Camera-based Plough Control

Duration of the Thesis: 6 months
Completion: July 2017
Supervisor: Dr. Gabriele Michalke (Robert Bosch GmbH), Dr. Michael Erz (Robert Bosch GmbH)
Supervisor & Examiner: Prof. Dr.-Ing. Norbert Haala

Nowadays, the tendency is to develop automated technologies, where the humans have little to no influence over the running processes. This is also valid in the agricultural sector, where the goal is to improve the efficiency and to individualize the actions based on the environment.  In the case of a tractor, for different tasks like ploughing, sowing or spraying various sensors are used for the equipment setup. However, they cannot provide enough information about the surroundings for a full automation. A better way to capture the environment, so that it can be understood and help personalize the actions, is by means of cameras.

Two cameras were used, in order to be able to recover the depth information. The implemented algorithms were based on grayscale segmentation and further, the identification and classification of objects in the image was performed in order to derive relevant information for the driver and the controlling unit in the tractor.

In order to evaluate the implemented algorithms, the previously results were compared to the data from a Lidar sensor. After analyzing the methods from the point of view of the accuracy, the root-mean-squared error was in cmrange.