Video image fusion method based on space-time grid

The invention relates to the technical field of video image fusion, in particular to a video image fusion method based on a space-time grid, which comprises the following steps of: performing a network event related video collection technology on videos through a CNN (Convolutional Neural Network) technology, realizing regional alignment and space-time alignment by using space-time grid coding, collecting related videos, performing video segmentation arrangement through the CNN technology, and performing video segmentation arrangement through the CNN technology. Whether input video data of the videos processed by the CNN technology conform to the same event point is judged, a person inputs a more specific time range and event characteristics of the event point into the GAN technology, and then whether the input video data of the videos processed by the CNN technology conform to the same event point is judged; and finally, the fused video is identified and analyzed through a video event point feature identific