This document details Datarock’s product Dominant Colour, which generates a single colour output that allows for a simple colour representation of each row image and is a particularly useful output that most 3D modelling packages can handle.
Contents
| Table of Contents | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Literature
This product is based on the following literature:
...
Title
...
Author
...
Year
Background
To identify a colour that best represents an image, we need to understand that colour can be numerically represented in many ways (https://en.wikipedia.org/wiki/Color_model ), called colour models. When an image, stored as a jpg or png or other file format, is loaded into a computer, the colour of each pixel can be numerically represented by passing saved information to any of these colour models.
One of the most intuitive colour models to think about is the RGB model (https://en.wikipedia.org/wiki/RGB_color_model ), where the red, green and blue primary colours of light are added together to form a distinct colour. Using this model, the colour of each pixel in our loaded image is therefore defined by an RGB triplet (R,G,B) of numerical information.
The purpose of the Datarock Dominant Colour product is to condense the varied colours represented in the core photograph to consistently determine the dominant colour of the core.
Dependent Models
The imagery only needs to have passed through Depth Registration in order to process Dominant Colour.
Data Processing
The following steps are taken to determine the Dominant Colour of a row image.
Take the output of the fracture detection model, “simple” classes only
Run a segmentation model to extract the fracture profile (or skeleton)
Determine its angle and curvature
For a given 3m interval, plot each measurement of angle and curvature on the plot shown below
Create 45° bins for fracture angles, slide left/right to minimise number of bins
If there are three or more fractures in a bin, a set is assigned. Other fractures are designated random.
Combine the measured values of each fracture with “complex” fracture detections as per the flowchart below.
...
Number of Sets
...
Random Joints Present?
...
Description
...
Jn
...
0
...
No
...
No joints
...
0.5
...
0
...
Yes
...
Random joints
...
1.0
...
1
...
No
...
One set
...
2.0
...
1
...
Yes
...
One set + random
...
3.0
...
2
...
No
...
Two sets
...
4.0
...
2
...
Yes
...
Two sets + random
...
6.0
...
3
...
No
...
Three sets
...
9.0
...
3
...
Yes
...
Three sets + random
...
12.0
...
4
...
No
...
>4 sets - heavily jointed
...
15.0
...
5
...
No
...
Crushed rock
...
20.0
Product Configuration Options
...
Configuration
...
Options
...
t_1 threshold (see flowchart above)
...
Threshold can be set to any value.
Default value is 30%
...
t_2 threshold (see flowchart above)
...
Threshold can be set to any value.
Default value is 20%
...
t_3 threshold (see flowchart above)
...
Threshold can be set to any value.
Default value is 10%
...
t_4 threshold (see flowchart above)
...
Threshold can be set to any value.
Default value is 0%
Output Intervals
Default interval length: 3.0m
Customisable interval available: No
User Data
User data may be uploaded to the platform via csv in the following format:
· HoleID_sampling_intervals_joint_sets.csv
CSV file to contain the following headers:
...
File Header
...
Description
...
depth_from
...
Start of interval
...
depth_to
...
End of interval
...
jn
...
Joint Set Number logged on site. This is an optional field.
Taking the depth registered row image (see Analytics Ready Data (Image Preparation + Depth Registration) for more information).
Extract every pixel from the image, and plot each RGB triplet onto a 3 axis plot.
Cluster all the RGB values into five clusters.
Determine the cluster with the highest number of pixels. This is the dominant cluster.
Find the geometric centre of the dominant cluster. The R, G and B values of this point is our Dominant Colour.
Convert the RGB value to hex and export both .
Example output
The images below show an example of the extracted row images. Key points:
Hole is 1,116m deep
Raw images were 380 megabytes
Processed data (csv format) is ~100KB
This results in a significantly compressed representation of the core images that can be used to plot in 3D modelling packages or further data analysis.
Depth registered row images | Dominant colour row images |
|---|---|
 |  |
Product Configuration Options
There are no configuration aspects to this product.
Output Intervals
Data intervals are based on the calculated depth of each row image.
User Data
Not currently able to upload to the platform.
Data Output
Results from this product is delivered in a batch nature.
...
The available CSV files include the following:
ProjectID_Joint_Set_Analysis_3mDominant_Colour.csv
File Header | Description | |
hole_id | Customer’s Hole ID | |
depth_from_m | Start of interval (metres) | |
depth_to_m | End of interval (metres) | |
jn | Joint Set Number as defined in the Q-System | |
jn_description | Joint Set Number description as defined in the Q-System | |
timestamp | Time of joint set calculations | |
version | A model version identifierR | Red component of the dominant colour |
G | Green component of the dominant colour | |
B | Blue component of the dominant colour | |
hex | Hex colour code for the dominant colour |
Product Limitations
Limitations | Comments |
Measurements are taken on simple, single interface fractures only.This method is based on measurements of simple, single interface fractures, and assumed parameters for broken zones. It is recommended that these assumptions be verified based on customer’s logging schemathe full row image | The full row image is ingested and to determine the dominant colour. Is most cases this is more than suitable for determining the dominant colour of the core. However, in some row imagery, non-rock areas in the image such as empty track or core block can dominate the image, resulting in the dominant colour representing these non-rock areas. |
Document Version
Version | Date | Author | Rationale | ||||
1 | 6 Sep 2022Dec 2023 | S Johnson | Initial release | ||||
2 | 9 Mar 29 Jan 2023 | S Johnson | Updated with Jn table | 3 | 18 July 2023 | S Johnson | Updated for custom intervals units for outputs |