Neural Network Illumination Estimation Without Geometric Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing simulation software for indoor illumination requires geometric/vectorial models, making it impractical for high-automation contexts or real-time control of ambient illumination.
Innovation Solution
A neural network is trained to predict illuminance values for a cloud of three-dimensional points representing an ambient with light sources, using ambient and emission data to estimate illumination intensity without requiring geometric/vectorial models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simulation software uses geometric/vectorial models to predict light quantity, then measurement precision of illumination is improved, but device complexity and time consumption increase significantly
Solution Approach 1:
The system performs preliminary action by pre-calculating and storing illumination characteristics in a lookup table during an offline phase. The precomputed data includes illuminance values for various light source positions and ambient configurations. During real-time operation, the system simply queries this precomputed table rather than performing full simulations, thus achieving fast illumination estimation without sacrificing accuracy.
Solution Approach 2:
The system creates a simplified point-cloud representation of the ambient environment as a copy of the full geometric model. Instead of processing complex vectorial models in real-time, the system works with this simplified point-cloud copy that contains essential spatial information. The lookup table stores illumination data corresponding to this point-cloud representation, enabling rapid querying and estimation without handling the full geometric complexity.
2Manufacturing precision
If simulation software requires geometric/vectorial models, then manufacturing precision of illumination simulation is improved, but ease of manufacture deteriorates due to need for CAD operators
Solution Approach 1:
The system applies self-service by automatically generating the point-cloud representation and performing illumination estimation without requiring external CAD operators. The point-cloud can be directly obtained from laser scanning or photogrammetry surveys, and the system autonomously processes this data to query the lookup table and generate illumination predictions, eliminating the need for manual geometric model creation and professional operator intervention.
Solution Approach 2:
The system uses a simplified point-cloud copy of the ambient environment instead of requiring full geometric/vectorial models. This point-cloud representation contains sufficient spatial information for illumination estimation but is much simpler to create and process. The lookup table is built based on this point-cloud representation, enabling the system to achieve accurate illumination simulation without complex model creation processes.
3Productivity
If automated high-quantity simulations are performed, then productivity of illumination analysis is improved, but loss of time increases due to computational complexity
Solution Approach 1:
The system performs preliminary action by pre-computing illumination characteristics for a comprehensive set of conditions and storing them in a lookup table during an offline phase. This allows the system to handle multiple simulation queries in real-time by simply looking up precomputed values rather than performing full calculations, thus achieving high productivity for automated batch processing without proportional increases in computational time.
Solution Approach 2:
The system extracts only the essential spatial and illumination characteristics needed for prediction, storing them in a compact lookup table format. Instead of processing complete geometric models for each simulation query, the system extracts and utilizes only the relevant precomputed illumination data corresponding to the query parameters, significantly reducing computational overhead while maintaining simulation throughput.
Data Source
AI summary
A method for estimating at least one illumination value of an ambient includes a) an acquisition phase, wherein ambient data which define a plurality of points detected in the ambient, emission data defining a first set of the points that represent light sources, and illuminance data defining a second set of the points that represent the illuminated ambient are read, b) a training phase, wherein a neural network is trained by inputting the ambient data and the emission data, forcing the output of the illuminance data, c) a determination phase, wherein second illuminance data are determined, by means of the neural network, on the basis of the ambient data and second emission data.