Window washing is a prime candidate for automation: it is repetitious, it requires little emotional intelligence, (particularly regarding multi-story buildings) it is dangerous to human window washers. Finally (and largely due to the danger), it is expensive, and the expense is related primarily to the tasks of getting the human window washer to the location and ensuring their safety. The project of automating window washing involves a number of sub-projects. Project one: figure out where the windows are. The success of this sub-project is essential for the success of the entire project.
In order to solve the problem of detecting windows we plan using a 3 pronged assault. First we plan on taking a picture of a surface and, Using OpenCV, detect all window like shapes on that surface. We will then tell the human user to move to each window in order to use our two sensors to determine if that contour is really a window. We are using Lidar which goes through windows along with an ultrasonic depth sensor, which doesn't, to determine the likelihood of that object being a window.
Plans for the future
text goes here