3D Image Reconstruction Using Adjoint Operator and Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current three-dimensional image reconstruction methods using remote sensing sensors face challenges with high computational requirements and memory consumption, especially when using machine learning algorithms, which limits their applicability due to the need for large simulations and memory-intensive matrix operations.
Innovation Solution
The method employs an adjoint operator to transform data into a one-dimensional vector, reducing computational complexity and enabling efficient processing, and allows for the use of unstructured meshes by representing images in U-, V-, and W-pixel units along coordinate axes, thereby simplifying the image reconstruction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning algorithms are used for three-dimensional image reconstruction, then accuracy and resolution of reconstructed images are improved, but computational time and memory consumption increase significantly
Solution Approach 1:
The patent segments the computational process into two distinct phases: (1) an offline training phase where the neural network is trained using simulated sensor data to learn the mapping between sensor responses and material properties, and (2) an online inference phase where the trained network rapidly reconstructs images from new sensor data. This segmentation allows the computationally intensive training to be performed once, while subsequent reconstructions benefit from fast inference, thus resolving the contradiction between accuracy and computational time.
Solution Approach 2:
The patent performs preliminary actions by pre-training the neural network with extensive simulated data before actual image reconstruction tasks. This preliminary training phase prepares the model to make accurate reconstructions with minimal computational effort during operation, effectively trading upfront computational investment for rapid subsequent processing.
2Measurement precision
If machine learning algorithms are used for three-dimensional image reconstruction, then accuracy and resolution of reconstructed images are improved, but memory consumption increases significantly
Solution Approach 1:
The patent extracts the complex computational burden into a separate offline training phase, removing the heavy memory requirements from the online reconstruction process. By separating training data generation and model training from the actual reconstruction task, the system achieves high accuracy while keeping operational memory consumption manageable.
3Ease of operation
If conventional matrix-based deep learning architectures are used, then image reconstruction can be performed, but computational complexity and memory requirements are excessively high
Solution Approach 1:
The patent substitutes the conventional matrix-based mechanical computational system with a physics-informed neural network that incorporates physical constraints and relationships directly into the network architecture. This substitution replaces heavy matrix operations with a more efficient representation that respects the underlying physics of the imaging problem, reducing computational complexity while maintaining reconstruction capability.
4Ease of manufacture
If a priori constraints are imposed on material property distribution to solve underdetermined systems, then the system becomes solvable, but image accuracy and reliability deteriorate
Solution Approach 1:
The patent changes the parameter representation from conventional matrix formats to a physics-informed neural network parameterization that inherently encodes physical relationships. This parameter transformation allows the system to solve underdetermined inverse problems without relying on arbitrary a priori constraints, thereby maintaining both solvability and image accuracy through physically consistent parameter relationships.
Data Source
AI summary
A method and system for three-dimensional reconstruction of material properties of a target using remotely located physical sensors is disclosed. The special technique disclosed here, enables an order of magnitude improvement in computational speed and memory requirements over current state-of-the-art artificial intelligence-based systems. When compared against state-of-the-art methods that do not use artificial intelligence, the improvement in accuracy and resolution enables deployment of order of magnitude cheaper data acquisition systems and/or provide the practical capability to image targets previously considered out-of-bounds. The use cases include but are not limited to oil field application systems such as the monitoring of pipeline health and integrity, leak, and spill extent delineation, seismic imaging systems, and for applications in agriculture, medical imaging, unexploded ordnance detection, mining, wind energy foundation studies, geotechnical work, groundwater systems, environmental science and engineering, and other problems where remote sensing-based image reconstruction is utilized/needed.