We describe a new method to map intra-plot soil surface heterogeneities at a 5 cm spatial resolution. Our approach unites aerial image classification acquired at very high spatial resolution (VHSR) with local soil sampling. VHSR aerial image processing, based on image classification, allows precise mapping of the spatial distribution of soil surfaces; soil sampling defines soil typology by physical and chemical characteristics. This method has been applied to a plot area located on the hillslopes of Burgundy vineyards (Monthelie, France), where decennial erosion data were already available, in order to assess the effect of soil surface characteristics and slope angle on erosion intensity and localisation. From this method, four classes of radiance were distinguished and interpreted as four soil surface state classes (SSC), defining specific areas within the studied plot. These SSCs have been characterised by their grain-size distribution, their organic carbon, calcium carbonate, and total nitrogen contents. By allowing soil surface states to be mapped at five centimetre resolution, this approach provides novel insights into the characterisation of soil patterns and into erosion analysis on cultivated hillslopes. Our work shows that the spatial distribution of soil erosion is related to the local slope steepness but also to the spatial distribution of stoniness that results from water and tillage erosion processes.