/Algae

Detection of algal bodies in coral-reef dataset

Primary LanguagePython

Coral Reef Analysis

This is a project undertaken at the Theoretical Ecology & Evolution Lab in the Center for Ecological Sciences, Indian Institute of Science, for the identification of algal colonies in images of coral reefs.

⚠️ Code is buggy ⚠️

Introduction:

  • This project aims to develop a tool to identify different algal species in the in a given coral reef image.

  • Currently, the tool is written in Python, incorporating the following libraries:

    • OpenCV - For analyzing, reading, writing of the image.
    • Pandas - For building image 'data frames' for ease of data manipulation.
    • Numpy - For building custom arrays throughout the program.

    Note: Code is still in it's nascent stages and has not been developed completely. Feel free to reach out if you would like to collaborate or have ideas to fix discrepancies.

Model Overview:

  1. A parser() function is used to read input parameters from the user. The parameters used right now are --file_name and --color_mode.

    • --file_name is used to locate the image which has to be analyzed.
    • --color_mode is used to specify the type of algal colony to be classified in the given sample.

    Note: Currently, the --color_mode parameter is not being used, and by default only red algae are being identified.

  2. The file_name argument is then passed onto the image_reader() which accepts the arguments from the parser() and extracts the image from the disc.

  3. The image is then viewed for the convenience of the user using the image_viewer() function.

  4. The image is then passed onto the pre_processing() functions, which performs the following functions as mentioned:

    • cv2.resize() which is used to resize the image, making it faster to carry out computations. The image is scaled to a quarter of the origin image size, along both the axes.
    • cv2.GaussianBlur() which is used to apply a Gaussian Blur on the image, thereby reducing the noise levels associated with it.

  5. The pandas dataframe is built next. The dimensions are equivalent to the dimensions of the image. Python passing arguments by reference. Hence, manipulation of the elements of the dataframe brings about changes in the channel values of the images.

  6. The pixel information of the image is then transferred into the pandas dataframe using the pixel_extractor() function, where different operations can be run on it to modify pixel values.

  7. We then apply the max_channel_value() function, to cycle through each of the pixel and then express only the channel which possess the maximum value in comparison to the rest of channels belonging to that pixel.

  8. To start out, we detect the presence of red algae in the images provided, hence filter the pixels which contain only the red channel as the maximum channel value. This is done using the red_value_determiner().

  9. In order to threshold and run analysis (such as regression), we transfer the pixels onto another pandas dataframe using the red_value_extractor() function.

Results:

  1. Sample_Image_1

    Fig 1.1 Original Sample_Image_1, post resizing

    Fig 1.2 Contoured Sample_Image_1, post resizing

  2. Sample_Image_2

    Fig 2.1 Original Sample_Image_2, post resizing

    Fig 2.2 Contoured Sample_Image_2, post resizing