Identify granite among other materials and group it by size. The AI model uses spectral analysis based on granite’s composition: quartz (70%), aluminum oxide (14%), potassium oxide (5%), and others.
Overview
Granite AI Analysis identifies granite among other materials and groups the quantity by size. The model reads the material through spectral analysis, using the composition of granite: quartz (70%), aluminum oxide (14%), potassium oxide (5%), and the rest.
The work also covers how that identification is measured. Lidar and a camera feed supply the dimensions, and the pieces are grouped before the data is sent on.
Training material came from quarries, so the model sees more than one source of stone.
The Challenge
Granite has to be separated from other materials, and the amount has to be grouped by size. A visual guess is not enough when the distinction depends on composition.
Dimensions need a live source: a video stream together with Lidar and camera data, without losing the link between what the camera sees and what the Lidar measures.
Our Solution
We built a model that identifies materials by spectral analysis, keyed to granite’s mix of quartz, aluminum oxide, potassium oxide and the remaining compounds.
Software connects to the video stream, takes Lidar and camera data for the dimensions, and a grouping system sends that result on.
Key Features
The model picks granite out from other materials using spectral analysis of its composition.
Identified granite is grouped by size so quantity can be read by how large the pieces are.
Identification follows granite’s mix: quartz (70%), aluminum oxide (14%), potassium oxide (5%), and others.
The Impact
The team built a dataset from various quarries, refined the model and Lidar controller, connected video with Lidar and camera data for dimensions, and added grouping with a system that sends the data on.
Quarries
Dataset from various quarries
Lidar
Model and controller refined
Stream
Video, Lidar and camera data
Size
Model for determining dimensions
Groups
Granite grouping and data sending