基于遥感影像的近海岸水产提取方法研究

发布时间：2018-07-01 12:11

本文选题：近海养殖区提取 + 基于正交子空间投影的约束能量最小化(OWCEM)　；参考：《中国科学院大学(中国科学院遥感与数字地球研究所)》2017年硕士论文

【摘要】：随着近海水产养殖业的发展,近海岸环境正面临着水体富营养化、原始生态群落被破坏等威胁。日益增长的水产养殖活动在追求经济利益中如何考虑到当地环境的承载力是养殖户和当地管理部门需要解决的问题。而利用遥感卫星影像对近海养殖区进行观测,是规划和管理近海水产养殖区的有效方式之一。不过传统的水产检测算法通常利用专业的解译人员配合遥感软件对影像进行人工解译,提取效率低,且人工成本较高,难以满足近海水产养殖区快速观测的要求。本文在研究近海水产养殖区在遥感影像上的光谱和空间特征,讨论了利用短重返周期、低数据成本的中等分辨率卫星提取近海水产养殖区的可行性,此外还对现有的水产提取算法进行了总结,并结合算法原理分析其优缺点和适用范围,对已有算法中存在的问题进行总结,并提出了基于改进约束能量最小化的水产自适应分割提取算法。算法利用Landsat 8 OLI中等分辨率卫星数据来进行水产提取方法的研究,不仅在最终的提取精度上高于传统的检测算法,而且整个提取流程的自动化程度高,减轻了传统影像解译人员的工作负担。相比于以往的水产检测方法仅能够识别出水产养殖区的大致范围,本文充分利用水产养殖区的光谱和空间特征,将养殖水域养殖带之间的非养殖像元进行剔除,大大降低了算法错分概率,提高了整体算法的提取精度,能有向有关部门和专家提供更为准确、有效的近海养殖信息。主要的研究工作如下:1、本章对现阶段国内外水产提取的方法进行了归纳总结,分别是目视解译法,特征指数提取法,面向对象分割提取法,无人机影像解译法和SAR影像解译法。通过分析不同提取方法间的算法原理,说明了各方法在水产养殖提取中的优缺点和主要解决的问题。除了介绍算法各自的原理步骤之外,还对常用几种算法的适用性和局限性进行了总结评述。2、针对以往水产提取算法误检率高、普适性较低等问题,本文提出了一种基于改进约束能量最小化的水产自适应分割提取算法(OWCEM-SDAS)。本文算法利通过多个波段组合的方式,构建了包含常用的水体指数MNDWI、MAWEI在内的多种指数,并结合叶绿素浓度和悬浮泥沙浓度指数使用正交加权约束能量最小化算法(OWCEM)对养殖区进行增强,然后通过原始影像分块,在每个子区域(块)中使用自适应阈值对水产养殖区进行分割提取,最后将各子区域(块)的提取结果综合,从而得到近海养殖区最终的水产提取。3、通过不同近海养殖区不同提取方法间的对比,利用混淆矩阵和kappa系数等指标量化提取结果,证明了本文算法在不同实验区复杂背景下的提取精度和稳定性都好于其他常用提取算法。
[Abstract]:With the development of offshore aquaculture, the coastal environment is facing the threat of eutrophication and destruction of primitive ecological communities. How to consider the bearing capacity of local environment in the pursuit of economic benefits in the growing aquaculture activities is a problem that farmers and local management departments need to solve. It is one of the effective ways to plan and manage the offshore aquaculture area by using remote sensing satellite image. However, the traditional fish detection algorithms usually use professional interpreters with remote sensing software for artificial interpretation of images, the extraction efficiency is low, and the labor cost is high, it is difficult to meet the requirements of rapid observation in offshore aquaculture areas. In this paper, the spectral and spatial characteristics of remote sensing images of offshore aquaculture area are studied, and the feasibility of extracting offshore aquaculture area by using medium resolution satellite with short re-entry period and low data cost is discussed. In addition, the existing aquatic extraction algorithms are summarized, and the advantages, disadvantages and applicable scope of the algorithms are analyzed, and the existing problems in the existing algorithms are summarized. An adaptive extraction algorithm based on improved constrained energy minimization is proposed. The algorithm uses Landsat 8 Oli medium resolution satellite data to carry on the aquatic product extraction method research, not only in the final extraction precision is higher than the traditional detection algorithm, but also the whole extraction process automation degree is high. It lightens the workload of the traditional image interpreter. Compared with the previous methods of aquaculture detection, we can only identify the approximate range of aquaculture areas. In this paper, we take full advantage of the spectral and spatial characteristics of aquaculture zones, and eliminate the non-culture pixels between the aquaculture zones. The algorithm greatly reduces the probability of misdivision, improves the accuracy of the whole algorithm, and can provide more accurate and effective offshore aquaculture information to the relevant departments and experts. The main research work is as follows: 1. In this chapter, the methods of aquatic extraction at home and abroad are summarized, including visual interpretation, feature index extraction, object oriented segmentation, UAV image interpretation and SAR image interpretation. By analyzing the principle of different extraction methods, the advantages and disadvantages of each method in aquiculture extraction and the main problems solved were explained. In addition to introducing the principle and steps of the algorithms, the applicability and limitation of several commonly used algorithms are summarized and commented. In view of the problems of high misdetection rate and low universality of the previous aquatic extraction algorithms, In this paper, an improved constrained energy minimization algorithm (OWCEM-SDAS) is proposed. In this paper, we construct a variety of indexes including the commonly used water index MNDWI and MAWEI by means of multi-band combination. Combined with chlorophyll concentration and suspended sediment concentration index, orthogonal weighted constrained energy minimization algorithm (OWCEM) is used to enhance the culture area, and then the original image is divided into blocks. In each sub-region (block), the adaptive threshold is used to segment and extract the aquaculture area. Finally, the extraction results of each sub-region (block) are synthesized. In this way, the final aquatic extraction. 3 in the offshore culture area was obtained. By comparing the different extraction methods in different offshore culture areas, the results were quantified by using the confusion matrix and kappa coefficient, and so on. It is proved that the extraction accuracy and stability of this algorithm are better than those of other common algorithms in different experimental areas.
【学位授予单位】：中国科学院大学(中国科学院遥感与数字地球研究所)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S951.4;TP751

【参考文献】