Research On Sea Ice Drift Detection Technology Based On Combination Of Feature Tracking And Pattern Matching

The detection of sea ice drift can not only predict the dynamic changes of global climate,but also prevent the occurrence of accidents such as sea ice crushing and destroying ships and oil platforms.Synthetic Aperture Radar(SAR)is widely used in sea ice drift detection because it is not affected by weather factors such as illumination and clouds and can perform satellite imaging all-day and all-weather.Because of the influence of incident angle,sea clutter,weather and other environmental factors on the SAR sea ice image,the brightness of two SAR sea ice images taken at different times in the same place is different,and the characteristics of the same sea ice in two SAR images are different due to the factors of sea ice deformation and rotation.More importantly,SAR-based imaging mechanism makes the same type of sea ice have similar features in SAR images.The above reasons greatly increase the difficulty of sea ice drift detection.Therefore,this paper has conducted a thorough study of SAR sea ice drift detection methods,and completed the following work:Firstly,the characteristics of common sea ice drift algorithms are compared and analyzed.Aiming at the problems that Sentinel-1 images are affected by incident angle,noise and environmental factors in the imaging process,a simulation experiment is carried out on the feature tracking method.The current common feature tracking algorithms are compared and analyzed from the aspects of image brightness,sea ice drift rotation and scale transformation,which provides a basis for selecting the suitable SAR sea ice drift detection method.At the same time,the characteristics and applications of common pattern matching methods are listed in theory,which provides a theoretical basis for the combination of feature tracking and pattern matching algorithm below.In order to solve the problem of high detection error rate of sea ice drift caused by the similarity of multiple areas in SAR sea ice images,a method that eliminates mismatches is proposed.This method is not only concise and efficient,but also has a higher matching accuracy than traditional ones,and can combine SIFT feature tracking method well,effectively improving the accuracy and timeliness of sea ice monitoring.Aiming at the problems of missing detection and error detection in the strong noise area of SAR image,the difference between HH and HV polarization data in sea ice drift detection is studied firstly,and the feature information detected by the two polarization data is effectively applied to sea ice drift detection by feature fusion method.The spatial distribution and coverage of sea ice drift vectors detected in this way are improved to a certain extent.In addition,a new algorithm based on triangulation is proposed,which combines feature tracking and pattern matching.It can be effectively applied to sea ice drift detection in high-resolution SAR images.From the comparisons between computational efficiency and matching accuracy,the algorithm can not only improve the speed of pattern matching detection,but also is not easily affected by scale transformation,illumination and angle.The detection results show that the proposed algorithm has higher robustness to image noise,which can further improve the detection accuracy and uniformity,and make up for the deficiency of the traditional feature tracking algorithm that the feature points in the local region are missing.In order to verify the performance of the combined algorithm of feature tracking and pattern matching,this method is compared with other classical methods of sea ice drift detection.The measured data show that the algorithm has high spatial coverage and uniformity in the application of sea ice drift detection in dual-polarized Sentinel-1 image.Compared with the pattern matching algorithm,the time of the algorithm is reduced by about 88%,and the detection is improved effectively.This paper also studies the applicability of the algorithm in the strong noise area of HH polarization and HV polarization data.The experimental results of different polarization data show that the sea ice drift vector obtained by this algorithm not only has higher coverage,but also reduces the root mean square error by about 10%.The detection accuracy is improved,and the robustness to noise is enhanced.Even under the interference of strip noise,the detection accuracy is still as high as 98%.It can be seen that the algorithm has universality for two polarization modes,which proves that the method can be effectively applied to sea ice drift monitoring.