Our team has received numerous well-meaning recommendations to use a combination of red and green or blue plus green to improve flag detection (reducing false positives created by other similarly colored objects in the background). This seems very logical and I’m sure it works under in a small range of conditions. Here are the following reasons we have found it unusable for competition. One is that the vision sensor only holds a limited number of detected objects which are prioritized by size. When the sensor has all flags plus other background object in its field of view, smaller (green) objects may not make the cut. Also, at some flag angles\distance the green stripes do not represent enough pixels to produce a detection. In addition, the green color produces the largest variation in color depending on lighting as well as being very glare sensitive. All of this means that if you are able produce a consistent and accurate red/blue detection (not always easy) but then also require it to also “see” the green signature, flags become undetectable in many situations. We also tried combining different similar signatures like when middle or high flags are very different in lighting. However, this appear to reduce the effectiveness of either signature when used separately. My theory is that the detection algorithms were not designed to be used in this manner. One process that does improve consistency is to take multiple samples. The downside to this approach is that is slows the process by the number of samples taken. It also becomes difficult to reconcile which readings are from the same object across samples, since in many instances the number of signatures detected changes.