| 本文已被：浏览 674次 下载 847次 |
码上扫一扫！ |
|
|
| 基于加速鲁棒特征图像匹配的云导风计算方法 |
|
孔德华1, 张东1,2, 张卓2,3, 宋志尧2,3,4
|
|
1.南京师范大学 海洋科学与工程学院, 江苏 南京 210023;2.江苏省地理信息资源开发与利用协同创新中心, 江苏 南京 210023;3.南京师范大学 虚拟地理环境教育部重点实验室, 江苏 南京 210023;4.大规模复杂系统数值模拟江苏省重点实验室, 江苏 南京 210023
|
|
| 摘要: |
| 利用云导风技术结合高分辨率气象卫星遥感数据获取风矢量,在监测台风等极端气象灾害方面具有重要应用。本文提出了一种基于加速鲁棒特征(speeded up robust features,SURF)图像匹配的云导风计算方法,利用SURF算法结合随机抽样一致算法(random sample consensus,RANSAC),提取并匹配两景连续时序云图的特征点,计算风矢量,并结合当地大气温度廓线指定云高,经质量控制得到云导风矢量。运用该方法模拟了2018年台风“山竹”的云导风矢量,以美国威斯康星大学气象卫星研究合作所(CIMSS)的大气运动矢量资料进行验证,结果表明:(1)风速和风向的相关系数分别为0.78和0.79,均方根误差分别为4.75 m·s–1和37.64°,平均绝对百分比误差分别为33.49%和22.55%,整体具有良好的模拟精度;(2)与CIMSS资料相比,基于特征点匹配的SURF云导风计算方法在反演密集云区的风矢量有明显优势,可有效提高云区内风矢量的数量,扩大风矢量的空间覆盖范围;(3)图像对比度增强处理对特征点的提取和风矢量的空间分布有重要影响,伽马变换因子γ=5时,能较好地平衡台风外围螺旋云带和中心附近云区的风矢量数量,反映台风风场的整体特征。该方法作为基于尺度不变特征变换的云导风计算方法的改进,可为利用卫星遥感影像数据进行云导风计算提供新的思路。 |
| 关键词: 加速鲁棒特征算法 图像匹配 云导风 风矢量 台风 |
| DOI：10.11759/hykx20211113001 |
| 分类号:TP751 |
| 基金项目:国家重点研发计划项目(2018YFB0505500,2018YFB0505502) |
|
| A cloud-motion-based wind retrieval method based on the SURF algorithm |
|
KONG De-hua1, ZHANG Dong1,2, ZHANG Zhuo2,3, SONG Zhi-yao2,3,4
|
|
1.College of Marine Science and Engineering, Nanjing Normal University, Nanjing 210023, China;2.Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China;3.Key Lab of Virtual Geographic Environment under the Ministry of Education, Nanjing Normal University, Nanjing 210023, China;4.Jiangsu Provincial Key Laboratory for Numerical Simulation of Large Scale Complex System, Nanjing 210023, China
|
| Abstract: |
| Cloud-motion-based wind retrieval technology combined with high-resolution meteorological satellite remote sensing data to obtain wind vectors has crucial applications in monitoring extreme meteorological disasters such as typhoons. In this study, a cloud-motion-based wind retrieval method based on the speeded-up robust features (SURF) image matching algorithm is proposed. The SURF and random sample consensus algorithms are used to extract and match the feature points of two consecutive time-series cloud images, calculate the wind vectors, and specify cloud height in combination with the local atmospheric temperature profile. The cloud-motion-based wind retrieval vectors are obtained through quality control. The proposed method is used to simulate the cloud- motion-based wind retrieval vectors of Typhoon “Mangkhut” in 2018, which is verified using the Cooperative Institute for Meteorological Satellite Studies (CIMSS) atmospheric motion vector data. The results indicate that: (1) The correlation coefficients of wind speed and wind direction are 0.78 and 0.79, respectively, the root mean square errors are 4.75 m·s−1 and 37.64°, respectively, and the average absolute percentage errors are 33.49% and 22.55%, respectively, which has good simulation accuracy overall. (2) Compared with the CIMSS data, the cloud-motion- based wind retrieval method based on the SURF feature matching algorithm has obvious advantages in retrieving wind vectors in dense cloud areas, which can effectively improve the number of wind vectors in cloud areas and expand the spatial coverage of wind vectors. (3) Image contrast enhancement highly impacts the extraction of feature points and the spatial distribution of wind vector γ = 5, which can better balance the number of wind vectors in the spiral cloud belt and the area near the center of the typhoon, reflecting the overall characteristics of the typhoon wind field. As an improvement of the cloud-motion-based wind retrieval method based on scale- invariant feature transformation, the proposed method can be a new approach for wind vector calculation using satellite remote sensing image data. |
| Key words: SURF algorithm image matching cloud-motion-based wind retrieval wind vectors typhoon |