| abstract |
The invention discloses an infrared cloud image cyclone analysis method, comprising (1) constructing a cyclone system detection network based on an SSD framework of a convolutional neural network, and using the network to automatically identify and locate the cyclone from each cloud image in a cloud image data set; (2) ) adopting the OTSU threshold segmentation algorithm to perform threshold segmentation on the cyclone, and performing area filtering on the result of the threshold segmentation to obtain the initial outline of the cyclone; (3) using the initial outline of the cyclone as an input, using Chan-Vese The model obtains the cyclone boundary of the cyclone; (4) uses the SURF algorithm to extract the cyclone feature points of the cyclone boundary; (5) uses the FLANN matcher to calculate the matching rate of each feature point of the cyclone at adjacent times through the Euclidean distance, This enables cyclone tracking. The invention draws on the MobileNet network and combines with the SSD detection framework to construct a cyclone detection network to identify the cloud image, and the identification accuracy is high and simpler. |