Mining and Metallurgy: Assignment Metallurgy

You need to assess the viability of an ore classification algorithm to be used in a new automatic ore sorting system for your company. The system will make use of computer vision that would derive textural descriptors from the ore surfaces to use as predictor variables for a classification model.

In your preliminary proof-of-concept study, you need to

a) Generate synthetic images of two types of ores

i. Type A should have a random assignment of pixel intensity values between 0 and 255 in 100 rows x 100 columns

ii. Type B should have pixel intensity values in 100 rows x 100 columns with correlated column values, ranging between 0 and 255 as well, i.e. every alternate column should have random pixel values between 0 and 255 and every column in between should have pixel values equal to the square root of the preceding column's simulated pixel values.

b) Use the simulated pixel intensities in the images to construct measurement vectors or variables.

c) Extract ore descriptors or features from the variables.

d) Design a model that can classify the ores correctly from images of the ore.

Your investigation should be presented in a summary report, which should contain at least

• a representative image of each of the two ores
• example of a measurement vector of each of the two ores
• a score plot of the features of the ores
• a brief description of the feature extraction method:
• a brief description of the classification model