Structure and land-use mapping (SNK) is a key component of rural development planning but still involves substantial manual effort. This study investigates the use of deep learning for automated SNK based on Bavarian open geodata. Digital orthophotos, normalized surface models, terrain models, and – for the first time – the direct integration of the ATKIS Basis-DLM as an additional input channel are used. An 11-channel model is compared to a 10-channel baseline. Results show consistent accuracy improvements, particularly for shrubs, transportation areas, arable land, and settlement/commercial areas, with less fragmented and more stable segmentations. The study demonstrates the practical benefit of integrating authoritative geodata for administrative mapping tasks.