dhanendras/computer-vision-project

Water level scale reader

Computer-Vision-Project

Water level scale reader !!

During my internship as a Data scientist within a Startup IT company in Oita Japan. I had the opportunity to conduct a fully-blown computer vision project from scratch while using a plethora of very useful tools available as libraries in Python. To only cite a few, Scipy.ndimage was used for morphology and image measurements modules, PIL for ImageStat module in order to calculate brightness, Sklearn.tree module in order to create a decision tree model that would select the correct parameters for the threshold image binarization, and last but not least, OpenCV was used to perform special shape detection and crucial image pre-processing.

The goal of the project is to be able to read the water level of an irrigation system. The input images were pictures of a graduated scale that has been carefully dipped into the water. The target was to count how many graduations are visible above water level. At first, there were a bunch of obstacles that were hindering image processing, such as overgrown grass, moving and static shadows, tilted scale because of wind blow and water flow, limited visibility because of night darkness etc… but thanks to perseverance and an introduction of machine learning (decision trees) into the operations of image processing, we were able to achieve the wanted results.

(Please refer to the png file uploaded into this repository for some sample results)

NOTE: If you want to copy any snipet of code from this repository, please ask permission first. As the code is not my sole propriety, it still belongs to the company I have conducted this project for. Should you have any inquiry, I'll be happy to help: [email protected]