基于主动轮廓模型的海洋航空遥感图像的边缘检测模型和最优化策略研究

项目名称： 基于主动轮廓模型的海洋航空遥感图像的边缘检测模型和最优化策略研究

项目编号： No.61501082

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 景雨

作者单位： 大连外国语大学

项目金额： 19万元

中文摘要： 海洋航空遥感图像的边缘检测技术是最关键的海洋污染监测技术之一，是辨别溢油、海冰和赤潮等海洋灾害存在、估算污染数量的前期工作。基于主动轮廓模型的边缘检测算法目前已经成为图像处理领域中一个重要的研究方向。针对海洋航空遥感图像中存在的噪声、低对比度、边界模糊和灰度不均匀性等问题，本项目对其边缘检测新的研究思路进行探索，即首先综合考虑图像的局部区域统计信息、边缘梯度信息、几何特征、凸性约束限定等问题，建立基于局部高斯拟合和灰度不均匀性纠正的遥感图像快速边缘检测模型，并从理论上借用了Chan的全局最小化优化思想来实现凸性转化。然后提出基于Split Bregman的最优化策略，该策略较之传统的梯度下降和加权全变差最优化方法，使边缘检测算法实现更简单，曲线收敛更快更稳定，边缘检测结果更准确，并采用机器学习的方法实现局部区域控制参数的自动选取，提高边缘检测算法的鲁棒性。

中文关键词： 海洋；遥感图像；边缘检测；主动轮廓模型

英文摘要： The technology of edge detection is an important tool for the location and acreage calculation of oil slick, sea ice and red tide on the sea by aerial remote sensing. Whenever we need to identify them, confirm the location or get the shape and acreage of them, we have to get the edge information of images firstly. Edge detection algorithm based on active contour model becomes one of hot topics in image processing. In view of high noise, blurring and intensity inhomogeneity of oil slick infrared aerial images, this project developes a new edge detection algorithm.Firstly, according to the local region statistics information, edge intensity information,geometrical characteristic and convex constraint criterias, a new edge detection model for ocean aerial IR image will be proposed based on the local gassian fitting and intensity inhomogeinty correction. Theoretically convex transformation is realized by the global optimization idea of Chan.Then, The optimization strategy based on Split Bregman will be proposed and compared with the traditional gradient descent and weighted total variation optimization methods, the proposed strategy make edge detection more simple, make curve converge faster and more stable, and make edge detection more accturate.After that,the selection of local region control parameter can be optimized by machine learning to improve the robustness of edge detection algorithm. At last, the efficiency of the algorithm are analyzed and evaluated.

英文关键词： ocean ;remote sensing image ;edge detection ;active contour model