论文部分内容阅读
海岸线变迁是沿海生态系统变化的重要指示因子,是国家海洋经济关注的重要组成部分。本文通过构建光谱角度—距离相似度模型,解决HJ-1B/IRS红外传感器在海岸线自动化提取应用中的可行性问题,以及当前方法应用于不同时相数据过程中的阈值不稳定性问题,拓展IRS传感器的应用领域和价值。光谱角度—距离相似度模型以多光谱像元归一化辐射值为向量元素,度量不同像元在单位空间距离上的角度相似性,以迭代方式分析水体样本像元与周边八邻域相邻像元的角度—距离相似性,通过相似性约束对水体样本进行区域生长以获取水岸分界线。通道辐射归一化分析表明采用反射率和量化等级最大值归一化的通道值能很好地反映地物随季节的变化;样本相似度分析表明以水体和非水体相似度两倍方差(0.01)为误差的生长阈值(0.98)适用于全时相影像水岸线提取,总体精度优于80%。验证数据分析表明,角度—距离相似度模型阈值稳定、不受时相影响。通过与常用的High Pass卷积滤波、Roberts卷积滤波、Sobel卷积滤波、Laplacian卷积滤波、FFT高通变换和Canny提取结果比对分析表明,High Pass、Laplacian和FFT变换无法应用于IRS传感器,Roberts和Sobel相对来说能较好的识别水陆边缘,Canny在正常噪声条件下也能有效识别水陆边缘。但这些算法在识别水陆边缘线的同时,也将内陆地物边缘线进行了识别,如何将内陆地物边缘线从识别结果中有效去除,是这些方面所面临的重要难点。比较而言,角—距相似度模型能很好的应用于IRS传感器的海岸线提取,对传感器的噪声并不敏感,在B4通道非正常水平噪声条件下也能提取出理想结果,而且后续处理简单,不存在内陆边缘线的干扰问题。光谱角度—距离相似度模型对海岸线识别精度较高、模型参数稳定,能有效地提升IRS传感器在海岸线提取方面应用价值。在实际应用过程中需要避免的是,既覆盖陆地又覆盖海洋的云团会遮挡地物光谱信息,造成海岸线无法有效分离,因此需要对影像数据进行有效的筛选。本文基于遥感影像提取的海岸线只是瞬时水边线,需要进一步结合海岸线的类型以及潮位数据和DEM等数据进行修正得到最终的海岸线。
Coastline changes are an important indicator of changes in coastal ecosystems and are an important part of the national marine economy. In this paper, the feasibility of HJ-1B / IRS infrared sensor in the coastline automatic extraction application is solved by constructing the spectral angle-distance similarity model, and the threshold instability problem of the current method applied to different time-phase data is extended. The IRS Sensor application areas and values. The spectral angle-distance similarity model uses the normalized radiation values of multi-spectral pixels as the vector elements to measure the angular similarity of different pixels in the unit space distance, and iteratively analyzes the relationship between the water sample pixels and the neighboring eight neighborhoods Pixel angle - distance similarity, through the similarity constraints of water samples for regional growth in order to obtain the waterfront demarcation line. The normalization of channel radiation shows that the normalized channel values reflect the changes of ground objects with the maximum of reflectance and quantification level. The analysis of sample similarity shows that the variance of water body and non-water body ) Is the growth threshold of error (0.98), which is suitable for waterfront extraction of full-time phase images with an overall accuracy of better than 80%. Verification data analysis shows that the angle-distance similarity model has a stable threshold and is independent of the time phase. Compared with the commonly used High Pass convolution filter, Roberts convolution filter, Sobel convolution filter, Laplacian convolution filter, FFT high-pass transform and Canny extraction results show that High Pass, Laplacian and FFT transform can not be applied to the IRS sensor, Roberts and Sobel are relatively good at identifying land and water edges, and Canny is also effective at identifying land and water edges under normal noise conditions. However, these algorithms recognize the marginal line of inland and outland while recognizing the marginal line of inland object, and how to effectively remove the inland edge line from the recognition result is an important and difficult point in these aspects. In contrast, the angular-distance similarity model can be applied to the coastline extraction of IRS sensors, which is not sensitive to the noise of the sensor. It can also extract the ideal results under the abnormal horizontal noise of the B4 channel, and the subsequent processing is simple , There is no interference of inland edge lines. Spectral angle-distance similarity model has higher coastline identification accuracy and stable model parameters, which can effectively enhance the application value of IRS sensor in coastline extraction. What needs to be avoided in the practical application is that the cloud covering both the land and the sea will obscure the spectral information of the terrestrial objects and cause the coastline to be effectively separated. Therefore, effective screening of the image data is required. In this paper, the coastline extracted based on remote sensing images is only an instantaneous waterline, and the final shoreline needs to be further corrected according to the type of coastline, tide level data and DEM data.