Detekce překážek s využitím hloubkových obrazů

Abstract

Due to the constant accidents caused in parking lots, on the roads, when moving the high-lift trucks in warehouse or controlling robot which maps surroundings and we don’t want to damage this robot, the parking assistance system helping the driver detect obstacles is necessary. In this thesis, we designed the system that finds and displays obstacles in front or behind the car to the driver from sequence of images from camera with depth image (Orbecc Astra). The principle of solving the problem is the thresholding of the input image, finding obstacles by changing the brightness in the depth image and, in the end, making obstacles more visible to warn driver about a possible collision. The application uses the OpenCV library from which we used some features to detect obstacles. During the design we wanted to create a system which could be used in real time to avoid the obstacles standing in the way.