A Structured Image Processing Operation Library to Automatically Isolate Weeds and Crops

We propose a computer-vision framework that isolate parts of the image pertaining to specific weeds and crops using image processing. We developed a structured Image Processing Operation (IPO) library, implemented in JSON format, for efficient representations of the algorithms needed to isolate the desired parts of the image. For each species, a specialized sequence of IPOs, along with their parameters, are obtained. Emphasis is placed on morphological image processing. The IPOs are mostly standard MATLAB functions, but specific and specialized image processing algorithms can be added as needed. We show preliminary success and discuss various aspects and complications.