Polarimetric SAR image classification method based on superpixels and full convolutional network

• Main Authors: GUAN JUNZHI, WANG YAHAN, CHAI XINGHUA, GAO FENG, CHEN YANQIAO
• Format: Patent
• Language: Chinese; English
• Published: 26.06.2020
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a polarimetric SAR image classification method based on superpixels and a full convolutional network, and mainly solves the problem of low classification precision of an existing polarimetric SAR image classification method. The method comprises the following steps: inputting a filtered polarimetric SAR image; extracting a T matrix and H/A/alpha decomposition features as original features; randomly selecting a part of marked samples as a training set, and taking the remaining marked samples as a test set; training a full convolutional network model by using the trainingset, classifying the whole polarimetric SAR image after the model training is completed, and marking a result as Result-SFCN; segmenting the whole polarimetric SAR image by using a superpixel algorithm to obtain a segmentation result; the Result-SFCN is corrected by using the segmentation result of the superpixel, and the corrected classification result is marked as Result-SLIC; according to theinformation entropy, obtai