Floodplain vegetation in dryland regions is important for maintaining biodiversity, providing habitat and ecological productivity. Environmental water has become an important tool in regulated rivers, particularly in dryland regions, to re-instate river-floodplain connectivity. Monitoring floodplain vegetation provides a means of evaluation against watering objectives, allows adaptive management, and has become a critical part of environmental water management. Detecting, mapping, and analysing the spatial distribution and features of floodplain vegetation, such as percent cover, height, canopy size and openness, provides a direct way to evaluate changes in response to water organisation. This study provides a systematic and integrated spatial framework to classify the floodplain shrub species like Duma florulenta (tangled lignum) and estimate a range of commonly used vegetation condition metrics by integrating vegetation indices (Visible Light Difference Vegetation Index, VDVI), machine learning (k-Nearest Neighbors, kNN) and ecological feature method (Canopy Height Model, CHM) combining drone Red-Green-Blue (RGB) images and LiDAR point datasets. The thresholds of VDVI and CHM were optimized and the kNN method was improved to enhance classification accuracy for the vegetation categories. A total of 18 (50 by 50m) plots at three different sites were established in the Mallee, Australia, in 2024. The results show that the spatial VDVI-kNN-CHM method (90.84%) was better than VDVI-kNN (69.94%) and VDVI (62.78%). The results highlight that the percent cover, and cover of high-quality lignum, average height, and patch area were all related to watering history, with sites managed with environmental water supporting nearly two times greater cover of larger and taller lignum plants. The approach presented novel solutions, consisting of image-based sampling and validation, kNN-based reason and analysis, which can be implemented and extended to other drone-related environmental and ecological investigations.