This is a low resolution heat map of occurrences across X/Y axes. They show clear "lanes" going through them, and I'd like to quantify their boarders/middle points. The naive approach would be obviously just slapping some boarders on by hand and calculate the average of all values in between them, but that doesn't feel clean enough: I'd like to do it all mathematically, but I'm not sure which approach is right. I thought about using KNN. Any thoughts or advice?