基于图像处理的红外云图地震预测算法研究

发布时间：2018-03-13 00:03

本文选题：地震预测　切入点：热红外遥感　出处：《华中科技大学》2016年硕士论文　论文类型：学位论文

【摘要】：全球自然灾害的发生数量逐步增长,每年有近2亿人深受此类灾难。我国自然灾害类型多,分布不均匀,70%以上的人口、80%以上的城市和工农业严重遭受自然灾害的威胁,因此开展地震预测的研究十分迫切。在过去的几十年里,虽然使用遥感图像预测地震已经取得了一些成就,但传统的预测方法存在一定的局限性,一是无法准确预测震中位置,二是都是人工或者半人工的实现预测。为了解决这两个问题,本文提出了一种跟踪异常云团出现位置和频率的热红外异常云识别地震预测方法。根据热红外异常云团地震预测理论,该方法主要分为热红外异常云的识别和跟踪两个部分。对于识别部分,分为样本训练和分类识别两个步骤。首先训练样本,选择确定的异常云作为正样本,非异常云作为负样本,计算其纹理特征向量,将纹理特征向量作为输入,是否是异常云作为输出,训练得到异常云的神经网络分类器。其次对每张输入云图做分类识别,先对输入热红外云图进行预处理,对可疑区域进行图像增强,并过滤掉非云区,再计算出云图中每个点周围的纹理特征向量并作为分类器的输入来对每个像素点进行分类,将分类结果聚类并过滤后提取出疑似异常云区域。对于跟踪部分,通过跟踪一段时间内的热红外云图,如果某个区域异常云复现频率较高,则可以认为这个位置有发生地震的可能,并根据异常云团中心的演变位置估计地震震中位置。论文通过地震反演实验结果证明热红外异常云团识别算法可有效实现自动地震预测。该方法不仅能准确预测地震中心,而且对震级和发震时间也有一定的预测作用。
[Abstract]:The number of natural disasters in the world is increasing step by step, and nearly 200 million people are affected by such disasters every year. In China, there are many types of natural disasters, and more than 80% of the population with uneven distribution are seriously threatened by natural disasters. Therefore, it is very urgent to carry out the research of earthquake prediction. In the past few decades, although some achievements have been made in using remote sensing images to predict earthquakes, there are some limitations in the traditional prediction methods. One is that the epicenter location cannot be accurately predicted. Second, both are artificial or semi-artificial predictions. In order to solve these two problems, In this paper, a method of seismic prediction based on thermal infrared anomaly cloud identification is proposed, which can track the location and frequency of abnormal cloud cluster, according to the theory of earthquake prediction of thermal infrared abnormal cloud cluster. The method is mainly divided into two parts: recognition and tracking of thermal infrared anomaly cloud. For the recognition part, it is divided into two steps: sample training and classification recognition. First, the sample is trained, and the determined abnormal cloud is selected as positive sample. The non-abnormal cloud is used as a negative sample to calculate its texture feature vector, the texture feature vector is taken as input, and whether the abnormal cloud is output is trained to obtain the neural network classifier of abnormal cloud. Secondly, every input cloud image is classified and recognized. The input thermal infrared cloud image is preprocessed, the suspicious area is enhanced, and the non-cloud region is filtered out. Then the texture feature vector around each point in the cloud image is calculated and classified as the input of the classifier. Clustering the classification results and filtering to extract the suspected abnormal cloud area. For the tracking part, by tracking the thermal infrared cloud image for a period of time, if the frequency of abnormal cloud reappearance in a region is higher, Then it can be considered that there is a possibility of an earthquake occurring in this location. The seismic epicenter location is estimated according to the evolution position of abnormal cloud cluster center. The experimental results of seismic inversion prove that the thermal infrared anomaly cloud cluster identification algorithm can effectively realize automatic earthquake prediction, and this method can not only accurately predict the seismic center, but also predict the seismic center accurately. Moreover, it can predict the magnitude and the time of earthquake occurrence.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P315.7;TP391.41

【参考文献】