Optical Sensor Detector Co-Design for IoT4Ag

Title
Motivation
Project Goals
Specifications
Test Results
Edge Detection
Segmentation Process
Engineering Standards
Video

With the rise of IoT-based precision agriculture, there is a growing demand for drones that can monitor vast tracts of farmland. Our project seeks to build and develop an automated system that can scan and recognize visual data from an array of sensors from a distance. This project is in conjunction with Dr. Kagan’s research in IoT for Precision Agriculture Engineering. Dr. Kagan and her team will advise us in this process as we assist them in their research endeavors. The sensors to be used in the project, which will change color according to changes in the environment it is placed in (pH, temperature, humidity, etc.), will be fabricated by the Kagan Group, while our team will focus on the method of efficiently and remotely detecting these changes.

As an Interdisciplinary Team consisting of mechanical, electrical, and computer engineering students, we aim to build an automated system of optical sensor detection that can capture pictures with incrementally varying distances, angles, speeds, and sampling rates. Initial tests will be run in a controlled setup, perhaps by eliminating variables like ambient light and using a wall with colored paper on a green background in lieu of actual plants growing on farmland. Once this nears completion, it is important to test this system under different lighting conditions and with different proxy sensors for varying brightness, opacity, size, color, or possibly border thickness. Accordingly, it may be beneficial to consider feedback control to adjust brightness, camera zoom, and distance, if the camera already does not possess these, and program them accordingly. This stage will be better defined once the system has gone through some testing. It will also be necessary to test different image recognition and processing algorithms to maximize success in recognizing color changes and mapping out sensor coordinates in terms of X and Y.

Our goal, ultimately, is to build an automated testing system that makes it easier to test different colored optical sensors, distances, and angles easily to ultimately optimize the success in recognizing color changes in these sensors. This consists of a remote-controlled vehicle with a mounted camera, a raspberry-pi interface that programs the vehicle to take a collection of pictures at various increments of distance and angles relative to the proxy sensors, and an image recognition algorithm that discerns the proxy sensors from the background and outputs meaningful data.