The invention discloses a comprehensive monitoring method for a regional
road traffic system in a cross-scale aerial platform and belongs to the field of
aviation monitoring. The method specifically comprises steps that firstly, a semantic segmentation model and an
object detection model are trained based on the
convolutional neural network; images are acquired by an airship in real time in the monitoring area, a video after frame extracting is sent to the semantic segmentation model frame by frame for
processing, and abnormal regions existing in each frame image are manually distinguished andlabeled; secondly, for a current frame image, the labeled adjacent abnormal regions are merged into a whole piece detection region, and a
drone is assigned to the detection area; each
drone goes to adesignated detection area, and the
object detection model is utilized to perform detailed detection of low-altitude monitoring; and lastly, the abnormal position is determined, messages are uploadedand returned back to the airship, and corresponding warning or
processing is performed. The method is advantaged in that through a
cooperative work mode of the airship and the
drone, the monitoring area is monitored roughly and in details, and complete coverage of the monitoring area is realized through relatively small cost.