Parallel Polarimetric SAR classification based on OpenCL

Abstract

The invention relates to a polarimetric SAR surface feature classification method based on OpenCL parallel, which comprises the following steps of: (1) inputting a polarimetric SAR image to be classified and a polarimetric SAR real surface feature class mark corresponding to the polarimetric SAR image; (2) removing speckle noise; (3) performing feature extraction; (4) generating a training sampleset and a test sample set; (5) pretreatment; (6) training support vector machine model; (7) configuring the OpenCL device end; (8) performing Parallel prediction test sample set landmark; (9) Coloringthe surface features of the test sample set; (10) Outputting the colored classification result map. The invention utilizes the multi-thread parallel processing of the OpenCL device to process the data to be predicted, and changes the original serial processing mode of the support vector machine prediction stage into the parallel processing mode, so as to reduce the time used in the prediction stage without affecting the classification accuracy of the test sample set.

Description

technical field [0001] The present invention belongs to the technical field of image processing, and further relates to a polarimetric synthetic aperture radar SAR (Polarimetric Synthetic Aperture Radar) object classification based on an Open Computing Language OpenCL (Open Computing Language) hardware device parallel processing prediction data in the technical field of image classification method. The invention can be used to extract the features of the polarimetric SAR image and use the features to classify the polarimetric SAR ground objects. Background technique [0002] The ground object classification method of polarimetric SAR image based on support vector machine is a very important classification method. However, due to the long running time of the support vector machine prediction stage, it is difficult to meet the requirements of the actual scene that needs to quickly obtain the classification result within seconds and perform subsequent image processing tasks. ...

AI Technical Summary

Problems solved by technology

This method can improve the accuracy and efficiency of polarimetric SAR image object classification to a certain extent. However, this method still has shortcomings: due to the complexity of the support vector machine model itself, the computational complexity of the prediction stage is high. As a result, the operation efficiency of the support vector machine prediction stage in this method is low and the operation time is long, which is difficult to meet the requirements of quickly obtaining classification results in seconds and performing subsequent image processing tasks in actual scenarios

However, since this method is based on the parallelized support vector machine implemented on the Spark platform, it is only suitable for distributed systems, and its portability is not high, which limits its application range

Images