Ball Position Transformation with Artificial Intelligence Based on Tensorflow Libraries

Research on wheeled soccer robots has been carried out by several researchers. This is due to the existence of national and international competitions. Previous research was to create a ball position transformation system with a modified method of neural network architecture. This research was developed by building an intelligent transformation system with the Tensorflow library. This transformation system aims to be able to directly measure the distance of objects in real terms without first changing the environmental image from an omni field to a flat plane with conventional camera calibration techniques. This process can replace manual calibration with a variety of field size changes The system can transform with mean error 0.0000026 on epoch 10000 using “conda-tensorflowneural network” libraries. It can transform the position of the ball from the omni space to the cartesian space. This system was implemented on wheeled soccer robot as keeper.