基于SR-NDVI和中值滤波的连云港海域紫菜筏架自动提取研究

Research on the Automatic Extraction of Floating Raft Aquaculture Based on SR-NDVI and Median Filtering in Lianyungang

快速准确地掌握紫菜筏架的个数和面积可为海域管理和海洋行政执法等提供决策支持。本文以连云港市紫菜养殖筏架为研究对象,利用Landsat-8卫星影像,采用分区域归一化植被指数(Sub-regional Normalized Difference Vegetation Index,SR-NDVI)和中值滤波(分别利用蓝光、绿光和红光波段)两种方法,开展紫菜筏架自动提取研究,并以无人机影像对实验结果开展精度验证,针对不同水深区域,从识别个数、面积和方法 3个角度进行比对分析。研究结果表明:浅水区SR-NDVI方法识别面积精度优于中值滤波方法,深水区恰恰相反;深水区水质稳定、受环境因素影响较小,提取精度较高,4种实验识别紫菜筏架误差均优于2个筏架,而浅水区提取精度相对较差;同时浅水区还存在较为严重的"筏架相连"现象。

The information of number and area of the laver rafts should be acquired quickly and accurately to provide decision-making support for sea use management and marine administrative law enforcement.In this paper,the number and area of the laver rafts in Lianyungang City are automatically extracted from Landsat-8 satellite imagery using the methods of Sub-regional Normalized Difference Vegetation Index(SR-NDVI)and median filtering on respective blue,green,and red images.Then the performance of the methods in different water depth is evaluated based on the UAV images.The results show that for extracting area,the performance of the SR-NDVI method is better than the median filtering method in shallow waters,and the opposite is true in deep waters.For the number extraction,the accuracy is relatively high(with error less than 2)in deep waters where the water quality is stable,and less affected by the environmental factors,whereas it is relatively poor in shallow waters where have a much serious phenomenon of"raft-frame connection".