The capacity to define an object and classify it in the context of its size, shape, texture and neighbours across multiple scales fundamentally changes the image processing paradigm. Classification algorithms are no longer limited to the concept and spatial scale of the ‘pixel’, which is inherently an artefact of the sensor used to generate the image. However identifying ‘real’ or existential objects, as distinct from objects that are an artefact of the image segmentation algorithm, remains a fundamental challenge. This study describes how the analysis of time series coefficients calculated from nationwide 250m MODIS EVI time series data can be used to address this challenge. Twelve statistical and phenological coefficients were used as input into a tailored object oriented image analysis algorithm. Small, detailed image objects were identified using statistical coefficients that contain a ‘hyper-resolution’ effect. These fine scale objects were fused based on climate and seasonality coefficients to form six broad climatic regions. The fine scale image objects within each of these broad regions were classified according to the International Standards Organisation (ISO) Land Cover Classification. Post classification fusion rules were then applied to preserve detailed polygons in the intensive land use regions of Australia, while aggregating up to the vegetation community and landscape unit scale in the more remote and less intensively managed parts of Australia. This approach which is currently being developed at Geoscience Australia, demonstrates the potential of combining tailored object oriented classifications with time series analysis imagery.