Improved Faster R-CNN Based Ship Detection Method for Port SAR Images

Abstract

The invention discloses an improved Faster R-CNN-based port SAR image offshore ship detection method, comprising the following steps: S1, selecting a SAR image sample set, and constructing a convolutional neural network; S2, inputting a target SAR image Convolutional neural network extracts feature maps; S3, generating candidate area suggestions; S4, target area pooling; S5, performing binary classification: calculate the target type according to the target area feature map obtained in S4, distinguish whether the target is a ship or a background, and form a classification Score; S6, border regression: according to the target area feature map obtained in S4, perform bounding box regression to refine the bounding box, and finally obtain the precise position of the detection frame; S7, eliminate false alarms, and obtain the final area where the target of interest exists. The present invention introduces a deep learning method for synthetic aperture radar target detection, changes the single method of selecting the final detection frame based on the threshold, combines the feature-based method with the pixel-based method, improves the detection rate, and reduces false alarms Rate.

Description

technical field [0001] The invention belongs to the field of radar target detection, in particular to an improved Faster R-CNN-based port SAR image offshore ship detection method. Background technique [0002] Synthetic Aperture Radar (SAR) is a high-resolution microwave imaging radar with all-weather and all-weather working capabilities. It is widely used in marine and hydrological observation, environmental and disaster monitoring, and land and sea tracking and rescue. and civilian areas. The port is an important facility in the military and civilian fields, and port ship detection has important strategic significance for port surveillance and maritime traffic. It is a challenging task in the radar field to detect and identify interested ship targets using SAR images in a port environment. [0003] The key problem of port ship target detection is that the port belongs to the mixed sea and land environment. For ships far away from the port, local threshold segmentation c...

AI Technical Summary

Problems solved by technology

These reasons lead to the contradiction between the detection rate and the false alarm rate of the ship detection system based on the deep learning method

Images