A detection method of marine objects in sar data based on otd_loglogistic

Abstract

The invention discloses an ocean target detection method based on OTD_Loglogistic for SAR data, which belongs to the field of ocean target detection. It adopts a detection method combining the initial detection based on the deep learning model and the CFAR fine detection based on Loglogistic, and first constructs the OceanTDA9 lightweight The deep learning model, based on the deep learning model, conducts the initial detection of marine targets, and then uses the CFAR method based on the Loglogistic model to perform fine detection of marine targets and extract the features of marine targets. The present invention combines the deep learning-based initial detection of marine targets with the Loglogistic-based CFAR method, which overcomes the shortcomings of the sliding window CFAR, which needs to count the pixels in the image one by one, which takes a long time to calculate, and cannot process the pixels on the edge of the image, resulting in missing detection at the edge. , which improves the detection efficiency and accuracy of marine targets.

Description

technical field [0001] The invention belongs to the field of ocean target detection, and in particular relates to an OTD_Loglogistic-based SAR data ocean target detection method. Background technique [0002] In recent years, neural networks have been applied to marine target detection, but deep neural networks have the defect of dimensionality disaster, which will reduce the detection speed. Ocean clutter on a calm sea surface can be modeled approximately using a Gaussian model. However, under complex sea conditions such as wind, waves, and tides, the SAR backscatter probability distribution has a long tail and no longer conforms to the Gaussian distribution. The traditional CFAR algorithm is based on the assumption that sea clutter in SAR data obeys a Gaussian distribution, but in complex sea conditions, the distribution of sea clutter is extremely complex, and this assumption is not scientific and universal. Moreover, CFAR needs to count the pixels in the image one by on...

AI Technical Summary

Problems solved by technology

Ocean clutter on a calm sea surface can be approximated using a Gaussian model. However, under complex sea conditions such as wind, waves, and tides, the SAR backscatter probability distribution has a long tail and no longer conforms to the Gaussian distribution.

The traditional CFAR algorithm is based on the assumption that sea clutter in SAR data obeys a Gaussian distribution, but in complex sea conditions, the distribution of sea clutter is extremely complex, and this assumption is not scientific and universal

Moreover, CFAR needs to count the pixels in the image one by one, and the calculation takes a long time. The sliding window cannot process the pixels on the edge of the image, which will cause the defect of missing detection at the edge.

Images