Deep Learning Pressure Map Prediction from 2D Photos
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pressure mapping technologies require physical contact between surfaces, making it difficult to obtain pressure maps for objects that cannot be brought into proximity, such as in medical devices or prosthetics for patients with mobility issues, and rely on expensive hardware like sensors and scanners.
Innovation Solution
A system that uses deep learning to generate pressure maps from 2D images and object parameters, eliminating the need for physical proximity and hardware by constructing a 3D model through photogrammetry and keypoint Deep Learning Networks, allowing for the prediction of pressure maps from series of 2D photographs taken with devices like smartphones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical contact pressure mapping systems are used, then measurement precision is improved, but device complexity and cost increase due to expensive sensors and scanners
Solution Approach 1:
The patent uses 2D photographs as optical copies of the object surface to create a digital representation, eliminating the need for physical contact with expensive sensors. The photogrammetry process reconstructs 3D geometry from these 2D images, and the deep learning network predicts pressure distributions from this digital model, achieving pressure mapping without physical pressure sensors
Solution Approach 2:
The patent replaces the mechanical pressure sensor system with an optical-digital system. Instead of using physical sensors that require mechanical contact, the system uses 2D imaging (optical) followed by photogrammetry and deep learning (digital processing) to predict pressure distributions, substituting a mechanical measurement system with a computational one
2Measurement precision
If physical contact pressure mapping is performed, then pressure distribution data is obtained, but ease of operation deteriorates when objects cannot be brought into proximity
Solution Approach 1:
The system creates a digital copy of the object through photogrammetry from 2D images, allowing pressure mapping to be performed on this virtual replica. This eliminates the need for the user to physically bring the object close to expensive scanning equipment, as the 2D photographs can be taken from a distance and processed digitally
Solution Approach 2:
The patent introduces a deep learning network as an intermediary that translates 2D visual information into pressure distribution predictions. This intermediary model, trained on pressure map data, allows the system to infer pressure information without direct physical measurement, bridging the gap between optical imaging and pressure measurement
3Measurement precision
If traditional pressure mapping systems are used, then pressure measurements are obtained, but loss of time increases due to requirement for physical proximity and travel
Solution Approach 1:
By using 2D photographs as input, the system creates a digital surrogate of the object that can be analyzed without requiring the user to travel to specialized facilities. The photogrammetry-reconstructed 3D model and subsequent deep learning analysis can be performed remotely and quickly, eliminating travel time while maintaining measurement capability
Solution Approach 2:
The deep learning network is pre-trained on pressure map data during an offline training phase. When a new object is measured, the already-trained model can immediately predict pressure distributions from 2D images without requiring time-consuming calibration or setup, enabling rapid measurement
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables the generation of accurate pressure maps without physical contact or specialized hardware, facilitating the creation of customized products and reducing the need for users to travel for measurements, improving accessibility and reducing costs.
Implementation Method 1
The images and object parameters are processed by a three-dimensional (3D) model generation module that uses photogrammetry
Data Source
AI summary
A structured 3D model of a real-world object is generated from a series of 2D photographs of the object, using photogrammetry, a keypoint detection deep learning network (DLN), and retopology. In addition, object parameters of the object are received. A pressure map of the object is then generated by a pressure estimation DLN based on the structured 3D model and the object parameters. The pressure estimation DLN was trained on structured 3D models, object parameters, and pressure maps of a plurality of objects belonging to a given object category. The pressure map of the real-world object can be used in downstream processes, such as custom manufacturing.