The present application relates to the technical field of
image processing, and more particularly to a city
terrain mapping method based on
image analysis, comprising: a UAV collects
terrain images and
point cloud data, the
terrain images are subjected to
radiation correction, geometric correction and
feature extraction to obtain multi-dimensional terrain features, the
point cloud data is subjected to cleaning, denoising and
standardization to serve as standardized
point cloud data; a three-dimensional network framework is constructed by combining the two and
TIN,
image texture is optimized to a city terrain three-dimensional model, and
terrain mapping is completed accordingly; the mapping accuracy is determined by the accuracy rate of multi-dimensional terrain
feature extraction, if it is not up to standard, it is determined whether to increase the ground point classification height threshold; if it is not necessary to increase, the
image acquisition integrity is determined by the building shadow coverage rate of the terrain images, if it is not up to standard, it is determined whether to increase the terrain feature texture
roughness coefficient; if it is not necessary to increase, the
TIN modeling triangular net edge number correlation redundancy rate is determined based on the
image frame loss rate. The present application improves the mapping accuracy of city terrain.