Thermal and Visible Light Camera Image Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing imaging devices face challenges in seamlessly integrating and processing thermal and visible light images for inspection purposes, particularly in aligning and matching pixel densities and fields of view between thermal and visible light cameras.
Innovation Solution
An imaging device with multiple cameras, where one camera captures thermal images and another captures visible light images, processes these images to align pixel densities and fields of view, allowing for the generation of composite images that represent a unified field of view, enabling effective inspection by centering pixel densities and adjusting image data to match the thermal and visible light fields.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a thermal camera and visible light camera are used together for inspection, then inspection capabilities are enhanced, but image alignment and pixel density matching become complex
Solution Approach 1:
The patent divides the image processing task into separate processing pipelines for thermal and visible light images. Each camera's images are processed independently through their respective processing circuits before being combined, allowing complex processing to be handled in manageable segments rather than as a single complex operation.
Solution Approach 2:
The patent introduces an intermediary processing step where images from both cameras are received and aligned to a common coordinate system before final combination. This intermediary alignment process acts as a mediator that resolves the complexity of directly combining misaligned images from different camera types.
2Measurement precision
If the second camera has higher pixel density than the first camera, then image detail is improved, but image processing complexity increases due to density mismatch
Solution Approach 1:
The patent applies different processing operations to different regions and images based on their specific characteristics. The visible light camera with higher pixel density undergoes downscaling processing, while the thermal camera images are processed differently to match the final output resolution, allowing each image type to be processed according to its local quality requirements.
Solution Approach 2:
The patent changes the pixel density parameter of the processed images to achieve uniformity. The visible light images are downscaled from their native high pixel density to match the thermal image density, and the thermal images are processed to achieve a common pixel density, thereby resolving the density mismatch through parameter transformation.
3Manufacturing precision
If the fields of view of two cameras are aligned, then composite image accuracy is improved, but camera positioning constraints increase
Solution Approach 1:
The patent resolves field of view alignment by transitioning from spatial alignment to coordinate system alignment. Instead of requiring precise physical positioning of cameras in space, the system mathematically maps both images to a common coordinate system, adding a computational dimension to solve the alignment problem.
Solution Approach 2:
The patent creates a virtual copy of the scene through coordinate transformation. By mapping both thermal and visible light images to a common coordinate system, the system creates an accurate virtual representation of the inspected object that maintains spatial relationships without requiring the physical cameras to be perfectly aligned.
Data Source
AI summary
An embodiment includes capturing, via a first camera, a first image having a first pixel density, a pixel of the first image corresponding to the first optical axis is substantially centered within the first image; capturing, via the second camera, a second image having a second pixel density that is greater than the first pixel density, where a pixel of the second image that corresponds to the first optical axis is off-center within the second image; processing the second image to generate a fourth image such that a pixel of the fourth image that corresponds to the first optical axis is substantially centered within the fourth image, where a pixel density of the first image is substantially equal to a pixel density of the fourth image, and where the fourth image represents a field of view that is substantially equal to field of view represented by the first image.