摘要
针对目前海上船舶油井提取多是使用已有的非实时陆地岸线提取海域,并且提取算法缺少在大尺度影像上搜索和查找可能存在目标能力的问题,提出一种基于可见光遥感数据的船舶油井检测策略。该策略主要包括综合形态学运算提取海域、目标有无判定算法、迭代最优阈值分割(TS)滑动窗口(SW)目标提取三个部分。探讨了目标有无判定算法中的参数设置和滑动窗口的大小设置,并将提取结果与人工目视解译结果进行了交叉对比验证。结果表明,该策略通过设置合理的参数,可使目标提取的真实精度达到0.981,相对精度达到0.954,表现出较高的实用性。
To improve situations as follows： in the strategies of offshore ship and well platform detection, most masks of sea zones use non-real time shoreline database; targets seeking algorithms lack capability of searching targets in large scale images, a strategy of ship and well platform detection based on optical remote sensing images was proposed. The strategy included building masks of sea zones using morphological operations, determining decision algorithms of targets＇ existence, and extracting targets＇ locations based on iterative optimal Threshold Segmentation （TS） in Sliding Windows （SW）. Parameters in the decision algorithm and the size of sliding window were analyzed, and the corresponding results were cross validated with that of artificial visual interpretation. The experimental results prove that the absolute accuracy of targets extraction arrives 0. 981 while relative accuracy of targets extraction arrives 0. 954 with proper parameters set. This strategy shows practical value.
出处
《计算机应用》
CSCD
北大核心
2013年第3期708-711,共4页
journal of Computer Applications
基金
国家海洋局海洋溢油鉴别与损害评估技术重点实验室开放基金资助项目(201208)
关键词
船舶油井检测
形态学运算
目标有无判定算法
迭代最优阈值分割
滑动窗口
ship and well platform detection
morphological operation
decision algorithm of target existence
iterativeoptimal Threshold Segmentation （TS）
Sliding Window （SW）
