Overview / Usage

This project was made for TATA Steel, India. Their entire steel plant is humongous and critical to the TATA firm as a whole. To ensure its smooth functioning, there are maintenance rounds which last an entire year and barely cover the entire plant altogether. The problem is further aggravated with rooftops of multi-story buildings and plants. To overcome these obstacles, drones were introduced to get an eagle eye's perspective of the various components needing maintenance thus making it more precise and less time-consuming. I was responsible for making the 2D stitched image of all images obtained from the drone along with its GPS coordinates.

Methodology / Approach

For the project, the pipeline was as follows

1. Arrange images according to GPS coordinates in a mesh
2. Extract features of images using conventional feature extractors or CNNs. CNNs were found to give better precision.
3. Mesh the images and monitor cost function for the mesh obtained
4. If the cost function is below a set threshold, merge images after scaling, rotation and translation (as the case might be)

Technologies Used

OpenCV
Tensorflow
Numpy