%0 Journal Article
%T 基于无人机遥感的海洋养殖区识别研究<br/>Mariculture Areas Recognition Based on UAV Remote Sensing
%A 连鹏
%A 刘阳
%A 何晓晴
%A 景有甫
%A 黄聪颖
%A 董浙燚
%A 宋旭辉
%A 徐安康
%J Open Journal of Fisheries Research
%P 179-188
%@ 2373-1451
%D 2019
%I Hans Publishing
%R 10.12677/OJFR.2019.64024
%X 无人机遥感技术已日趋成为地理信息系统获取数据的重要工具，它以高精度、高灵活性和快速提取多维特征的优点而广泛应用于规划设计、风险评估和整体优化等领域中。本文利用无人机遥感以及基于机器学习的图像识别技术对山东省青岛市灵山岛海洋养殖区域及其沿海居民区进行了地物识别。同时，使用基于支持向量机(SVM)的方法对沿海各个功能区域进行了目标空间信息计算并得到了该区域的各项特征要素参数。本文基于监督分类，使用35%的样本像素作为训练集，测试了在不同光谱和空间条件下模型模拟效果的稳健性。结果表明：对于小样本容量，快速进行地物识别分析的要求下，SVM表现出较高的判准性(可达76.56%)，但对小样品而言，模型效果受训练集的影响较大。对于现场调查和结果验证，将来通过大数据可以进一步提高对物体识别的精度。<br />
UAV remote sensing technology has become an increasingly important tool for GIS to obtain data. It is widely applied in various fields such as planning and design, risk assessment and overall optimi-zation, with high accuracy, strong flexibility and rapid extraction of multi-dimensional features. In this paper, UAV remote sensing and image recognition technology based on machine learning are used to model mariculture areas along the coast of Lingshan Island in Qingdao, Shandong and to identify the marine aquaculture areas and coastal residents. In addition, the method based on sup-port vector machine is used to process spatial information of each functional area along the coast and consequently get the parameters of each feature element of the area. In this paper, based on the supervised classification, 35% of sample pixels are used as training set to test the robustness of the model simulation effect under different spectral details and spatial details. The results show that: for small sample size and the requirement of fast ground object identification and analysis, SVM shows a high accuracy (up to 76.56%), but for small samples, the model effect is greatly affect-ed by the training set. In the future, about field investigation and verification of results, combina-tion of big data could improve the object identification accuracy.
%K 无人机遥感，机器学习，海洋养殖区，支持向量机，监督分类<br/>UAV Remote Sensing
%K Machine Learning
%K Mariculture Areas
%K SVM
%K Supervised Classification
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=33402