MIE.15 – The Low Cost Vision System

The Low Cost Vision System is used to assist a UR5e with detecting objects laid out on a platform. In this project, the team aims to build a low cost vision system that can determine the orientation and size of an object for the customer: DesignHawk Innovations. The camera would be connected to a controller, and one way to program the vision system is through Python. The end goal of this project was building a low cost vision system from scratch that could be attached to a (robotic arm). This would allow the customer to save funds and use the finances for other projects. The design alternatives were using a PiCam, other coding languages like C++ and platforms like Visual Studio or OpenCV. The team’s methodology was to first take a mass amount of pictures of the blocks laid out on the tray in different positions and orientation. Next is to train the system by using Roboflow to determine the types of objects present. The final step is to implement the training into the program allowing the controller and camera to detect the objects. The final design consists of using a Raspberry Pi to connect and control an Intel RealSense depth camera. This design would be able to identify the object’s depth and size.