My name is Michael McDermott. I'm an undergrad studying Mechanical Engineering, and have just completed ten weeks of research under the office of the Stanford Precourt Institute for Energy Research. This summer was spent doing research specifically for the Hot Bucket project. The project itself was focused on creating a reliable and cheap lighting system that could be used in rural India, for the direct purpose of analyzing crop values. Cost reduction, replicability, and robustness were key aspects of the project, so most design decisions employed that as a crucial deciding factor. My research focused on creating the proper lighting environment for the computer vision system, making the target environment evenly illuminated, while also reducing overall power consumption.
The start of my work was heavily involved on analyzing the differences between differing LED, or 'light-emitting-diode', types. It quickly became apparent that surface mounted LED's were the best way forward, due to their incredibly high ratio of brightness intensity, as compared to their power usage, and small package size. In each experiment, the angle of the LED's was varied in order to determine the optimum lighting angle. This variable was important because we wanted to test the effect that different lighting had on the photos. Evenly diffused light was the end goal, so that the effects of dark spots would have negligible effect on the AI when testing photos. A higher angle would lead to less high-angle light, and thus would effectively light the 'undercut' sections of produce, such as with apples, or oranges. This would also lead to a more diffuse light spread, but with each increase in angle, the overall light level would be diminished. Finding an optimal balance here was key.
Later similar brightness tests were conducted to compare power draw at differing light intensity levels. Increasing power consumption to increase brightness brought diminishing returns, and each power gain would lead to less increase in brightness after a point. Determining the optimum current draw then, was also of critical importance, in order to make the most efficient use of our limited power source, while also maintaining an adequate lighting level in the photo environment
The rest of the work was a mix of mechanical and electrical design, getting the system to work effectively, while also keeping power consumption low. The angle blocks needed to be developed such that they would secure the LED circuit boards at the correct angle and be easily modifiable. The small camera we utilized had to be properly secured, and the seams adequately waterproofed. The power from the battery needed to be increased in order to drive the LED's, all while staying power efficient and keeping a small package size.
Overall, there were many things considered in constructing this project, but I believe that the final outcome was worth it. The current prototype should be easy to repair, and should work well for when phot test runs are initialized in the following year. I hope that the work I have achieved leaves room for improvement, but also that it helped to answer many long-standing questions that the researchers had with the system itself. I believe, personally, that I succeeded in doing so, but only time will tell. Overall, the project was incredibly fun to work on, and although the work was tedious at times, I truly hope that my work on the project helps to contribute to a final product that helps farmers to make the most out of their limited resources.