Three-dimensional reconstruction method, medium and system for power distribution network unmanned aerial vehicle

The invention discloses a three-dimensional reconstruction method for a power distribution network unmanned aerial vehicle, a medium and a system. The method comprises the following steps: acquiring a video sequence acquired by the unmanned aerial vehicle from a power distribution network environment; generating a cost amount according to the video sequence; predicting a mask of the moving object according to the cost amount; obtaining a depth estimation result according to the cost amount and the mask; and according to the depth estimation result and the unmanned aerial vehicle attitude parameters, inter-frame mapping is carried out and a point cloud is generated so as to realize three-dimensional reconstruction. According to the method, for the power distribution network environment video sequence recorded by the unmanned aerial vehicle, the self-supervised multi-stage training deep learning technology is used, three-dimensional reconstruction of the power distribution network environment is realized, and t