Wide area analysis of video image for ground truth of satellite image

测量南极海冰的季节周期和年际变化对全球气候和船舶航行的调查具有重要意义。虽然卫星数据提供了大量关于冰况的信息，但仍然需要实地验证。地面真值参考数据的分辨率远高于NOAA卫星数据。为了将地面真值数据与卫星数据进行比较，需要对地面数据进行连续的补片，从而扩大地面真值数据的分析范围；两者都需要进行定量分析。在视频图像分析中，提出了一种利用船舶视频图像测量大范围海冰特征的方法。船上的摄像机拍下了海冰。通过几何变换和模板匹配得到连续视频图像。可以连续地获得冰的大小和沿船路线的浓度。另一方面，来自气象卫星的更多观测提供了了解极地表面状况的最佳手段，因为它们在很大范围内都是均匀的。然而，在极地地区，云、雪和冰在可见光通道中具有几乎相同的反照率，在红外通道中具有相同的亮温。因此，仅使用卫星图像的灰度阈值很难区分这些区域。本文提出了一种对云层、海冰和地面进行分类的方法。本研究基于对NOAA/AVHRR在南极洲的红外图像的分析。该算法包括两个主要的方法：图像特征提取和分类算法。最小距离分类器使用5个图像特征将该区域分为3个类别之一。为了提高分类精度，改变最小距离分类器的阈值边界。分类区和误分类区均随阈值水平的增加而减小。一般来说，错误率将根据目的而定。利用视频图像分析海冰特征和利用卫星数据提取海冰面积这两种方法都有助于检验这两种数据的相关性。少

Measurement of the seasonal cycle and interannual variations of Antarctic sea ice is important for investigations of global climate and ship navigation. While satellite data give a great amount of information about ice conditions, there still remains a need for in situ validations. Resolution of ground truth reference data is much higher than NOAA satellite data. In order to compare the ground truth data with satellite data, analyzing area of ground truth data have to be expanded by patching the ground based data continuously ; and both need to be analyzed quantitatively.In the video image analysis, a technique for measuring sea ice characteristics over a wide area using video images taken from a ship is proposed. The sea ice was photographed by video camera from the ship. Continuous video images are obtained using geometric transformation and template matching. Both size of the ice and concentration along the ship's route can be obtained continuously.The other hand, remote sensing obs … More ervation from a meteorological satellite offers the best available means to understand polar surface conditions, because of their homogeneity over a wide area. However, in the polar region, cloud, snow and ice have almost the same albedo in the visible channel and the same brightness temperature in the infrared channel. Therefore, it is difficult to distinguish among these regions using only the threshold of gray level of a satellite image. In this work, a method to classify clud, sea ice and ground is proposed. This study is based upon analysis of the NOAA/AVHRR infrared images in Antarctica. The algorithm consists of two major approaches : extraction of image features and a classification algorithm. Minimum distance classifier was used to classify that region into one of three categories using five image features. To improve the classification accuracy, threshold boundaries for minimum distance classifier were changed. Both classified and misclassfied areas were decreased with increasing the threshold levels. In geneal, the error rates will be decided according to the purpose.Since both methods of anarysis of sea ice characteristics using video image and extraction of sea ice area using satellite data, it is useful for exanining tke correlation of these two data. Less