A Fast Target Extraction Method for SAR Azimuth Estimation

Abstract

The invention discloses a fast target extraction method for SAR azimuth angle estimation, comprising the following steps: S1, determining a seed threshold and a growth threshold; S2, taking a pixel point whose gray value is greater than the seed threshold as a target seed point; S3, Add the pixels whose gray value around the seed point is greater than the growth threshold as a new target seed point to the target linked list; S4, use the pixels around the seed point that meet the set neighborhood conditions as new target seed points and add them to Target linked list; S5. Output a binary matrix indicating whether the image pixel is the target point as the final target image. The present invention does not need to assume the distribution of the target and the noise, and overcomes the problem of low segmentation accuracy of the traditional method; it uses the two-time growth method based on the target linked list to find the target, while ensuring the accuracy, it suppresses the problem of target extraction. The false alarm phenomenon avoids the subsequent "false" target removal processing and improves the target extraction speed.

Description

technical field [0001] The invention relates to the image segmentation problem in the technical field of image processing, in particular to a fast target extraction method for synthetic aperture radar (Synthetic Aperture Radar, SAR) azimuth estimation. Background technique [0002] Synthetic aperture radar has the advantages of all-day, all-weather and strong penetrating ability, and has become an important means of military reconnaissance. In recent years, researches on Automatic Target Recognition (ATR) using high-resolution SAR images continue to emerge. [0003] The SAR target image is very sensitive to the azimuth of the radar imaging, and the images obtained by the same target at different azimuths are very different. In the traditional SAR ATR system, a large number of SAR template images of different azimuths are stored. By combining the target to be recognized with the template image Matching is performed to complete the identification. Therefore, pre-estimating t...

AI Technical Summary

Problems solved by technology

The minimum value method assumes that there are obvious peaks and valleys between the double peaks of the gray histogram of the target and the background, and it determines the threshold of segmentation by finding the minimum value of the curve; the minimum error method assumes that the probability density function of the target and the background is a certain Distribution (such as normal distribution), it determines the segmentation threshold by minimizing the probability of misjudgment; the maximum between-class variance method determines the threshold by maximizing the variance between the two classes of segmentation, the calculation is simple and effective, but its fatal flaw When the gray level difference between the target and the background is not obvious, serious misjudgment will occur; the constant false alarm method is a method often used in practical applications. It uses the statistical characteristics of pixels to ensure a constant false alarm rate. To estimate the threshold, in the case of noise interference, false alarms and missing alarms will be generated, so that the segmented target contains many non-target points, and the constant false alarm method is also very sensitive to the size of the target

In short, although the threshold method has good timeliness, but because the SAR image is polluted by coherent speckle noise, the current threshold method's assumptions about the gray distribution of the target and the background are not necessarily true, and its segmentation accuracy is often not high; in addition, the traditional threshold method After determining the threshold, the method takes all pixels whose gray value is greater than the threshold as the target

Due to the existence of noise, there are often a few pixels outside the target whose gray value is greater than the threshold, and these non-target points will be detected as targets. In order to remove these "pseudo" targets, the traditional threshold method often requires subsequent processing to truly Complete target extraction, which also increases the computational complexity of target extraction

Images