Urban vegetation carbon sequestration capability estimation method based on multi-source and multi-mode technology

The invention discloses an urban vegetation carbon sequestration capability estimation method based on a multi-source and multi-mode technology, and relates to the field of machine learning and remote sensing. Comprising the following steps that firstly, historical remote sensing data, photosynthetically active radiation (PAR) data and road network (OpenStreetMap) data of a research area are directly obtained, and an initial no-label data set is formed; secondly, utilizing an MAE pre-training model and an RNN pre-training model to train the wave band data and the PAR data respectively; thirdly, data of the latest day are input into the trained model, and reconstruction data are obtained; and finally, inputting the reconstruction data into an NPP function model, and completing accurate simulation of the urban vegetation carbon sequestration capability. According to the method, multi-source data with relatively low cost is acquired and combined with a large-scale deep neural network, so that the carbon sequestr