Grains Recognition

About

The project was implemented during 24h hackaton competition in Brno organised by UnIT Brno on 6-7.4.2018.

Task

Implement an application which is able to load and analyze a picture containing grains in TIFF format. The application is supposed to find all grains and calculate their properties which will be exported in CSV format.

Requirements:

CLI parameters:
- Path to a file to be analyzed (--image-path)
- Path to a CSV file for exporting the calculations (--csv-path)
Each line of the CSV file contains informations about one grain.
Information about grains to be collected:
- Width = it's a width of a bounded box in pixels
- Height = it's a height of a bounded box in pixels
- Max Length = flowline of the most distance between two points on the grain, the flowline can go through outside of the grain - in case of a irregular shape of the grain
- Thickness = perpendicular flowline on the Max Length distance, the flowline mustn't go outside of the grain
Usage of opensource libraries for image recognition, filtration is not allowed. They can be used only for loading an image.
The grains which are located at the borders of the image - only a part of the grain is shown and the rest is outside of the image - can't be processed
The grains which are overlapping are processed as one grain.

Prerequisites

Python 3.6 or higher (but it may work even with older versions of Python 3)
pip3

Install

Install requirements which are listed in the requirements.txt
```
$ ./tools/install_venv.sh
```
The script creates a .venv where all requirements are installed, source it:
```
$ source .venv/bin/activate
```

Run

$ ./main.py [-h] --image-path IMAGE_PATH --csv-path CSV_PATH [--resize RESIZE]

Options

-h, --help => show this help message and exit
--image-path IMAGE_PATH => Path to an image to be processed
--csv-path CSV_PATH => Path where csv file will be stored
--resize RESIZE => Percentage to scale picture down

Process Flow

An image in .TIFF format is loaded
(optional) If --resize argument is present, the loaded image is resized
Blur filter is applied
Threshold is calculated by OTSU algorithm
Thresholding - the threshold is applied on the image
Each grain is wrapped inside a bounded box
Border points are discovered
Lengths and thickness are calculated

Authors

kopecmartin/grains-recognition