基于改进Canny算子的声呐图像边缘检测

Sonar Image Edge Detection Based on Improved Canny Operator

摘要: 目标的识别与解译是声呐图像处理中的一项重要任务。受到海底混响噪声和复杂背景的干扰，声呐图像有对比度差、分辨率低、噪声严重的问题，给声呐图像的目标识别带来了显著的挑战。为此，本文提出了一种改进的Canny算子边缘检测方法。通过LEE滤波器去除噪声，其次用OTSU阈值法划分高低阈值，最后运用Laplacian算子检测离群点并将其去除。本文将该方法命名为Canny-Laplacian算子(C-L算子)。本文使用多种边缘检测方法与改进的方法对真实声呐图像进行了处理，并在连续性、完整性和准确性上进行了对比。本文使用多种评价指标进行定量分析，证明了本文方法在声呐图像的边缘检测上有更好的效果。

Abstract: Target recognition and interpretation is an important task in sonar image processing. Due to the interference of underwater reverberation noise and complex backgrounds, sonar images suffer from poor contrast, low resolution, and severe noise, posing significant challenges to target recognition in sonar images. Therefore, this article proposes an improved Canny operator edge detection method. Remove noise through LEE filters, then divide high and low thresholds using OTSU thresholding method, and finally use Laplacian operator to detect outliers and remove them. This article names the method Canny Laplacian operator (C-L operator). This article uses multiple edge detection methods and improved methods to process real sonar images, and compares them in terms of continuity, integrity, and accuracy. This article uses multiple evaluation indicators for quantitative analysis, proving that our method has better performance in edge detection of sonar images.