基于Kirsch算子和多级阈值分割的河道轮廓识别和提取算法  

River Contour Recognition and Extraction Algorithm Based on Kirsch Operator and Multi-level Threshold Segmentation

作　　者：周少君 佘海龙 ZHOU Shaojun;SHE Hailong(Water Authority of Yugu Autonomous County of Sunan,Zhangye 734400,China)

机构地区：[1]肃南裕固族自治县水务局,甘肃张掖734400

出　　处：《人民珠江》2023年第7期109-117,共9页Pearl River

摘　　要：针对高分辨率遥感图像,在复杂背景下河道轮廓智能化提取精度低的问题。提出了采用Kirsch算子和阈值分割相结合的方式,对灰度图进行预处理和预分割。再依据河流的形态特征,对图像进行形态学处理,去除大部分背景噪声。最终对主要区域进行二次阈值分割,实现了准确的河道轮廓识别和提取。通过识别不同地形条件下的河流样本,证明了该方法的适用性。利用客观的评价指标,验证了河流区域的识别精度超过了90%,远高于简单的阈值分割方法。这满足了高精度河道提取需求,为智能化快速提取河道信息提供了参考。The intelligent extraction accuracy of river contour from high-resolution remote sensing images under complex backgrounds is low.Thus,a method of combining the Kirsch operator with threshold segmentation is proposed to preprocess and pre-segment the gray images.Then,according to the morphological characteristics of the river,the image is morphologically processed to remove most of the background noise.Finally,the second threshold segmentation is conducted on the main area to realize accurate river contour recognition and extraction.The applicability of this method is proven by recognizing river samples in different terrain conditions.By employing objective evaluation indexes,the recognition accuracy of river area is verified to exceed 90%,which is much higher than the simple threshold segmentation method.This meets the needs of high-precision river channel extraction and provides references for intelligent and rapid extraction of river channel information.

关 键 词：河道提取 KIRSCH算子 多级阈值分割 形态学处理 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]