Unmonitored or partially supervised learning of a classifier for the environment detection of a vehicle

摘要

The invention relates to a computer-implemented method for learning a classifier for objects in the vicinity of a vehicle for a system for assisted or autonomous driving. The method includes determining patterns for learning the classifier from a time series of environmental data. Objects are also tracked in the environmental data. Based on this, patterns that can be assigned to the same object are associated. In one embodiment, first classifications of patterns are determined with an initial classifier. Based on these first classifications, the classifier can be learned iteratively, taking into account the associations between the patterns.