a model of the sets of parallel tracks as captured in the image. Selected pixels in the model are mapped to corresponding points on the grid using a mapping based on the plurality of parameters, with the initial values used as inputs to the mapping. An error function is determined based on a discrepancy between mapped grid coordinates of the points and known grid coordinates. The initial values of the parameters are updated based on the error function and the updated values are stored for mapping pixels in images of the grid section captured by the camera to corresponding points on the grid of the storage system.