SAR Sea Target Detection via Poisson Distribution Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
SAR images of sea areas suffer from speckle noise, which complicates target detection and requires pre-processing or a priori knowledge of ground conditions, making it difficult to accurately identify targets like ships or oil films without modifying useful signal characteristics.
Innovation Solution
A method that computes a reference Poisson distribution for pixel magnitudes in a SAR image assuming no targets are present, allowing for target detection by comparing this distribution with the actual distribution of selected pixels, without the need for speckle noise reduction or prior knowledge of ground conditions, using a detection threshold based on standard deviation to determine the presence of targets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If speckle noise reduction filtering is applied to SAR images, then image quality is improved, but useful signal characteristics are modified
Solution Approach 1:
The patent applies the Poisson distribution model to the speckle noise itself, treating the harmful noise pattern as a useful statistical signature. By modeling the noise characteristics rather than removing them, the method converts the speckle noise from a harmful factor into a beneficial tool for detection, where the noise statistics provide the basis for identifying target regions through deviation from the expected Poisson distribution.
2Measurement precision
If pre-processing or a priori knowledge of ground conditions is required, then target detection accuracy is improved, but device complexity and processing requirements increase
Solution Approach 1:
The method is self-service in that it uses the inherent statistical properties of the SAR image data itself to perform detection. The Poisson distribution model is fitted directly to the image pixel intensities, and targets are detected based on deviations from this model. No external pre-processing filters, ground truth data, or a priori knowledge of sea conditions are required - the algorithm uses only the image data and its own statistical characteristics to achieve accurate target detection.
Data Source
Figure 1~2
Figure 3~5
Figure 6~7
AI summary
Disclosed herein is a method (13) of detecting a target in a sea area based on a Synthetic Aperture Radar (SAR) image thereof. The Synthetic Aperture Radar (SAR) image is made up of pixels, each having a respective magnitude. The method comprises computing a first reference quantity (block 14) which characterizes a Poisson distribution assumed for the magnitudes that the pixels in the Synthetic Aperture Radar (SAR) image would have if the sea area were free of targets. The method further comprises selecting pixels in the Synthetic Aperture Radar (SAR) image (block 15), computing a real quantity (block 16) which characterizes a real statistical distribution of the magnitudes of the selected pixels, and detecting (block 17) a target in the sea area based on the computed first reference and real quantities. The selected pixels are in a one and the same sub-image of the Synthetic Aperture Radar (SAR) image, and detecting (block 17) comprises detecting a target in a sea subarea of the sea area, the sea subarea being represented by the sub-image.