Freespace detection in a driver assistance system of a motor vehicle with a neural network

A method for operating a driver assistance system of a motor vehicle, comprising:

• a) capturing an environment of the motor vehicle by a capturing device of the driver assistance system;
• b) detecting an accessible freespace in the captured environment by a computing device of the driver assistance system;
• c) detecting and classifying a plurality of objects in the captured environment that are located at a border of the freespace by a neural network of the driver assistance system to produce a plurality of classified objects;
• d) assigning a part of the border of the freespace to a classified object among the plurality of classified objects;
• and
• e) categorizing a part of the freespace adjacent to the part of the border that is assigned to the classified object in dependence upon the class of the classified object,
• wherein a respective risk is assigned to the categories for the freespace that the part of freespace is categorized into a risk for a collision with the classified object adjacent to the respective part of freespace,
• wherein in the classifying of the at least one object according to method step c), more than one class is assigned to the classified object and a confidence value that indicates the probability for the respective class assigned to the classified object to be true is assigned to each class assigned to the classified object,
• wherein in the categorizing of the part of the freespace according to method step e), the confidence values for the classes of the classified object are compared,
• wherein the freespace is categorized into the category for the freespace that corresponds to the class with the highest confidence value, when this confidence value differs from the second highest confidence value of the classes assigned to the classified object more than a preset threshold, and
• wherein the freespace is categorized into the category that the highest respective risk is assigned to, when the highest confidence value differs from the second highest confidence value less than the preset threshold.