A fast underwater target extraction method based on side-scan sonar imaging

Abstract

The present invention discloses an underwater target rapid extraction method for side-scan sonar imaging. The method can perform real-time analysis on a side-scan sonar image and quickly extract an underwater target. According to the method, adaptive non-linear complex diffusion model filtering is performed on a side-scan sonar image of which the multiplicative noises are prominent, so that a noise problem in the side-scan sonar image can be improved; the de-noised sonar image is pre-segmented, and the target region, seabed reverberation region and shade region of the side-scan sonar image are automatically determined; the contours of the pre-segmentation results are redefined, so that time for constantly re-defining an initial level set in level set model evolution can be decreased in a segmentation process; a variation level set model-based evolution and segmentation are performed on the target region and the shade region; and binarization is performed on a target to be extracted, clear segmentation results of the three regions are displayed for an operator. According to the underwater target rapid extraction method for the side-scan sonar imaging of the invention, target extraction is performed on a real-time side-scan sonar, and therefore, the method has the advantages of high detection real-time performance, high detection stability and high accuracy, provides a reliable guarantee for underwater detection and target tracking.

Description

technical field [0001] The invention relates to the technical field of sonar image processing, in particular to a method for quickly extracting underwater targets in side-scan sonar imaging. Background technique [0002] Because of its long detection distance and strong penetrating power, sonar imaging has been widely used in oceanographic mapping and underwater target detection and tracking. In the long-range detection of underwater targets, use sonar equipment to scan and image, extract target highlight areas and shadow areas in side-scan sonar images, and identify suspicious targets according to the area and size of the area and combined with device parameter settings and measurement. However, the interference of the reverberation signal, the influence of the ocean environment noise and the self-noise of the ship lead to prominent speckle noise and blurred target edges in the sonar image obtained by imaging, which brings great difficulties to target segmentation and dete...

AI Technical Summary

Problems solved by technology

However, the interference of the reverberation signal, the influence of the ocean environment noise and the ship's self-noise lead to prominent speckle noise and blurred target edges in the sonar image obtained by imaging, which brings great difficulties to target segmentation and detection.

In addition, as the key to target recognition and measurement, existing segmentation algorithms often cannot balance speed and accuracy.

The segmentation speed based on threshold, mathematical morphology, and clustering is fast, but it is prone to noise interference and large errors; the segmentation based on Markov random field and statistics has high accuracy, but requires prior knowledge, that is, the size and shape of the target and other features; the segmentation based on the geometric active contour model can extract edges accurately, but the contour evolution speed is slow

Images