Calibrating Non-overlapping Camera Parameters via 3D Scene Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for calibrating relative parameters of cameras with non-overlapping fields of view require a common calibration object and known scene structures, which limits their applicability and efficiency.
Innovation Solution
A computer-implemented method that reconstructs a three-dimensional scene using target images and determines relative position and attitude information between cameras without the need for a calibration object, utilizing feature point matching and plane-based structure from motion for extrinsic calibration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional calibration methods using common calibration objects are employed, then calibration accuracy is improved, but the method becomes inapplicable to cameras with non-overlapping fields of view
Solution Approach 1:
The patent introduces a coordinate system transformation model as an intermediary to bridge cameras with non-overlapping fields of view. Instead of requiring direct observation of common calibration objects, the method uses coordinate transformations between different camera coordinate systems to achieve calibration, allowing each camera to independently capture images of calibration objects at different positions and times
Solution Approach 2:
The calibration process is segmented into multiple independent stages: each camera independently captures calibration images at different positions, then coordinate transformations are applied to integrate the segmented calibration data from different cameras into a unified calibration result, enabling calibration without simultaneous overlapping field of view
2Reliability
If calibration objects and known scene structures are required, then calibration reliability is improved, but the calibration process complexity and time consumption increase
Solution Approach 1:
The patent employs universal calibration objects (such as planar structures or feature-rich objects) that can be captured by any single camera in the system. These calibration objects serve multiple purposes: they provide feature points for coordinate transformation, enable self-calibration of individual cameras, and facilitate the integration of multiple camera coordinate systems, eliminating the need for complex specialized calibration setups
Solution Approach 2:
The system performs self-calibration by automatically capturing calibration images with the calibration object, extracting feature points, computing coordinate transformations, and determining relative position and attitude parameters without requiring manual intervention or complex pre-configuration of calibration scenes
3Productivity
If simultaneous image collection by multiple cameras is required, then calibration efficiency is improved, but the method cannot be applied to systems with non-overlapping fields of view
Solution Approach 1:
The patent transforms the static requirement of simultaneous image collection into a dynamic process where cameras can capture calibration images at different times and positions. The coordinate system transformation model dynamically integrates these temporally and spatially separated images, allowing flexible calibration schedules that adapt to the operational constraints of non-overlapping field of view camera systems
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The embodiments of the present disclosure provide a method for calibrating relative parameters of a collector, an apparatus for calibrating relative parameters of a collector and a storage medium. The technical solution of the present disclosure may determine target position and attitude information of a target image collector in a calibration coordinate system when each target image is collected, determine the first position information of a spatial point in a three-dimensional scene point cloud in the calibration coordinate system and the second position information a projection point of the spatial point in each target image, and determine a relative position and attitude value between the target image collector and the reference collector based on the target position and attitude information corresponding to each target image, the first position information and the second position information.