| The study of sea ice drift in the Bohai Sea is crucial for economic development,sea ice disaster monitoring and global change.The revisiting period of polar-orbiting satellites and sun-synchronous orbit satellites is long,and the timeliness of monitoring sea ice drift in the Bohai Sea is insufficient.GF-4 is China's first optical remote sensing satellite in geostationary orbit.Its unique advantages of short imaging time interval(20 seconds)and high resolution(50 meters)provide better data support for monitoring sea ice drift with rapid dynamic changes.Based on GF-4 optical remote sensing satellite image,this paper carried out the monitoring research on the Bohai Sea ice drift,and the main work is as follows:(1)An analysis of the optimal GF-4 band combination for improving sea ice/seawater contrast was performed.There are five bands in the GF-4 image.Based on the interclass separability,the three methods of Euclidean distance,Pakistan distance and mean distance are used as the metrics to select the spectrum that can maximize the sea ice/seawater contrast.The combined results of the screening were verified by imaging expert interpretation and ice-water variance comparison.Experiments show that the optimal band combination of this paper can significantly improve the sea ice/seawater contrast.(2)The sea ice drift detection method based on the combination of SIFT and NCC is realized.In this paper,the principles of two methods are compared and analyzed,and the corresponding sea ice drift detection experiments are carried out respectively.The SIFT algorithm has the advantages of large number of feature point extraction and strong anti-noise interference,while the NCC algorithm has the advantage of high matching accuracy.Therefore,this paper combines the advantages of the two algorithms to extract feature points based on SIFT algorithm.On this basis,NCC algorithm is used for matching.Based on the combined algorithm in this paper,the number,speed and accuracy of extracting sea ice drift points have been significantly improved.The experimental data set was made by visual interpretation.Based on the data set,the developed method in this paper was compared and verified with the two classical methods of SIFT and NCC.Compared with the SIFT algorithm,the accuracy of feature point matching is improved from about 25%to about 55%.Compared with the NCC algorithm,the number of sea ice drift points correctly identified by the method in this paper is about 3 times of the former.(3)The positioning accuracy of the same name point of the GF-4 satellite in the Bohai region was evaluated.The experiments in this paper show that the direction and speed of sea ice drift based on GF-4 extraction are greatly affected by the accuracy of the data registration.In order to accurately analyze its influence,this paper carried out a study on the positioning accuracy evaluation of homonymic points.In this paper,the sea area of Liaodong Bay,the main ice area of Bohai Sea,was divided into three regions:the bottom of the bay,the east coast and the west coast.In the three regions,ground markers with time intervals of 1 minute,3 hours,4 hours and 24 hours were selected,and their spatial deviations,total deviations and directions were counted.Based on the statistical information,the offset direction and frequency of the same name point were analyzed by rose graph,the total offset of the same name point was compared by histogram,and the relationship between the spatial offset of the same name point and the length of the time interval was analyzed.The evaluation results of this paper show that:the deviation directions of different bay areas are inconsistent for data with time interval less than 3 hours.For data with time interval more than 4 hours,the main offset directions of the three bay areas are all due east;the land migration in the three areas of Liaodong Bay is mainly from 60 to 80 meters,with the maximum is no more than 100 meters.Most are offset by 1.2 pixels,up to maximum of no more than 2 pixels. |
PDF Full Text Request